diff --git a/sys/netinet/sctp_cc_functions.c b/sys/netinet/sctp_cc_functions.c index 28bbe5a5ced5..ddd2fd85aea1 100644 --- a/sys/netinet/sctp_cc_functions.c +++ b/sys/netinet/sctp_cc_functions.c @@ -1,1728 +1,1712 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include __FBSDID("$FreeBSD$"); static void sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *assoc; uint32_t cwnd_in_mtu; assoc = &stcb->asoc; cwnd_in_mtu = SCTP_BASE_SYSCTL(sctp_initial_cwnd); if (cwnd_in_mtu == 0) { /* Using 0 means that the value of RFC 4960 is used. */ net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND)); } else { /* * We take the minimum of the burst limit and the initial * congestion window. */ if ((assoc->max_burst > 0) && (cwnd_in_mtu > assoc->max_burst)) cwnd_in_mtu = assoc->max_burst; net->cwnd = (net->mtu - sizeof(struct sctphdr)) * cwnd_in_mtu; } if (stcb->asoc.sctp_cmt_on_off == 2) { /* In case of resource pooling initialize appropriately */ net->cwnd /= assoc->numnets; if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) { net->cwnd = net->mtu - sizeof(struct sctphdr); } } net->ssthresh = assoc->peers_rwnd; SDT_PROBE(sctp, cwnd, net, init, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, 0, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION); } } static void sctp_cwnd_update_after_fr(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_nets *net; uint32_t t_ssthresh, t_cwnd; /* MT FIXME: Don't compute this over and over again */ t_ssthresh = 0; t_cwnd = 0; if (asoc->sctp_cmt_on_off == 2) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { t_ssthresh += net->ssthresh; t_cwnd += net->cwnd; } } /*- * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && * (net->fast_retran_loss_recovery == 0))) */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if ((asoc->fast_retran_loss_recovery == 0) || (asoc->sctp_cmt_on_off > 0)) { /* out of a RFC2582 Fast recovery window? */ if (net->net_ack > 0) { /* * per section 7.2.3, are there any * destinations that had a fast retransmit * to them. If so what we need to do is * adjust ssthresh and cwnd. */ struct sctp_tmit_chunk *lchk; int old_cwnd = net->cwnd; if (asoc->sctp_cmt_on_off == 2) { net->ssthresh = (uint32_t) (((uint64_t) 4 * (uint64_t) net->mtu * (uint64_t) net->ssthresh) / (uint64_t) t_ssthresh); if ((net->cwnd > t_cwnd / 2) && (net->ssthresh < net->cwnd - t_cwnd / 2)) { net->ssthresh = net->cwnd - t_cwnd / 2; } if (net->ssthresh < net->mtu) { net->ssthresh = net->mtu; } } else { net->ssthresh = net->cwnd / 2; if (net->ssthresh < (net->mtu * 2)) { net->ssthresh = 2 * net->mtu; } } net->cwnd = net->ssthresh; SDT_PROBE(sctp, cwnd, net, fr, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR); } lchk = TAILQ_FIRST(&asoc->send_queue); net->partial_bytes_acked = 0; /* Turn on fast recovery window */ asoc->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ asoc->fast_recovery_tsn = asoc->sending_seq - 1; } else { asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } /* * CMT fast recovery -- per destination * recovery variable. */ net->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ net->fast_recovery_tsn = asoc->sending_seq - 1; } else { net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } - /* - * Disable Nonce Sum Checking and store the - * resync tsn - */ - asoc->nonce_sum_check = 0; - asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1; - sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); } } else if (net->net_ack > 0) { /* * Mark a peg that we WOULD have done a cwnd * reduction but RFC2582 prevented this action. */ SCTP_STAT_INCR(sctps_fastretransinrtt); } } } static void sctp_cwnd_update_after_sack(struct sctp_tcb *stcb, struct sctp_association *asoc, int accum_moved, int reneged_all, int will_exit) { struct sctp_nets *net; int old_cwnd; uint32_t t_ssthresh, t_cwnd, incr; /* MT FIXME: Don't compute this over and over again */ t_ssthresh = 0; t_cwnd = 0; if (stcb->asoc.sctp_cmt_on_off == 2) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { t_ssthresh += net->ssthresh; t_cwnd += net->cwnd; } } /******************************/ /* update cwnd and Early FR */ /******************************/ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code. Need to debug. */ if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) { net->will_exit_fast_recovery = 1; } } #endif if (SCTP_BASE_SYSCTL(sctp_early_fr)) { /* * So, first of all do we need to have a Early FR * timer running? */ if ((!TAILQ_EMPTY(&asoc->sent_queue) && (net->ref_count > 1) && (net->flight_size < net->cwnd)) || (reneged_all)) { /* * yes, so in this case stop it if its * running, and then restart it. Reneging * all is a special case where we want to * run the Early FR timer and then force the * last few unacked to be sent, causing us * to illicit a sack with gaps to force out * the others. */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck2); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_20); } SCTP_STAT_INCR(sctps_earlyfrstrid); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net); } else { /* No, stop it if its running */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck3); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_21); } } } /* if nothing was acked on this destination skip it */ if (net->net_ack == 0) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); } continue; } if (net->net_ack2 > 0) { /* * Karn's rule applies to clearing error count, this * is optional. */ net->error_count = 0; if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) == SCTP_ADDR_NOT_REACHABLE) { /* addr came good */ net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE; net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED); /* now was it the primary? if so restore */ if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) { (void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net); } } /* * JRS 5/14/07 - If CMT PF is on and the destination * is in PF state, set the destination to active * state and set the cwnd to one or two MTU's based * on whether PF1 or PF2 is being used. * * Should we stop any running T3 timer here? */ if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { net->dest_state &= ~SCTP_ADDR_PF; old_cwnd = net->cwnd; net->cwnd = net->mtu * asoc->sctp_cmt_pf; SDT_PROBE(sctp, cwnd, net, ack, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n", net, net->cwnd); /* * Since the cwnd value is explicitly set, * skip the code that updates the cwnd * value. */ goto skip_cwnd_update; } } #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code */ /* * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery * && net->will_exit_fast_recovery == 0) { @@@ Do something * } else if (sctp_cmt_on_off == 0 && * asoc->fast_retran_loss_recovery && will_exit == 0) { */ #endif if (asoc->fast_retran_loss_recovery && (will_exit == 0) && (asoc->sctp_cmt_on_off == 0)) { /* * If we are in loss recovery we skip any cwnd * update */ goto skip_cwnd_update; } /* * CMT: CUC algorithm. Update cwnd if pseudo-cumack has * moved. */ if (accum_moved || ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { /* If the cumulative ack moved we can proceed */ if (net->cwnd <= net->ssthresh) { /* We are in slow start */ if (net->flight_size + net->net_ack >= net->cwnd) { old_cwnd = net->cwnd; if (stcb->asoc.sctp_cmt_on_off == 2) { uint32_t limit; limit = (uint32_t) (((uint64_t) net->mtu * (uint64_t) SCTP_BASE_SYSCTL(sctp_L2_abc_variable) * (uint64_t) net->ssthresh) / (uint64_t) t_ssthresh); incr = (uint32_t) (((uint64_t) net->net_ack * (uint64_t) net->ssthresh) / (uint64_t) t_ssthresh); if (incr > limit) { incr = limit; } if (incr == 0) { incr = 1; } } else { incr = net->net_ack; if (incr > net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable)) { incr = net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable); } } net->cwnd += incr; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, incr, SCTP_CWND_LOG_FROM_SS); } SDT_PROBE(sctp, cwnd, net, ack, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_SS); } } } else { /* We are in congestion avoidance */ uint32_t incr; /* * Add to pba */ net->partial_bytes_acked += net->net_ack; if ((net->flight_size + net->net_ack >= net->cwnd) && (net->partial_bytes_acked >= net->cwnd)) { net->partial_bytes_acked -= net->cwnd; old_cwnd = net->cwnd; if (asoc->sctp_cmt_on_off == 2) { incr = (uint32_t) (((uint64_t) net->mtu * (uint64_t) net->ssthresh) / (uint64_t) t_ssthresh); if (incr == 0) { incr = 1; } } else { incr = net->mtu; } net->cwnd += incr; SDT_PROBE(sctp, cwnd, net, ack, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_CA); } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_CA); } } } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_NO_CUMACK); } } skip_cwnd_update: /* * NOW, according to Karn's rule do we need to restore the * RTO timer back? Check our net_ack2. If not set then we * have a ambiguity.. i.e. all data ack'd was sent to more * than one place. */ if (net->net_ack2) { /* restore any doubled timers */ net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1; if (net->RTO < stcb->asoc.minrto) { net->RTO = stcb->asoc.minrto; } if (net->RTO > stcb->asoc.maxrto) { net->RTO = stcb->asoc.maxrto; } } } } static void sctp_cwnd_update_after_timeout(struct sctp_tcb *stcb, struct sctp_nets *net) { int old_cwnd = net->cwnd; uint32_t t_ssthresh, t_cwnd; /* MT FIXME: Don't compute this over and over again */ t_ssthresh = 0; t_cwnd = 0; if (stcb->asoc.sctp_cmt_on_off == 2) { struct sctp_nets *lnet; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { t_ssthresh += lnet->ssthresh; t_cwnd += lnet->cwnd; } net->ssthresh = (uint32_t) (((uint64_t) 4 * (uint64_t) net->mtu * (uint64_t) net->ssthresh) / (uint64_t) t_ssthresh); if ((net->cwnd > t_cwnd / 2) && (net->ssthresh < net->cwnd - t_cwnd / 2)) { net->ssthresh = net->cwnd - t_cwnd / 2; } if (net->ssthresh < net->mtu) { net->ssthresh = net->mtu; } } else { net->ssthresh = max(net->cwnd / 2, 4 * net->mtu); } net->cwnd = net->mtu; net->partial_bytes_acked = 0; SDT_PROBE(sctp, cwnd, net, to, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX); } } static void -sctp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net) +sctp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, + int in_window, int num_pkt_lost) { int old_cwnd = net->cwnd; - SCTP_STAT_INCR(sctps_ecnereducedcwnd); - net->ssthresh = net->cwnd / 2; - if (net->ssthresh < net->mtu) { - net->ssthresh = net->mtu; - /* here back off the timer as well, to slow us down */ - net->RTO <<= 1; - } - net->cwnd = net->ssthresh; - SDT_PROBE(sctp, cwnd, net, ecn, - stcb->asoc.my_vtag, - ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), - net, - old_cwnd, net->cwnd); - if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { - sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + if (in_window == 0) { + SCTP_STAT_INCR(sctps_ecnereducedcwnd); + net->ssthresh = net->cwnd / 2; + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + /* here back off the timer as well, to slow us down */ + net->RTO <<= 1; + } + net->cwnd = net->ssthresh; + SDT_PROBE(sctp, cwnd, net, ecn, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } } } static void sctp_cwnd_update_after_packet_dropped(struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_pktdrop_chunk *cp, uint32_t * bottle_bw, uint32_t * on_queue) { uint32_t bw_avail; int rtt, incr; int old_cwnd = net->cwnd; /* need real RTT for this calc */ rtt = ((net->lastsa >> 2) + net->lastsv) >> 1; /* get bottle neck bw */ *bottle_bw = ntohl(cp->bottle_bw); /* and whats on queue */ *on_queue = ntohl(cp->current_onq); /* * adjust the on-queue if our flight is more it could be that the * router has not yet gotten data "in-flight" to it */ if (*on_queue < net->flight_size) *on_queue = net->flight_size; /* calculate the available space */ bw_avail = (*bottle_bw * rtt) / 1000; if (bw_avail > *bottle_bw) { /* * Cap the growth to no more than the bottle neck. This can * happen as RTT slides up due to queues. It also means if * you have more than a 1 second RTT with a empty queue you * will be limited to the bottle_bw per second no matter if * other points have 1/2 the RTT and you could get more * out... */ bw_avail = *bottle_bw; } if (*on_queue > bw_avail) { /* * No room for anything else don't allow anything else to be * "added to the fire". */ int seg_inflight, seg_onqueue, my_portion; net->partial_bytes_acked = 0; /* how much are we over queue size? */ incr = *on_queue - bw_avail; if (stcb->asoc.seen_a_sack_this_pkt) { /* * undo any cwnd adjustment that the sack might have * made */ net->cwnd = net->prev_cwnd; } /* Now how much of that is mine? */ seg_inflight = net->flight_size / net->mtu; seg_onqueue = *on_queue / net->mtu; my_portion = (incr * seg_inflight) / seg_onqueue; /* Have I made an adjustment already */ if (net->cwnd > net->flight_size) { /* * for this flight I made an adjustment we need to * decrease the portion by a share our previous * adjustment. */ int diff_adj; diff_adj = net->cwnd - net->flight_size; if (diff_adj > my_portion) my_portion = 0; else my_portion -= diff_adj; } /* * back down to the previous cwnd (assume we have had a sack * before this packet). minus what ever portion of the * overage is my fault. */ net->cwnd -= my_portion; /* we will NOT back down more than 1 MTU */ if (net->cwnd <= net->mtu) { net->cwnd = net->mtu; } /* force into CA */ net->ssthresh = net->cwnd - 1; } else { /* * Take 1/4 of the space left or max burst up .. whichever * is less. */ incr = (bw_avail - *on_queue) >> 2; if ((stcb->asoc.max_burst > 0) && (stcb->asoc.max_burst * net->mtu < incr)) { incr = stcb->asoc.max_burst * net->mtu; } net->cwnd += incr; } if (net->cwnd > bw_avail) { /* We can't exceed the pipe size */ net->cwnd = bw_avail; } if (net->cwnd < net->mtu) { /* We always have 1 MTU */ net->cwnd = net->mtu; } if (net->cwnd - old_cwnd != 0) { /* log only changes */ SDT_PROBE(sctp, cwnd, net, pd, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); } } } static void sctp_cwnd_update_after_output(struct sctp_tcb *stcb, struct sctp_nets *net, int burst_limit) { int old_cwnd = net->cwnd; if (net->ssthresh < net->cwnd) net->ssthresh = net->cwnd; net->cwnd = (net->flight_size + (burst_limit * net->mtu)); SDT_PROBE(sctp, cwnd, net, bl, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_BRST); } } static void sctp_cwnd_update_after_fr_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int old_cwnd = net->cwnd; sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED); /* * make a small adjustment to cwnd and force to CA. */ if (net->cwnd > net->mtu) /* drop down one MTU after sending */ net->cwnd -= net->mtu; if (net->cwnd < net->ssthresh) /* still in SS move to CA */ net->ssthresh = net->cwnd - 1; SDT_PROBE(sctp, cwnd, net, fr, stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, old_cwnd, net->cwnd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (old_cwnd - net->cwnd), SCTP_CWND_LOG_FROM_FR); } } struct sctp_hs_raise_drop { int32_t cwnd; int32_t increase; int32_t drop_percent; }; #define SCTP_HS_TABLE_SIZE 73 struct sctp_hs_raise_drop sctp_cwnd_adjust[SCTP_HS_TABLE_SIZE] = { {38, 1, 50}, /* 0 */ {118, 2, 44}, /* 1 */ {221, 3, 41}, /* 2 */ {347, 4, 38}, /* 3 */ {495, 5, 37}, /* 4 */ {663, 6, 35}, /* 5 */ {851, 7, 34}, /* 6 */ {1058, 8, 33}, /* 7 */ {1284, 9, 32}, /* 8 */ {1529, 10, 31}, /* 9 */ {1793, 11, 30}, /* 10 */ {2076, 12, 29}, /* 11 */ {2378, 13, 28}, /* 12 */ {2699, 14, 28}, /* 13 */ {3039, 15, 27}, /* 14 */ {3399, 16, 27}, /* 15 */ {3778, 17, 26}, /* 16 */ {4177, 18, 26}, /* 17 */ {4596, 19, 25}, /* 18 */ {5036, 20, 25}, /* 19 */ {5497, 21, 24}, /* 20 */ {5979, 22, 24}, /* 21 */ {6483, 23, 23}, /* 22 */ {7009, 24, 23}, /* 23 */ {7558, 25, 22}, /* 24 */ {8130, 26, 22}, /* 25 */ {8726, 27, 22}, /* 26 */ {9346, 28, 21}, /* 27 */ {9991, 29, 21}, /* 28 */ {10661, 30, 21}, /* 29 */ {11358, 31, 20}, /* 30 */ {12082, 32, 20}, /* 31 */ {12834, 33, 20}, /* 32 */ {13614, 34, 19}, /* 33 */ {14424, 35, 19}, /* 34 */ {15265, 36, 19}, /* 35 */ {16137, 37, 19}, /* 36 */ {17042, 38, 18}, /* 37 */ {17981, 39, 18}, /* 38 */ {18955, 40, 18}, /* 39 */ {19965, 41, 17}, /* 40 */ {21013, 42, 17}, /* 41 */ {22101, 43, 17}, /* 42 */ {23230, 44, 17}, /* 43 */ {24402, 45, 16}, /* 44 */ {25618, 46, 16}, /* 45 */ {26881, 47, 16}, /* 46 */ {28193, 48, 16}, /* 47 */ {29557, 49, 15}, /* 48 */ {30975, 50, 15}, /* 49 */ {32450, 51, 15}, /* 50 */ {33986, 52, 15}, /* 51 */ {35586, 53, 14}, /* 52 */ {37253, 54, 14}, /* 53 */ {38992, 55, 14}, /* 54 */ {40808, 56, 14}, /* 55 */ {42707, 57, 13}, /* 56 */ {44694, 58, 13}, /* 57 */ {46776, 59, 13}, /* 58 */ {48961, 60, 13}, /* 59 */ {51258, 61, 13}, /* 60 */ {53677, 62, 12}, /* 61 */ {56230, 63, 12}, /* 62 */ {58932, 64, 12}, /* 63 */ {61799, 65, 12}, /* 64 */ {64851, 66, 11}, /* 65 */ {68113, 67, 11}, /* 66 */ {71617, 68, 11}, /* 67 */ {75401, 69, 10}, /* 68 */ {79517, 70, 10}, /* 69 */ {84035, 71, 10}, /* 70 */ {89053, 72, 10}, /* 71 */ {94717, 73, 9} /* 72 */ }; static void sctp_hs_cwnd_increase(struct sctp_tcb *stcb, struct sctp_nets *net) { int cur_val, i, indx, incr; cur_val = net->cwnd >> 10; indx = SCTP_HS_TABLE_SIZE - 1; #ifdef SCTP_DEBUG printf("HS CC CAlled.\n"); #endif if (cur_val < sctp_cwnd_adjust[0].cwnd) { /* normal mode */ if (net->net_ack > net->mtu) { net->cwnd += net->mtu; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_SS); } } else { net->cwnd += net->net_ack; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_FROM_SS); } } } else { for (i = net->last_hs_used; i < SCTP_HS_TABLE_SIZE; i++) { if (cur_val < sctp_cwnd_adjust[i].cwnd) { indx = i; break; } } net->last_hs_used = indx; incr = ((sctp_cwnd_adjust[indx].increase) << 10); net->cwnd += incr; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, incr, SCTP_CWND_LOG_FROM_SS); } } } static void sctp_hs_cwnd_decrease(struct sctp_tcb *stcb, struct sctp_nets *net) { int cur_val, i, indx; int old_cwnd = net->cwnd; cur_val = net->cwnd >> 10; if (cur_val < sctp_cwnd_adjust[0].cwnd) { /* normal mode */ net->ssthresh = net->cwnd / 2; if (net->ssthresh < (net->mtu * 2)) { net->ssthresh = 2 * net->mtu; } net->cwnd = net->ssthresh; } else { /* drop by the proper amount */ net->ssthresh = net->cwnd - (int)((net->cwnd / 100) * sctp_cwnd_adjust[net->last_hs_used].drop_percent); net->cwnd = net->ssthresh; /* now where are we */ indx = net->last_hs_used; cur_val = net->cwnd >> 10; /* reset where we are in the table */ if (cur_val < sctp_cwnd_adjust[0].cwnd) { /* feel out of hs */ net->last_hs_used = 0; } else { for (i = indx; i >= 1; i--) { if (cur_val > sctp_cwnd_adjust[i - 1].cwnd) { break; } } net->last_hs_used = indx; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR); } } static void sctp_hs_cwnd_update_after_fr(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_nets *net; /* * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && * (net->fast_retran_loss_recovery == 0))) */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if ((asoc->fast_retran_loss_recovery == 0) || (asoc->sctp_cmt_on_off > 0)) { /* out of a RFC2582 Fast recovery window? */ if (net->net_ack > 0) { /* * per section 7.2.3, are there any * destinations that had a fast retransmit * to them. If so what we need to do is * adjust ssthresh and cwnd. */ struct sctp_tmit_chunk *lchk; sctp_hs_cwnd_decrease(stcb, net); lchk = TAILQ_FIRST(&asoc->send_queue); net->partial_bytes_acked = 0; /* Turn on fast recovery window */ asoc->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ asoc->fast_recovery_tsn = asoc->sending_seq - 1; } else { asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } /* * CMT fast recovery -- per destination * recovery variable. */ net->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ net->fast_recovery_tsn = asoc->sending_seq - 1; } else { net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } - /* - * Disable Nonce Sum Checking and store the - * resync tsn - */ - asoc->nonce_sum_check = 0; - asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1; - sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); } } else if (net->net_ack > 0) { /* * Mark a peg that we WOULD have done a cwnd * reduction but RFC2582 prevented this action. */ SCTP_STAT_INCR(sctps_fastretransinrtt); } } } static void sctp_hs_cwnd_update_after_sack(struct sctp_tcb *stcb, struct sctp_association *asoc, int accum_moved, int reneged_all, int will_exit) { struct sctp_nets *net; /******************************/ /* update cwnd and Early FR */ /******************************/ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code. Need to debug. */ if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) { net->will_exit_fast_recovery = 1; } } #endif if (SCTP_BASE_SYSCTL(sctp_early_fr)) { /* * So, first of all do we need to have a Early FR * timer running? */ if ((!TAILQ_EMPTY(&asoc->sent_queue) && (net->ref_count > 1) && (net->flight_size < net->cwnd)) || (reneged_all)) { /* * yes, so in this case stop it if its * running, and then restart it. Reneging * all is a special case where we want to * run the Early FR timer and then force the * last few unacked to be sent, causing us * to illicit a sack with gaps to force out * the others. */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck2); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_20); } SCTP_STAT_INCR(sctps_earlyfrstrid); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net); } else { /* No, stop it if its running */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck3); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_21); } } } /* if nothing was acked on this destination skip it */ if (net->net_ack == 0) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); } continue; } if (net->net_ack2 > 0) { /* * Karn's rule applies to clearing error count, this * is optional. */ net->error_count = 0; if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) == SCTP_ADDR_NOT_REACHABLE) { /* addr came good */ net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE; net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED); /* now was it the primary? if so restore */ if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) { (void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net); } } /* * JRS 5/14/07 - If CMT PF is on and the destination * is in PF state, set the destination to active * state and set the cwnd to one or two MTU's based * on whether PF1 or PF2 is being used. * * Should we stop any running T3 timer here? */ if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { net->dest_state &= ~SCTP_ADDR_PF; net->cwnd = net->mtu * asoc->sctp_cmt_pf; SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n", net, net->cwnd); /* * Since the cwnd value is explicitly set, * skip the code that updates the cwnd * value. */ goto skip_cwnd_update; } } #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code */ /* * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery * && net->will_exit_fast_recovery == 0) { @@@ Do something * } else if (sctp_cmt_on_off == 0 && * asoc->fast_retran_loss_recovery && will_exit == 0) { */ #endif if (asoc->fast_retran_loss_recovery && (will_exit == 0) && (asoc->sctp_cmt_on_off == 0)) { /* * If we are in loss recovery we skip any cwnd * update */ goto skip_cwnd_update; } /* * CMT: CUC algorithm. Update cwnd if pseudo-cumack has * moved. */ if (accum_moved || ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { /* If the cumulative ack moved we can proceed */ if (net->cwnd <= net->ssthresh) { /* We are in slow start */ if (net->flight_size + net->net_ack >= net->cwnd) { sctp_hs_cwnd_increase(stcb, net); } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_SS); } } } else { /* We are in congestion avoidance */ net->partial_bytes_acked += net->net_ack; if ((net->flight_size + net->net_ack >= net->cwnd) && (net->partial_bytes_acked >= net->cwnd)) { net->partial_bytes_acked -= net->cwnd; net->cwnd += net->mtu; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_CA); } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_CA); } } } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_NO_CUMACK); } } skip_cwnd_update: /* * NOW, according to Karn's rule do we need to restore the * RTO timer back? Check our net_ack2. If not set then we * have a ambiguity.. i.e. all data ack'd was sent to more * than one place. */ if (net->net_ack2) { /* restore any doubled timers */ net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1; if (net->RTO < stcb->asoc.minrto) { net->RTO = stcb->asoc.minrto; } if (net->RTO > stcb->asoc.maxrto) { net->RTO = stcb->asoc.maxrto; } } } } /* * H-TCP congestion control. The algorithm is detailed in: * R.N.Shorten, D.J.Leith: * "H-TCP: TCP for high-speed and long-distance networks" * Proc. PFLDnet, Argonne, 2004. * http://www.hamilton.ie/net/htcp3.pdf */ static int use_rtt_scaling = 1; static int use_bandwidth_switch = 1; static inline int between(uint32_t seq1, uint32_t seq2, uint32_t seq3) { return seq3 - seq2 >= seq1 - seq2; } static inline uint32_t htcp_cong_time(struct htcp *ca) { return sctp_get_tick_count() - ca->last_cong; } static inline uint32_t htcp_ccount(struct htcp *ca) { return htcp_cong_time(ca) / ca->minRTT; } static inline void htcp_reset(struct htcp *ca) { ca->undo_last_cong = ca->last_cong; ca->undo_maxRTT = ca->maxRTT; ca->undo_old_maxB = ca->old_maxB; ca->last_cong = sctp_get_tick_count(); } #ifdef SCTP_NOT_USED static uint32_t htcp_cwnd_undo(struct sctp_tcb *stcb, struct sctp_nets *net) { net->htcp_ca.last_cong = net->htcp_ca.undo_last_cong; net->htcp_ca.maxRTT = net->htcp_ca.undo_maxRTT; net->htcp_ca.old_maxB = net->htcp_ca.undo_old_maxB; return max(net->cwnd, ((net->ssthresh / net->mtu << 7) / net->htcp_ca.beta) * net->mtu); } #endif static inline void measure_rtt(struct sctp_tcb *stcb, struct sctp_nets *net) { uint32_t srtt = net->lastsa >> 3; /* keep track of minimum RTT seen so far, minRTT is zero at first */ if (net->htcp_ca.minRTT > srtt || !net->htcp_ca.minRTT) net->htcp_ca.minRTT = srtt; /* max RTT */ if (net->fast_retran_ip == 0 && net->ssthresh < 0xFFFF && htcp_ccount(&net->htcp_ca) > 3) { if (net->htcp_ca.maxRTT < net->htcp_ca.minRTT) net->htcp_ca.maxRTT = net->htcp_ca.minRTT; if (net->htcp_ca.maxRTT < srtt && srtt <= net->htcp_ca.maxRTT + MSEC_TO_TICKS(20)) net->htcp_ca.maxRTT = srtt; } } static void measure_achieved_throughput(struct sctp_tcb *stcb, struct sctp_nets *net) { uint32_t now = sctp_get_tick_count(); if (net->fast_retran_ip == 0) net->htcp_ca.bytes_acked = net->net_ack; if (!use_bandwidth_switch) return; /* achieved throughput calculations */ /* JRS - not 100% sure of this statement */ if (net->fast_retran_ip == 1) { net->htcp_ca.bytecount = 0; net->htcp_ca.lasttime = now; return; } net->htcp_ca.bytecount += net->net_ack; if (net->htcp_ca.bytecount >= net->cwnd - ((net->htcp_ca.alpha >> 7 ? : 1) * net->mtu) && now - net->htcp_ca.lasttime >= net->htcp_ca.minRTT && net->htcp_ca.minRTT > 0) { uint32_t cur_Bi = net->htcp_ca.bytecount / net->mtu * hz / (now - net->htcp_ca.lasttime); if (htcp_ccount(&net->htcp_ca) <= 3) { /* just after backoff */ net->htcp_ca.minB = net->htcp_ca.maxB = net->htcp_ca.Bi = cur_Bi; } else { net->htcp_ca.Bi = (3 * net->htcp_ca.Bi + cur_Bi) / 4; if (net->htcp_ca.Bi > net->htcp_ca.maxB) net->htcp_ca.maxB = net->htcp_ca.Bi; if (net->htcp_ca.minB > net->htcp_ca.maxB) net->htcp_ca.minB = net->htcp_ca.maxB; } net->htcp_ca.bytecount = 0; net->htcp_ca.lasttime = now; } } static inline void htcp_beta_update(struct htcp *ca, uint32_t minRTT, uint32_t maxRTT) { if (use_bandwidth_switch) { uint32_t maxB = ca->maxB; uint32_t old_maxB = ca->old_maxB; ca->old_maxB = ca->maxB; if (!between(5 * maxB, 4 * old_maxB, 6 * old_maxB)) { ca->beta = BETA_MIN; ca->modeswitch = 0; return; } } if (ca->modeswitch && minRTT > (uint32_t) MSEC_TO_TICKS(10) && maxRTT) { ca->beta = (minRTT << 7) / maxRTT; if (ca->beta < BETA_MIN) ca->beta = BETA_MIN; else if (ca->beta > BETA_MAX) ca->beta = BETA_MAX; } else { ca->beta = BETA_MIN; ca->modeswitch = 1; } } static inline void htcp_alpha_update(struct htcp *ca) { uint32_t minRTT = ca->minRTT; uint32_t factor = 1; uint32_t diff = htcp_cong_time(ca); if (diff > (uint32_t) hz) { diff -= hz; factor = 1 + (10 * diff + ((diff / 2) * (diff / 2) / hz)) / hz; } if (use_rtt_scaling && minRTT) { uint32_t scale = (hz << 3) / (10 * minRTT); scale = min(max(scale, 1U << 2), 10U << 3); /* clamping ratio to * interval [0.5,10]<<3 */ factor = (factor << 3) / scale; if (!factor) factor = 1; } ca->alpha = 2 * factor * ((1 << 7) - ca->beta); if (!ca->alpha) ca->alpha = ALPHA_BASE; } /* After we have the rtt data to calculate beta, we'd still prefer to wait one * rtt before we adjust our beta to ensure we are working from a consistent * data. * * This function should be called when we hit a congestion event since only at * that point do we really have a real sense of maxRTT (the queues en route * were getting just too full now). */ static void htcp_param_update(struct sctp_tcb *stcb, struct sctp_nets *net) { uint32_t minRTT = net->htcp_ca.minRTT; uint32_t maxRTT = net->htcp_ca.maxRTT; htcp_beta_update(&net->htcp_ca, minRTT, maxRTT); htcp_alpha_update(&net->htcp_ca); /* * add slowly fading memory for maxRTT to accommodate routing * changes etc */ if (minRTT > 0 && maxRTT > minRTT) net->htcp_ca.maxRTT = minRTT + ((maxRTT - minRTT) * 95) / 100; } static uint32_t htcp_recalc_ssthresh(struct sctp_tcb *stcb, struct sctp_nets *net) { htcp_param_update(stcb, net); return max(((net->cwnd / net->mtu * net->htcp_ca.beta) >> 7) * net->mtu, 2U * net->mtu); } static void htcp_cong_avoid(struct sctp_tcb *stcb, struct sctp_nets *net) { /*- * How to handle these functions? * if (!tcp_is_cwnd_limited(sk, in_flight)) RRS - good question. * return; */ if (net->cwnd <= net->ssthresh) { /* We are in slow start */ if (net->flight_size + net->net_ack >= net->cwnd) { if (net->net_ack > (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable))) { net->cwnd += (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable)); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_SS); } } else { net->cwnd += net->net_ack; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_FROM_SS); } } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_SS); } } } else { measure_rtt(stcb, net); /* * In dangerous area, increase slowly. In theory this is * net->cwnd += alpha / net->cwnd */ /* What is snd_cwnd_cnt?? */ if (((net->partial_bytes_acked / net->mtu * net->htcp_ca.alpha) >> 7) * net->mtu >= net->cwnd) { /*- * Does SCTP have a cwnd clamp? * if (net->snd_cwnd < net->snd_cwnd_clamp) - Nope (RRS). */ net->cwnd += net->mtu; net->partial_bytes_acked = 0; htcp_alpha_update(&net->htcp_ca); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_CA); } } else { net->partial_bytes_acked += net->net_ack; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_NOADV_CA); } } net->htcp_ca.bytes_acked = net->mtu; } } #ifdef SCTP_NOT_USED /* Lower bound on congestion window. */ static uint32_t htcp_min_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net) { return net->ssthresh; } #endif static void htcp_init(struct sctp_tcb *stcb, struct sctp_nets *net) { memset(&net->htcp_ca, 0, sizeof(struct htcp)); net->htcp_ca.alpha = ALPHA_BASE; net->htcp_ca.beta = BETA_MIN; net->htcp_ca.bytes_acked = net->mtu; net->htcp_ca.last_cong = sctp_get_tick_count(); } static void sctp_htcp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net) { /* * We take the max of the burst limit times a MTU or the * INITIAL_CWND. We then limit this to 4 MTU's of sending. */ net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND)); net->ssthresh = stcb->asoc.peers_rwnd; htcp_init(stcb, net); if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION); } } static void sctp_htcp_cwnd_update_after_sack(struct sctp_tcb *stcb, struct sctp_association *asoc, int accum_moved, int reneged_all, int will_exit) { struct sctp_nets *net; /******************************/ /* update cwnd and Early FR */ /******************************/ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code. Need to debug. */ if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) { net->will_exit_fast_recovery = 1; } } #endif if (SCTP_BASE_SYSCTL(sctp_early_fr)) { /* * So, first of all do we need to have a Early FR * timer running? */ if ((!TAILQ_EMPTY(&asoc->sent_queue) && (net->ref_count > 1) && (net->flight_size < net->cwnd)) || (reneged_all)) { /* * yes, so in this case stop it if its * running, and then restart it. Reneging * all is a special case where we want to * run the Early FR timer and then force the * last few unacked to be sent, causing us * to illicit a sack with gaps to force out * the others. */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck2); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_20); } SCTP_STAT_INCR(sctps_earlyfrstrid); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net); } else { /* No, stop it if its running */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck3); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_21); } } } /* if nothing was acked on this destination skip it */ if (net->net_ack == 0) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); } continue; } if (net->net_ack2 > 0) { /* * Karn's rule applies to clearing error count, this * is optional. */ net->error_count = 0; if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) == SCTP_ADDR_NOT_REACHABLE) { /* addr came good */ net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE; net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED); /* now was it the primary? if so restore */ if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) { (void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net); } } /* * JRS 5/14/07 - If CMT PF is on and the destination * is in PF state, set the destination to active * state and set the cwnd to one or two MTU's based * on whether PF1 or PF2 is being used. * * Should we stop any running T3 timer here? */ if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { net->dest_state &= ~SCTP_ADDR_PF; net->cwnd = net->mtu * asoc->sctp_cmt_pf; SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n", net, net->cwnd); /* * Since the cwnd value is explicitly set, * skip the code that updates the cwnd * value. */ goto skip_cwnd_update; } } #ifdef JANA_CMT_FAST_RECOVERY /* * CMT fast recovery code */ /* * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery * && net->will_exit_fast_recovery == 0) { @@@ Do something * } else if (sctp_cmt_on_off == 0 && * asoc->fast_retran_loss_recovery && will_exit == 0) { */ #endif if (asoc->fast_retran_loss_recovery && will_exit == 0 && (asoc->sctp_cmt_on_off == 0)) { /* * If we are in loss recovery we skip any cwnd * update */ goto skip_cwnd_update; } /* * CMT: CUC algorithm. Update cwnd if pseudo-cumack has * moved. */ if (accum_moved || ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { htcp_cong_avoid(stcb, net); measure_achieved_throughput(stcb, net); } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_NO_CUMACK); } } skip_cwnd_update: /* * NOW, according to Karn's rule do we need to restore the * RTO timer back? Check our net_ack2. If not set then we * have a ambiguity.. i.e. all data ack'd was sent to more * than one place. */ if (net->net_ack2) { /* restore any doubled timers */ net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1; if (net->RTO < stcb->asoc.minrto) { net->RTO = stcb->asoc.minrto; } if (net->RTO > stcb->asoc.maxrto) { net->RTO = stcb->asoc.maxrto; } } } } static void sctp_htcp_cwnd_update_after_fr(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_nets *net; /* * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && * (net->fast_retran_loss_recovery == 0))) */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if ((asoc->fast_retran_loss_recovery == 0) || (asoc->sctp_cmt_on_off > 0)) { /* out of a RFC2582 Fast recovery window? */ if (net->net_ack > 0) { /* * per section 7.2.3, are there any * destinations that had a fast retransmit * to them. If so what we need to do is * adjust ssthresh and cwnd. */ struct sctp_tmit_chunk *lchk; int old_cwnd = net->cwnd; /* JRS - reset as if state were changed */ htcp_reset(&net->htcp_ca); net->ssthresh = htcp_recalc_ssthresh(stcb, net); net->cwnd = net->ssthresh; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR); } lchk = TAILQ_FIRST(&asoc->send_queue); net->partial_bytes_acked = 0; /* Turn on fast recovery window */ asoc->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ asoc->fast_recovery_tsn = asoc->sending_seq - 1; } else { asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } /* * CMT fast recovery -- per destination * recovery variable. */ net->fast_retran_loss_recovery = 1; if (lchk == NULL) { /* Mark end of the window */ net->fast_recovery_tsn = asoc->sending_seq - 1; } else { net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1; } - /* - * Disable Nonce Sum Checking and store the - * resync tsn - */ - asoc->nonce_sum_check = 0; - asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1; - sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); } } else if (net->net_ack > 0) { /* * Mark a peg that we WOULD have done a cwnd * reduction but RFC2582 prevented this action. */ SCTP_STAT_INCR(sctps_fastretransinrtt); } } } static void sctp_htcp_cwnd_update_after_timeout(struct sctp_tcb *stcb, struct sctp_nets *net) { int old_cwnd = net->cwnd; /* JRS - reset as if the state were being changed to timeout */ htcp_reset(&net->htcp_ca); net->ssthresh = htcp_recalc_ssthresh(stcb, net); net->cwnd = net->mtu; net->partial_bytes_acked = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX); } } static void sctp_htcp_cwnd_update_after_fr_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int old_cwnd; old_cwnd = net->cwnd; sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED); net->htcp_ca.last_cong = sctp_get_tick_count(); /* * make a small adjustment to cwnd and force to CA. */ if (net->cwnd > net->mtu) /* drop down one MTU after sending */ net->cwnd -= net->mtu; if (net->cwnd < net->ssthresh) /* still in SS move to CA */ net->ssthresh = net->cwnd - 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, (old_cwnd - net->cwnd), SCTP_CWND_LOG_FROM_FR); } } static void sctp_htcp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, - struct sctp_nets *net) + struct sctp_nets *net, int in_window, int num_pkt_lost) { int old_cwnd; old_cwnd = net->cwnd; /* JRS - reset hctp as if state changed */ - htcp_reset(&net->htcp_ca); - SCTP_STAT_INCR(sctps_ecnereducedcwnd); - net->ssthresh = htcp_recalc_ssthresh(stcb, net); - if (net->ssthresh < net->mtu) { - net->ssthresh = net->mtu; - /* here back off the timer as well, to slow us down */ - net->RTO <<= 1; - } - net->cwnd = net->ssthresh; - if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { - sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + if (in_window == 0) { + htcp_reset(&net->htcp_ca); + SCTP_STAT_INCR(sctps_ecnereducedcwnd); + net->ssthresh = htcp_recalc_ssthresh(stcb, net); + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + /* here back off the timer as well, to slow us down */ + net->RTO <<= 1; + } + net->cwnd = net->ssthresh; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } } } struct sctp_cc_functions sctp_cc_functions[] = { { .sctp_set_initial_cc_param = sctp_set_initial_cc_param, .sctp_cwnd_update_after_sack = sctp_cwnd_update_after_sack, .sctp_cwnd_update_after_fr = sctp_cwnd_update_after_fr, .sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout, .sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo, .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, .sctp_cwnd_update_after_fr_timer = sctp_cwnd_update_after_fr_timer }, { .sctp_set_initial_cc_param = sctp_set_initial_cc_param, .sctp_cwnd_update_after_sack = sctp_hs_cwnd_update_after_sack, .sctp_cwnd_update_after_fr = sctp_hs_cwnd_update_after_fr, .sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout, .sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo, .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, .sctp_cwnd_update_after_fr_timer = sctp_cwnd_update_after_fr_timer }, { .sctp_set_initial_cc_param = sctp_htcp_set_initial_cc_param, .sctp_cwnd_update_after_sack = sctp_htcp_cwnd_update_after_sack, .sctp_cwnd_update_after_fr = sctp_htcp_cwnd_update_after_fr, .sctp_cwnd_update_after_timeout = sctp_htcp_cwnd_update_after_timeout, .sctp_cwnd_update_after_ecn_echo = sctp_htcp_cwnd_update_after_ecn_echo, .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, .sctp_cwnd_update_after_fr_timer = sctp_htcp_cwnd_update_after_fr_timer } }; diff --git a/sys/netinet/sctp_constants.h b/sys/netinet/sctp_constants.h index 988ec9e434c8..719da6453843 100644 --- a/sys/netinet/sctp_constants.h +++ b/sys/netinet/sctp_constants.h @@ -1,1051 +1,1050 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_constants.h,v 1.17 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_constants_h__ #define __sctp_constants_h__ /* IANA assigned port number for SCTP over UDP encapsulation */ /* For freebsd we cannot bind the port at * startup. Otherwise what will happen is * we really won't be bound. The user must * put it into the sysctl... or we need * to build a special timer for this to allow * us to wait 1 second or so after the system * comes up. */ #define SCTP_OVER_UDP_TUNNELING_PORT 0 /* Number of packets to get before sack sent by default */ #define SCTP_DEFAULT_SACK_FREQ 2 /* Address limit - This variable is calculated * based on an 65535 byte max ip packet. We take out 100 bytes * for the cookie, 40 bytes for a v6 header and 32 * bytes for the init structure. A second init structure * for the init-ack and then finally a third one for the * imbedded init. This yeilds 100+40+(3 * 32) = 236 bytes. * This leaves 65299 bytes for addresses. We throw out the 299 bytes. * Now whatever we send in the INIT() we need to allow to get back in the * INIT-ACK plus all the values from INIT and INIT-ACK * listed in the cookie. Plus we need some overhead for * maybe copied parameters in the COOKIE. If we * allow 1080 addresses, and each side has 1080 V6 addresses * that will be 21600 bytes. In the INIT-ACK we will * see the INIT-ACK 21600 + 43200 in the cookie. This leaves * about 500 bytes slack for misc things in the cookie. */ #define SCTP_ADDRESS_LIMIT 1080 /* We need at least 2k of space for us, inits * larger than that lets abort. */ #define SCTP_LARGEST_INIT_ACCEPTED (65535 - 2048) /* Number of addresses where we just skip the counting */ #define SCTP_COUNT_LIMIT 40 #define SCTP_ZERO_COPY_TICK_DELAY (((100 * hz) + 999) / 1000) #define SCTP_ZERO_COPY_SENDQ_TICK_DELAY (((100 * hz) + 999) / 1000) /* Number of ticks to delay before running * iterator on an address change. */ #define SCTP_ADDRESS_TICK_DELAY 2 #define SCTP_VERSION_STRING "KAME-BSD 1.1" /* #define SCTP_AUDITING_ENABLED 1 used for debug/auditing */ #define SCTP_AUDIT_SIZE 256 #define SCTP_KTRHEAD_NAME "sctp_iterator" #define SCTP_KTHREAD_PAGES 0 /* If you support Multi-VRF how big to * make the initial array of VRF's to. */ #define SCTP_DEFAULT_VRF_SIZE 4 /* constants for rto calc */ #define sctp_align_safe_nocopy 0 #define sctp_align_unsafe_makecopy 1 /* JRS - Values defined for the HTCP algorithm */ #define ALPHA_BASE (1<<7) /* 1.0 with shift << 7 */ #define BETA_MIN (1<<6) /* 0.5 with shift << 7 */ #define BETA_MAX 102 /* 0.8 with shift << 7 */ /* Places that CWND log can happen from */ #define SCTP_CWND_LOG_FROM_FR 1 #define SCTP_CWND_LOG_FROM_RTX 2 #define SCTP_CWND_LOG_FROM_BRST 3 #define SCTP_CWND_LOG_FROM_SS 4 #define SCTP_CWND_LOG_FROM_CA 5 #define SCTP_CWND_LOG_FROM_SAT 6 #define SCTP_BLOCK_LOG_INTO_BLK 7 #define SCTP_BLOCK_LOG_OUTOF_BLK 8 #define SCTP_BLOCK_LOG_CHECK 9 #define SCTP_STR_LOG_FROM_INTO_STRD 10 #define SCTP_STR_LOG_FROM_IMMED_DEL 11 #define SCTP_STR_LOG_FROM_INSERT_HD 12 #define SCTP_STR_LOG_FROM_INSERT_MD 13 #define SCTP_STR_LOG_FROM_INSERT_TL 14 #define SCTP_STR_LOG_FROM_MARK_TSN 15 #define SCTP_STR_LOG_FROM_EXPRS_DEL 16 #define SCTP_FR_LOG_BIGGEST_TSNS 17 #define SCTP_FR_LOG_STRIKE_TEST 18 #define SCTP_FR_LOG_STRIKE_CHUNK 19 #define SCTP_FR_T3_TIMEOUT 20 #define SCTP_MAP_PREPARE_SLIDE 21 #define SCTP_MAP_SLIDE_FROM 22 #define SCTP_MAP_SLIDE_RESULT 23 #define SCTP_MAP_SLIDE_CLEARED 24 #define SCTP_MAP_SLIDE_NONE 25 #define SCTP_FR_T3_MARK_TIME 26 #define SCTP_FR_T3_MARKED 27 #define SCTP_FR_T3_STOPPED 28 #define SCTP_FR_MARKED 30 #define SCTP_CWND_LOG_NOADV_SS 31 #define SCTP_CWND_LOG_NOADV_CA 32 #define SCTP_MAX_BURST_APPLIED 33 #define SCTP_MAX_IFP_APPLIED 34 #define SCTP_MAX_BURST_ERROR_STOP 35 #define SCTP_INCREASE_PEER_RWND 36 #define SCTP_DECREASE_PEER_RWND 37 #define SCTP_SET_PEER_RWND_VIA_SACK 38 #define SCTP_LOG_MBCNT_INCREASE 39 #define SCTP_LOG_MBCNT_DECREASE 40 #define SCTP_LOG_MBCNT_CHKSET 41 #define SCTP_LOG_NEW_SACK 42 #define SCTP_LOG_TSN_ACKED 43 #define SCTP_LOG_TSN_REVOKED 44 #define SCTP_LOG_LOCK_TCB 45 #define SCTP_LOG_LOCK_INP 46 #define SCTP_LOG_LOCK_SOCK 47 #define SCTP_LOG_LOCK_SOCKBUF_R 48 #define SCTP_LOG_LOCK_SOCKBUF_S 49 #define SCTP_LOG_LOCK_CREATE 50 #define SCTP_LOG_INITIAL_RTT 51 #define SCTP_LOG_RTTVAR 52 #define SCTP_LOG_SBALLOC 53 #define SCTP_LOG_SBFREE 54 #define SCTP_LOG_SBRESULT 55 #define SCTP_FR_DUPED 56 #define SCTP_FR_MARKED_EARLY 57 #define SCTP_FR_CWND_REPORT 58 #define SCTP_FR_CWND_REPORT_START 59 #define SCTP_FR_CWND_REPORT_STOP 60 #define SCTP_CWND_LOG_FROM_SEND 61 #define SCTP_CWND_INITIALIZATION 62 #define SCTP_CWND_LOG_FROM_T3 63 #define SCTP_CWND_LOG_FROM_SACK 64 #define SCTP_CWND_LOG_NO_CUMACK 65 #define SCTP_CWND_LOG_FROM_RESEND 66 #define SCTP_FR_LOG_CHECK_STRIKE 67 #define SCTP_SEND_NOW_COMPLETES 68 #define SCTP_CWND_LOG_FILL_OUTQ_CALLED 69 #define SCTP_CWND_LOG_FILL_OUTQ_FILLS 70 #define SCTP_LOG_FREE_SENT 71 #define SCTP_NAGLE_APPLIED 72 #define SCTP_NAGLE_SKIPPED 73 #define SCTP_WAKESND_FROM_SACK 74 #define SCTP_WAKESND_FROM_FWDTSN 75 #define SCTP_NOWAKE_FROM_SACK 76 #define SCTP_CWNDLOG_PRESEND 77 #define SCTP_CWNDLOG_ENDSEND 78 #define SCTP_AT_END_OF_SACK 79 #define SCTP_REASON_FOR_SC 80 #define SCTP_BLOCK_LOG_INTO_BLKA 81 #define SCTP_ENTER_USER_RECV 82 #define SCTP_USER_RECV_SACKS 83 #define SCTP_SORECV_BLOCKSA 84 #define SCTP_SORECV_BLOCKSB 85 #define SCTP_SORECV_DONE 86 #define SCTP_SACK_RWND_UPDATE 87 #define SCTP_SORECV_ENTER 88 #define SCTP_SORECV_ENTERPL 89 #define SCTP_MBUF_INPUT 90 #define SCTP_MBUF_IALLOC 91 #define SCTP_MBUF_IFREE 92 #define SCTP_MBUF_ICOPY 93 #define SCTP_MBUF_SPLIT 94 #define SCTP_SORCV_FREECTL 95 #define SCTP_SORCV_DOESCPY 96 #define SCTP_SORCV_DOESLCK 97 #define SCTP_SORCV_DOESADJ 98 #define SCTP_SORCV_BOTWHILE 99 #define SCTP_SORCV_PASSBF 100 #define SCTP_SORCV_ADJD 101 #define SCTP_UNKNOWN_MAX 102 #define SCTP_RANDY_STUFF 103 #define SCTP_RANDY_STUFF1 104 #define SCTP_STRMOUT_LOG_ASSIGN 105 #define SCTP_STRMOUT_LOG_SEND 106 #define SCTP_FLIGHT_LOG_DOWN_CA 107 #define SCTP_FLIGHT_LOG_UP 108 #define SCTP_FLIGHT_LOG_DOWN_GAP 109 #define SCTP_FLIGHT_LOG_DOWN_RSND 110 #define SCTP_FLIGHT_LOG_UP_RSND 111 #define SCTP_FLIGHT_LOG_DOWN_RSND_TO 112 #define SCTP_FLIGHT_LOG_DOWN_WP 113 #define SCTP_FLIGHT_LOG_UP_REVOKE 114 #define SCTP_FLIGHT_LOG_DOWN_PDRP 115 #define SCTP_FLIGHT_LOG_DOWN_PMTU 116 #define SCTP_SACK_LOG_NORMAL 117 #define SCTP_SACK_LOG_EXPRESS 118 #define SCTP_MAP_TSN_ENTERS 119 #define SCTP_THRESHOLD_CLEAR 120 #define SCTP_THRESHOLD_INCR 121 #define SCTP_FLIGHT_LOG_DWN_WP_FWD 122 #define SCTP_FWD_TSN_CHECK 123 #define SCTP_LOG_MAX_TYPES 124 /* * To turn on various logging, you must first enable 'options KTR' and * you might want to bump the entires 'options KTR_ENTRIES=80000'. * To get something to log you define one of the logging defines. * (see LINT). * * This gets the compile in place, but you still need to turn the * logging flag on too in the sysctl (see in sctp.h). */ #define SCTP_LOG_EVENT_UNKNOWN 0 #define SCTP_LOG_EVENT_CWND 1 #define SCTP_LOG_EVENT_BLOCK 2 #define SCTP_LOG_EVENT_STRM 3 #define SCTP_LOG_EVENT_FR 4 #define SCTP_LOG_EVENT_MAP 5 #define SCTP_LOG_EVENT_MAXBURST 6 #define SCTP_LOG_EVENT_RWND 7 #define SCTP_LOG_EVENT_MBCNT 8 #define SCTP_LOG_EVENT_SACK 9 #define SCTP_LOG_LOCK_EVENT 10 #define SCTP_LOG_EVENT_RTT 11 #define SCTP_LOG_EVENT_SB 12 #define SCTP_LOG_EVENT_NAGLE 13 #define SCTP_LOG_EVENT_WAKE 14 #define SCTP_LOG_MISC_EVENT 15 #define SCTP_LOG_EVENT_CLOSE 16 #define SCTP_LOG_EVENT_MBUF 17 #define SCTP_LOG_CHUNK_PROC 18 #define SCTP_LOG_ERROR_RET 19 #define SCTP_LOG_MAX_EVENT 20 #define SCTP_LOCK_UNKNOWN 2 /* number of associations by default for zone allocation */ #define SCTP_MAX_NUM_OF_ASOC 40000 /* how many addresses per assoc remote and local */ #define SCTP_SCALE_FOR_ADDR 2 /* default AUTO_ASCONF mode enable(1)/disable(0) value (sysctl) */ #define SCTP_DEFAULT_AUTO_ASCONF 1 /* default MULTIPLE_ASCONF mode enable(1)/disable(0) value (sysctl) */ #define SCTP_DEFAULT_MULTIPLE_ASCONFS 0 /* default MOBILITY_BASE mode enable(1)/disable(0) value (sysctl) */ #define SCTP_DEFAULT_MOBILITY_BASE 0 /* default MOBILITY_FASTHANDOFF mode enable(1)/disable(0) value (sysctl) */ #define SCTP_DEFAULT_MOBILITY_FASTHANDOFF 0 /* * Theshold for rwnd updates, we have to read (sb_hiwat >> * SCTP_RWND_HIWAT_SHIFT) before we will look to see if we need to send a * window update sack. When we look, we compare the last rwnd we sent vs the * current rwnd. It too must be greater than this value. Using 3 divdes the * hiwat by 8, so for 200k rwnd we need to read 24k. For a 64k rwnd we need * to read 8k. This seems about right.. I hope :-D.. we do set a * min of a MTU on it so if the rwnd is real small we will insist * on a full MTU of 1500 bytes. */ #define SCTP_RWND_HIWAT_SHIFT 3 /* How much of the rwnd must the * message be taking up to start partial delivery. * We calculate this by shifing the hi_water (recv_win) * left the following .. set to 1, when a message holds * 1/2 the rwnd. If we set it to 2 when a message holds * 1/4 the rwnd...etc.. */ #define SCTP_PARTIAL_DELIVERY_SHIFT 1 /* * default HMAC for cookies, etc... use one of the AUTH HMAC id's * SCTP_HMAC is the HMAC_ID to use * SCTP_SIGNATURE_SIZE is the digest length */ #define SCTP_HMAC SCTP_AUTH_HMAC_ID_SHA1 #define SCTP_SIGNATURE_SIZE SCTP_AUTH_DIGEST_LEN_SHA1 #define SCTP_SIGNATURE_ALOC_SIZE SCTP_SIGNATURE_SIZE /* * the SCTP protocol signature this includes the version number encoded in * the last 4 bits of the signature. */ #define PROTO_SIGNATURE_A 0x30000000 #define SCTP_VERSION_NUMBER 0x3 #define MAX_TSN 0xffffffff /* how many executions every N tick's */ #define SCTP_ITERATOR_MAX_AT_ONCE 20 /* number of clock ticks between iterator executions */ #define SCTP_ITERATOR_TICKS 1 /* * option: If you comment out the following you will receive the old behavior * of obeying cwnd for the fast retransmit algorithm. With this defined a FR * happens right away with-out waiting for the flightsize to drop below the * cwnd value (which is reduced by the FR to 1/2 the inflight packets). */ #define SCTP_IGNORE_CWND_ON_FR 1 /* * Adds implementors guide behavior to only use newest highest update in SACK * gap ack's to figure out if you need to stroke a chunk for FR. */ #define SCTP_NO_FR_UNLESS_SEGMENT_SMALLER 1 /* default max I can burst out after a fast retransmit, 0 disables it */ #define SCTP_DEF_MAX_BURST 0 /* IP hdr (20/40) + 12+2+2 (enet) + sctp common 12 */ #define SCTP_FIRST_MBUF_RESV 68 /* Packet transmit states in the sent field */ #define SCTP_DATAGRAM_UNSENT 0 #define SCTP_DATAGRAM_SENT 1 #define SCTP_DATAGRAM_RESEND1 2 /* not used (in code, but may * hit this value) */ #define SCTP_DATAGRAM_RESEND2 3 /* not used (in code, but may * hit this value) */ #define SCTP_DATAGRAM_RESEND 4 #define SCTP_DATAGRAM_ACKED 10010 #define SCTP_DATAGRAM_MARKED 20010 #define SCTP_FORWARD_TSN_SKIP 30010 /* chunk output send from locations */ #define SCTP_OUTPUT_FROM_USR_SEND 0 #define SCTP_OUTPUT_FROM_T3 1 #define SCTP_OUTPUT_FROM_INPUT_ERROR 2 #define SCTP_OUTPUT_FROM_CONTROL_PROC 3 #define SCTP_OUTPUT_FROM_SACK_TMR 4 #define SCTP_OUTPUT_FROM_SHUT_TMR 5 #define SCTP_OUTPUT_FROM_HB_TMR 6 #define SCTP_OUTPUT_FROM_SHUT_ACK_TMR 7 #define SCTP_OUTPUT_FROM_ASCONF_TMR 8 #define SCTP_OUTPUT_FROM_STRRST_TMR 9 #define SCTP_OUTPUT_FROM_AUTOCLOSE_TMR 10 #define SCTP_OUTPUT_FROM_EARLY_FR_TMR 11 #define SCTP_OUTPUT_FROM_STRRST_REQ 12 #define SCTP_OUTPUT_FROM_USR_RCVD 13 #define SCTP_OUTPUT_FROM_COOKIE_ACK 14 #define SCTP_OUTPUT_FROM_DRAIN 15 #define SCTP_OUTPUT_FROM_CLOSING 16 /* SCTP chunk types are moved sctp.h for application (NAT, FW) use */ /* align to 32-bit sizes */ #define SCTP_SIZE32(x) ((((x)+3) >> 2) << 2) #define IS_SCTP_CONTROL(a) ((a)->chunk_type != SCTP_DATA) #define IS_SCTP_DATA(a) ((a)->chunk_type == SCTP_DATA) /* SCTP parameter types */ /*************0x0000 series*************/ #define SCTP_HEARTBEAT_INFO 0x0001 #define SCTP_IPV4_ADDRESS 0x0005 #define SCTP_IPV6_ADDRESS 0x0006 #define SCTP_STATE_COOKIE 0x0007 #define SCTP_UNRECOG_PARAM 0x0008 #define SCTP_COOKIE_PRESERVE 0x0009 #define SCTP_HOSTNAME_ADDRESS 0x000b #define SCTP_SUPPORTED_ADDRTYPE 0x000c /* draft-ietf-stewart-tsvwg-strreset-xxx */ #define SCTP_STR_RESET_OUT_REQUEST 0x000d #define SCTP_STR_RESET_IN_REQUEST 0x000e #define SCTP_STR_RESET_TSN_REQUEST 0x000f #define SCTP_STR_RESET_RESPONSE 0x0010 #define SCTP_STR_RESET_ADD_STREAMS 0x0011 #define SCTP_MAX_RESET_PARAMS 2 #define SCTP_STREAM_RESET_TSN_DELTA 0x1000 /*************0x4000 series*************/ /*************0x8000 series*************/ #define SCTP_ECN_CAPABLE 0x8000 -/* ECN Nonce: draft-ladha-sctp-ecn-nonce */ -#define SCTP_ECN_NONCE_SUPPORTED 0x8001 + /* draft-ietf-tsvwg-auth-xxx */ #define SCTP_RANDOM 0x8002 #define SCTP_CHUNK_LIST 0x8003 #define SCTP_HMAC_LIST 0x8004 /* * draft-ietf-tsvwg-addip-sctp-xx param=0x8008 len=0xNNNN Byte | Byte | Byte * | Byte Byte | Byte ... * * Where each byte is a chunk type extension supported. For example, to support * all chunks one would have (in hex): * * 80 01 00 09 C0 C1 80 81 82 00 00 00 * * Has the parameter. C0 = PR-SCTP (RFC3758) C1, 80 = ASCONF (addip draft) 81 * = Packet Drop 82 = Stream Reset 83 = Authentication */ #define SCTP_SUPPORTED_CHUNK_EXT 0x8008 /*************0xC000 series*************/ #define SCTP_PRSCTP_SUPPORTED 0xc000 /* draft-ietf-tsvwg-addip-sctp */ #define SCTP_ADD_IP_ADDRESS 0xc001 #define SCTP_DEL_IP_ADDRESS 0xc002 #define SCTP_ERROR_CAUSE_IND 0xc003 #define SCTP_SET_PRIM_ADDR 0xc004 #define SCTP_SUCCESS_REPORT 0xc005 #define SCTP_ULP_ADAPTATION 0xc006 /* behave-nat-draft */ #define SCTP_HAS_NAT_SUPPORT 0xc007 #define SCTP_NAT_VTAGS 0xc008 /* Notification error codes */ #define SCTP_NOTIFY_DATAGRAM_UNSENT 0x0001 #define SCTP_NOTIFY_DATAGRAM_SENT 0x0002 #define SCTP_FAILED_THRESHOLD 0x0004 #define SCTP_HEARTBEAT_SUCCESS 0x0008 #define SCTP_RESPONSE_TO_USER_REQ 0x0010 #define SCTP_INTERNAL_ERROR 0x0020 #define SCTP_SHUTDOWN_GUARD_EXPIRES 0x0040 #define SCTP_RECEIVED_SACK 0x0080 #define SCTP_PEER_FAULTY 0x0100 #define SCTP_ICMP_REFUSED 0x0200 /* bits for TOS field */ #define SCTP_ECT0_BIT 0x02 #define SCTP_ECT1_BIT 0x01 #define SCTP_CE_BITS 0x03 /* below turns off above */ #define SCTP_FLEXIBLE_ADDRESS 0x20 #define SCTP_NO_HEARTBEAT 0x40 /* mask to get sticky */ #define SCTP_STICKY_OPTIONS_MASK 0x0c /* * SCTP states for internal state machine XXX (should match "user" values) */ #define SCTP_STATE_EMPTY 0x0000 #define SCTP_STATE_INUSE 0x0001 #define SCTP_STATE_COOKIE_WAIT 0x0002 #define SCTP_STATE_COOKIE_ECHOED 0x0004 #define SCTP_STATE_OPEN 0x0008 #define SCTP_STATE_SHUTDOWN_SENT 0x0010 #define SCTP_STATE_SHUTDOWN_RECEIVED 0x0020 #define SCTP_STATE_SHUTDOWN_ACK_SENT 0x0040 #define SCTP_STATE_SHUTDOWN_PENDING 0x0080 #define SCTP_STATE_CLOSED_SOCKET 0x0100 #define SCTP_STATE_ABOUT_TO_BE_FREED 0x0200 #define SCTP_STATE_PARTIAL_MSG_LEFT 0x0400 #define SCTP_STATE_WAS_ABORTED 0x0800 #define SCTP_STATE_IN_ACCEPT_QUEUE 0x1000 #define SCTP_STATE_MASK 0x007f #define SCTP_GET_STATE(asoc) ((asoc)->state & SCTP_STATE_MASK) #define SCTP_SET_STATE(asoc, newstate) ((asoc)->state = ((asoc)->state & ~SCTP_STATE_MASK) | newstate) #define SCTP_CLEAR_SUBSTATE(asoc, substate) ((asoc)->state &= ~substate) #define SCTP_ADD_SUBSTATE(asoc, substate) ((asoc)->state |= substate) /* SCTP reachability state for each address */ #define SCTP_ADDR_REACHABLE 0x001 #define SCTP_ADDR_NOT_REACHABLE 0x002 #define SCTP_ADDR_NOHB 0x004 #define SCTP_ADDR_BEING_DELETED 0x008 #define SCTP_ADDR_NOT_IN_ASSOC 0x010 #define SCTP_ADDR_WAS_PRIMARY 0x020 #define SCTP_ADDR_SWITCH_PRIMARY 0x040 #define SCTP_ADDR_OUT_OF_SCOPE 0x080 #define SCTP_ADDR_DOUBLE_SWITCH 0x100 #define SCTP_ADDR_UNCONFIRMED 0x200 #define SCTP_ADDR_REQ_PRIMARY 0x400 /* JRS 5/13/07 - Added potentially failed state for CMT PF */ #define SCTP_ADDR_PF 0x800 /* bound address types (e.g. valid address types to allow) */ #define SCTP_BOUND_V6 0x01 #define SCTP_BOUND_V4 0x02 /* * what is the default number of mbufs in a chain I allow before switching to * a cluster */ #define SCTP_DEFAULT_MBUFS_IN_CHAIN 5 /* How long a cookie lives in milli-seconds */ #define SCTP_DEFAULT_COOKIE_LIFE 60000 /* resource limit of streams */ #define MAX_SCTP_STREAMS 2048 /* Maximum the mapping array will grow to (TSN mapping array) */ #define SCTP_MAPPING_ARRAY 512 /* size of the inital malloc on the mapping array */ #define SCTP_INITIAL_MAPPING_ARRAY 16 /* how much we grow the mapping array each call */ #define SCTP_MAPPING_ARRAY_INCR 32 /* * Here we define the timer types used by the implementation as arguments in * the set/get timer type calls. */ #define SCTP_TIMER_INIT 0 #define SCTP_TIMER_RECV 1 #define SCTP_TIMER_SEND 2 #define SCTP_TIMER_HEARTBEAT 3 #define SCTP_TIMER_PMTU 4 #define SCTP_TIMER_MAXSHUTDOWN 5 #define SCTP_TIMER_SIGNATURE 6 /* * number of timer types in the base SCTP structure used in the set/get and * has the base default. */ #define SCTP_NUM_TMRS 7 /* timer types */ #define SCTP_TIMER_TYPE_NONE 0 #define SCTP_TIMER_TYPE_SEND 1 #define SCTP_TIMER_TYPE_INIT 2 #define SCTP_TIMER_TYPE_RECV 3 #define SCTP_TIMER_TYPE_SHUTDOWN 4 #define SCTP_TIMER_TYPE_HEARTBEAT 5 #define SCTP_TIMER_TYPE_COOKIE 6 #define SCTP_TIMER_TYPE_NEWCOOKIE 7 #define SCTP_TIMER_TYPE_PATHMTURAISE 8 #define SCTP_TIMER_TYPE_SHUTDOWNACK 9 #define SCTP_TIMER_TYPE_ASCONF 10 #define SCTP_TIMER_TYPE_SHUTDOWNGUARD 11 #define SCTP_TIMER_TYPE_AUTOCLOSE 12 #define SCTP_TIMER_TYPE_EVENTWAKE 13 #define SCTP_TIMER_TYPE_STRRESET 14 #define SCTP_TIMER_TYPE_INPKILL 15 #define SCTP_TIMER_TYPE_EARLYFR 17 #define SCTP_TIMER_TYPE_ASOCKILL 18 #define SCTP_TIMER_TYPE_ADDR_WQ 19 #define SCTP_TIMER_TYPE_ZERO_COPY 20 #define SCTP_TIMER_TYPE_ZCOPY_SENDQ 21 #define SCTP_TIMER_TYPE_PRIM_DELETED 22 /* add new timers here - and increment LAST */ #define SCTP_TIMER_TYPE_LAST 23 #define SCTP_IS_TIMER_TYPE_VALID(t) (((t) > SCTP_TIMER_TYPE_NONE) && \ ((t) < SCTP_TIMER_TYPE_LAST)) /* max number of TSN's dup'd that I will hold */ #define SCTP_MAX_DUP_TSNS 20 /* * Here we define the types used when setting the retry amounts. */ /* How many drop re-attempts we make on INIT/COOKIE-ECHO */ #define SCTP_RETRY_DROPPED_THRESH 4 /* * Maxmium number of chunks a single association can have on it. Note that * this is a squishy number since the count can run over this if the user * sends a large message down .. the fragmented chunks don't count until * AFTER the message is on queue.. it would be the next send that blocks * things. This number will get tuned up at boot in the sctp_init and use the * number of clusters as a base. This way high bandwidth environments will * not get impacted by the lower bandwidth sending a bunch of 1 byte chunks */ #define SCTP_ASOC_MAX_CHUNKS_ON_QUEUE 512 /* The conversion from time to ticks and vice versa is done by rounding * upwards. This way we can test in the code the time to be positive and * know that this corresponds to a positive number of ticks. */ #define MSEC_TO_TICKS(x) ((hz == 1000) ? x : ((((x) * hz) + 999) / 1000)) #define TICKS_TO_MSEC(x) ((hz == 1000) ? x : ((((x) * 1000) + (hz - 1)) / hz)) #define SEC_TO_TICKS(x) ((x) * hz) #define TICKS_TO_SEC(x) (((x) + (hz - 1)) / hz) /* * Basically the minimum amount of time before I do a early FR. Making this * value to low will cause duplicate retransmissions. */ #define SCTP_MINFR_MSEC_TIMER 250 /* The floor this value is allowed to fall to when starting a timer. */ #define SCTP_MINFR_MSEC_FLOOR 20 /* init timer def = 1 sec */ #define SCTP_INIT_SEC 1 /* send timer def = 1 seconds */ #define SCTP_SEND_SEC 1 /* recv timer def = 200ms */ #define SCTP_RECV_MSEC 200 /* 30 seconds + RTO (in ms) */ #define SCTP_HB_DEFAULT_MSEC 30000 /* Max time I will wait for Shutdown to complete */ #define SCTP_DEF_MAX_SHUTDOWN_SEC 180 /* * This is how long a secret lives, NOT how long a cookie lives how many * ticks the current secret will live. */ #define SCTP_DEFAULT_SECRET_LIFE_SEC 3600 #define SCTP_RTO_UPPER_BOUND (60000) /* 60 sec in ms */ #define SCTP_RTO_LOWER_BOUND (300) /* 0.3 sec is ms */ #define SCTP_RTO_INITIAL (3000) /* 3 sec in ms */ #define SCTP_INP_KILL_TIMEOUT 20/* number of ms to retry kill of inpcb */ #define SCTP_ASOC_KILL_TIMEOUT 10 /* number of ms to retry kill of inpcb */ #define SCTP_DEF_MAX_INIT 8 #define SCTP_DEF_MAX_SEND 10 #define SCTP_DEF_MAX_PATH_RTX 5 #define SCTP_DEF_PMTU_RAISE_SEC 600 /* 10 min between raise attempts */ /* How many streams I request initally by default */ #define SCTP_OSTREAM_INITIAL 10 /* * How many smallest_mtu's need to increase before a window update sack is * sent (should be a power of 2). */ /* Send window update (incr * this > hiwat). Should be a power of 2 */ #define SCTP_MINIMAL_RWND (4096) /* minimal rwnd */ #define SCTP_ADDRMAX 24 /* SCTP DEBUG Switch parameters */ #define SCTP_DEBUG_TIMER1 0x00000001 #define SCTP_DEBUG_TIMER2 0x00000002 /* unused */ #define SCTP_DEBUG_TIMER3 0x00000004 /* unused */ #define SCTP_DEBUG_TIMER4 0x00000008 #define SCTP_DEBUG_OUTPUT1 0x00000010 #define SCTP_DEBUG_OUTPUT2 0x00000020 #define SCTP_DEBUG_OUTPUT3 0x00000040 #define SCTP_DEBUG_OUTPUT4 0x00000080 #define SCTP_DEBUG_UTIL1 0x00000100 #define SCTP_DEBUG_UTIL2 0x00000200 /* unused */ #define SCTP_DEBUG_AUTH1 0x00000400 #define SCTP_DEBUG_AUTH2 0x00000800 /* unused */ #define SCTP_DEBUG_INPUT1 0x00001000 #define SCTP_DEBUG_INPUT2 0x00002000 #define SCTP_DEBUG_INPUT3 0x00004000 #define SCTP_DEBUG_INPUT4 0x00008000 /* unused */ #define SCTP_DEBUG_ASCONF1 0x00010000 #define SCTP_DEBUG_ASCONF2 0x00020000 #define SCTP_DEBUG_OUTPUT5 0x00040000 /* unused */ #define SCTP_DEBUG_XXX 0x00080000 /* unused */ #define SCTP_DEBUG_PCB1 0x00100000 #define SCTP_DEBUG_PCB2 0x00200000 /* unused */ #define SCTP_DEBUG_PCB3 0x00400000 #define SCTP_DEBUG_PCB4 0x00800000 #define SCTP_DEBUG_INDATA1 0x01000000 #define SCTP_DEBUG_INDATA2 0x02000000 /* unused */ #define SCTP_DEBUG_INDATA3 0x04000000 /* unused */ #define SCTP_DEBUG_CRCOFFLOAD 0x08000000 /* unused */ #define SCTP_DEBUG_USRREQ1 0x10000000 /* unused */ #define SCTP_DEBUG_USRREQ2 0x20000000 /* unused */ #define SCTP_DEBUG_PEEL1 0x40000000 #define SCTP_DEBUG_XXXXX 0x80000000 /* unused */ #define SCTP_DEBUG_ALL 0x7ff3ffff #define SCTP_DEBUG_NOISY 0x00040000 /* What sender needs to see to avoid SWS or we consider peers rwnd 0 */ #define SCTP_SWS_SENDER_DEF 1420 /* * SWS is scaled to the sb_hiwat of the socket. A value of 2 is hiwat/4, 1 * would be hiwat/2 etc. */ /* What receiver needs to see in sockbuf or we tell peer its 1 */ #define SCTP_SWS_RECEIVER_DEF 3000 #define SCTP_INITIAL_CWND 4380 #define SCTP_DEFAULT_MTU 1500 /* emergency default MTU */ /* amount peer is obligated to have in rwnd or I will abort */ #define SCTP_MIN_RWND 1500 #define SCTP_DEFAULT_MAXSEGMENT 65535 #define SCTP_CHUNK_BUFFER_SIZE 512 #define SCTP_PARAM_BUFFER_SIZE 512 /* small chunk store for looking at chunk_list in auth */ #define SCTP_SMALL_CHUNK_STORE 260 #define SCTP_DEFAULT_MINSEGMENT 512 /* MTU size ... if no mtu disc */ #define SCTP_HOW_MANY_SECRETS 2 /* how many secrets I keep */ #define SCTP_NUMBER_OF_SECRETS 8 /* or 8 * 4 = 32 octets */ #define SCTP_SECRET_SIZE 32 /* number of octets in a 256 bits */ /* * SCTP upper layer notifications */ #define SCTP_NOTIFY_ASSOC_UP 1 #define SCTP_NOTIFY_ASSOC_DOWN 2 #define SCTP_NOTIFY_INTERFACE_DOWN 3 #define SCTP_NOTIFY_INTERFACE_UP 4 #define SCTP_NOTIFY_DG_FAIL 5 #define SCTP_NOTIFY_STRDATA_ERR 6 #define SCTP_NOTIFY_ASSOC_ABORTED 7 #define SCTP_NOTIFY_PEER_OPENED_STREAM 8 #define SCTP_NOTIFY_STREAM_OPENED_OK 9 #define SCTP_NOTIFY_ASSOC_RESTART 10 #define SCTP_NOTIFY_HB_RESP 11 #define SCTP_NOTIFY_ASCONF_SUCCESS 12 #define SCTP_NOTIFY_ASCONF_FAILED 13 #define SCTP_NOTIFY_PEER_SHUTDOWN 14 #define SCTP_NOTIFY_ASCONF_ADD_IP 15 #define SCTP_NOTIFY_ASCONF_DELETE_IP 16 #define SCTP_NOTIFY_ASCONF_SET_PRIMARY 17 #define SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION 18 #define SCTP_NOTIFY_INTERFACE_CONFIRMED 20 #define SCTP_NOTIFY_STR_RESET_RECV 21 #define SCTP_NOTIFY_STR_RESET_SEND 22 #define SCTP_NOTIFY_STR_RESET_FAILED_OUT 23 #define SCTP_NOTIFY_STR_RESET_FAILED_IN 24 #define SCTP_NOTIFY_AUTH_NEW_KEY 25 #define SCTP_NOTIFY_AUTH_FREE_KEY 26 #define SCTP_NOTIFY_SPECIAL_SP_FAIL 27 #define SCTP_NOTIFY_NO_PEER_AUTH 28 #define SCTP_NOTIFY_SENDER_DRY 29 #define SCTP_NOTIFY_STR_RESET_ADD_OK 30 #define SCTP_NOTIFY_STR_RESET_ADD_FAIL 31 #define SCTP_NOTIFY_STR_RESET_INSTREAM_ADD_OK 32 #define SCTP_NOTIFY_MAX 32 /* This is the value for messages that are NOT completely * copied down where we will start to split the message. * So, with our default, we split only if the piece we * want to take will fill up a full MTU (assuming * a 1500 byte MTU). */ #define SCTP_DEFAULT_SPLIT_POINT_MIN 2904 /* ABORT CODES and other tell-tale location * codes are generated by adding the below * to the instance id. */ /* File defines */ #define SCTP_FROM_SCTP_INPUT 0x10000000 #define SCTP_FROM_SCTP_PCB 0x20000000 #define SCTP_FROM_SCTP_INDATA 0x30000000 #define SCTP_FROM_SCTP_TIMER 0x40000000 #define SCTP_FROM_SCTP_USRREQ 0x50000000 #define SCTP_FROM_SCTPUTIL 0x60000000 #define SCTP_FROM_SCTP6_USRREQ 0x70000000 #define SCTP_FROM_SCTP_ASCONF 0x80000000 #define SCTP_FROM_SCTP_OUTPUT 0x90000000 #define SCTP_FROM_SCTP_PEELOFF 0xa0000000 #define SCTP_FROM_SCTP_PANDA 0xb0000000 #define SCTP_FROM_SCTP_SYSCTL 0xc0000000 /* Location ID's */ #define SCTP_LOC_1 0x00000001 #define SCTP_LOC_2 0x00000002 #define SCTP_LOC_3 0x00000003 #define SCTP_LOC_4 0x00000004 #define SCTP_LOC_5 0x00000005 #define SCTP_LOC_6 0x00000006 #define SCTP_LOC_7 0x00000007 #define SCTP_LOC_8 0x00000008 #define SCTP_LOC_9 0x00000009 #define SCTP_LOC_10 0x0000000a #define SCTP_LOC_11 0x0000000b #define SCTP_LOC_12 0x0000000c #define SCTP_LOC_13 0x0000000d #define SCTP_LOC_14 0x0000000e #define SCTP_LOC_15 0x0000000f #define SCTP_LOC_16 0x00000010 #define SCTP_LOC_17 0x00000011 #define SCTP_LOC_18 0x00000012 #define SCTP_LOC_19 0x00000013 #define SCTP_LOC_20 0x00000014 #define SCTP_LOC_21 0x00000015 #define SCTP_LOC_22 0x00000016 #define SCTP_LOC_23 0x00000017 #define SCTP_LOC_24 0x00000018 #define SCTP_LOC_25 0x00000019 #define SCTP_LOC_26 0x0000001a #define SCTP_LOC_27 0x0000001b #define SCTP_LOC_28 0x0000001c #define SCTP_LOC_29 0x0000001d #define SCTP_LOC_30 0x0000001e #define SCTP_LOC_31 0x0000001f #define SCTP_LOC_32 0x00000020 #define SCTP_LOC_33 0x00000021 /* Free assoc codes */ #define SCTP_NORMAL_PROC 0 #define SCTP_PCBFREE_NOFORCE 1 #define SCTP_PCBFREE_FORCE 2 /* From codes for adding addresses */ #define SCTP_ADDR_IS_CONFIRMED 8 #define SCTP_ADDR_DYNAMIC_ADDED 6 #define SCTP_IN_COOKIE_PROC 100 #define SCTP_ALLOC_ASOC 1 #define SCTP_LOAD_ADDR_2 2 #define SCTP_LOAD_ADDR_3 3 #define SCTP_LOAD_ADDR_4 4 #define SCTP_LOAD_ADDR_5 5 #define SCTP_DONOT_SETSCOPE 0 #define SCTP_DO_SETSCOPE 1 /* This value determines the default for when * we try to add more on the send queue., if * there is room. This prevents us from cycling * into the copy_resume routine to often if * we have not got enough space to add a decent * enough size message. Note that if we have enough * space to complete the message copy we will always * add to the message, no matter what the size. Its * only when we reach the point that we have some left * to add, there is only room for part of it that we * will use this threshold. Its also a sysctl. */ #define SCTP_DEFAULT_ADD_MORE 1452 #ifndef SCTP_PCBHASHSIZE /* default number of association hash buckets in each endpoint */ #define SCTP_PCBHASHSIZE 256 #endif #ifndef SCTP_TCBHASHSIZE #define SCTP_TCBHASHSIZE 1024 #endif #ifndef SCTP_CHUNKQUEUE_SCALE #define SCTP_CHUNKQUEUE_SCALE 10 #endif /* clock variance is 1 ms */ #define SCTP_CLOCK_GRANULARITY 1 #define IP_HDR_SIZE 40 /* we use the size of a IP6 header here this * detracts a small amount for ipv4 but it * simplifies the ipv6 addition */ /* Argument magic number for sctp_inpcb_free() */ /* third argument */ #define SCTP_CALLED_DIRECTLY_NOCMPSET 0 #define SCTP_CALLED_AFTER_CMPSET_OFCLOSE 1 #define SCTP_CALLED_FROM_INPKILL_TIMER 2 /* second argument */ #define SCTP_FREE_SHOULD_USE_ABORT 1 #define SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE 0 #ifndef IPPROTO_SCTP #define IPPROTO_SCTP 132 /* the Official IANA number :-) */ #endif /* !IPPROTO_SCTP */ #define SCTP_MAX_DATA_BUNDLING 256 /* modular comparison */ /* See RFC 1982 for details. */ #define SCTP_SSN_GT(a, b) (((a < b) && ((b - a) > (1<<15))) || \ ((a > b) && ((a - b) < (1<<15)))) #define SCTP_SSN_GE(a, b) (SCTP_SSN_GT(a, b) || (a == b)) #define SCTP_TSN_GT(a, b) (((a < b) && ((b - a) > (1<<31))) || \ ((a > b) && ((a - b) < (1<<31)))) #define SCTP_TSN_GE(a, b) (SCTP_TSN_GT(a, b) || (a == b)) /* Mapping array manipulation routines */ #define SCTP_IS_TSN_PRESENT(arry, gap) ((arry[(gap >> 3)] >> (gap & 0x07)) & 0x01) #define SCTP_SET_TSN_PRESENT(arry, gap) (arry[(gap >> 3)] |= (0x01 << ((gap & 0x07)))) #define SCTP_UNSET_TSN_PRESENT(arry, gap) (arry[(gap >> 3)] &= ((~(0x01 << ((gap & 0x07)))) & 0xff)) #define SCTP_CALC_TSN_TO_GAP(gap, tsn, mapping_tsn) do { \ if (tsn >= mapping_tsn) { \ gap = tsn - mapping_tsn; \ } else { \ gap = (MAX_TSN - mapping_tsn) + tsn + 1; \ } \ } while(0) #define SCTP_RETRAN_DONE -1 #define SCTP_RETRAN_EXIT -2 /* * This value defines the number of vtag block time wait entry's per list * element. Each entry will take 2 4 byte ints (and of course the overhead * of the next pointer as well). Using 15 as an example will yield * ((8 * * 15) + 8) or 128 bytes of overhead for each timewait block that gets * initialized. Increasing it to 31 would yeild 256 bytes per block. */ #define SCTP_NUMBER_IN_VTAG_BLOCK 15 /* * If we use the STACK option, we have an array of this size head pointers. * This array is mod'd the with the size to find which bucket and then all * entries must be searched to see if the tag is in timed wait. If so we * reject it. */ #define SCTP_STACK_VTAG_HASH_SIZE 32 /* * Number of seconds of time wait for a vtag. */ #define SCTP_TIME_WAIT 60 #define SCTP_SEND_BUFFER_SPLITTING 0x00000001 #define SCTP_RECV_BUFFER_SPLITTING 0x00000002 /* The system retains a cache of free chunks such to * cut down on calls the memory allocation system. There * is a per association limit of free items and a overall * system limit. If either one gets hit then the resource * stops being cached. */ #define SCTP_DEF_ASOC_RESC_LIMIT 10 #define SCTP_DEF_SYSTEM_RESC_LIMIT 1000 /*- * defines for socket lock states. * Used by __APPLE__ and SCTP_SO_LOCK_TESTING */ #define SCTP_SO_LOCKED 1 #define SCTP_SO_NOT_LOCKED 0 #define SCTP_HOLDS_LOCK 1 #define SCTP_NOT_LOCKED 0 /*- * For address locks, do we hold the lock? */ #define SCTP_ADDR_LOCKED 1 #define SCTP_ADDR_NOT_LOCKED 0 #define IN4_ISPRIVATE_ADDRESS(a) \ ((((uint8_t *)&(a)->s_addr)[0] == 10) || \ ((((uint8_t *)&(a)->s_addr)[0] == 172) && \ (((uint8_t *)&(a)->s_addr)[1] >= 16) && \ (((uint8_t *)&(a)->s_addr)[1] <= 32)) || \ ((((uint8_t *)&(a)->s_addr)[0] == 192) && \ (((uint8_t *)&(a)->s_addr)[1] == 168))) #define IN4_ISLOOPBACK_ADDRESS(a) \ ((((uint8_t *)&(a)->s_addr)[0] == 127) && \ (((uint8_t *)&(a)->s_addr)[1] == 0) && \ (((uint8_t *)&(a)->s_addr)[2] == 0) && \ (((uint8_t *)&(a)->s_addr)[3] == 1)) #if defined(_KERNEL) #define SCTP_GETTIME_TIMEVAL(x) (getmicrouptime(x)) #define SCTP_GETPTIME_TIMEVAL(x) (microuptime(x)) #endif /*#if defined(__FreeBSD__) || defined(__APPLE__)*/ /*#define SCTP_GETTIME_TIMEVAL(x) { \*/ /* (x)->tv_sec = ticks / 1000; \*/ /* (x)->tv_usec = (ticks % 1000) * 1000; \*/ /*}*/ /*#else*/ /*#define SCTP_GETTIME_TIMEVAL(x) (microtime(x))*/ /*#endif __FreeBSD__ */ #if defined(_KERNEL) || defined(__Userspace__) #define sctp_sowwakeup(inp, so) \ do { \ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ inp->sctp_flags |= SCTP_PCB_FLAGS_WAKEOUTPUT; \ } else { \ sowwakeup(so); \ } \ } while (0) #define sctp_sowwakeup_locked(inp, so) \ do { \ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ SOCKBUF_UNLOCK(&((so)->so_snd)); \ inp->sctp_flags |= SCTP_PCB_FLAGS_WAKEOUTPUT; \ } else { \ sowwakeup_locked(so); \ } \ } while (0) #define sctp_sorwakeup(inp, so) \ do { \ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ inp->sctp_flags |= SCTP_PCB_FLAGS_WAKEINPUT; \ } else { \ sorwakeup(so); \ } \ } while (0) #define sctp_sorwakeup_locked(inp, so) \ do { \ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ inp->sctp_flags |= SCTP_PCB_FLAGS_WAKEINPUT; \ SOCKBUF_UNLOCK(&((so)->so_rcv)); \ } else { \ sorwakeup_locked(so); \ } \ } while (0) #endif /* _KERNEL || __Userspace__ */ #endif diff --git a/sys/netinet/sctp_header.h b/sys/netinet/sctp_header.h index 2e94fb73a0f6..4a33b0d97e70 100644 --- a/sys/netinet/sctp_header.h +++ b/sys/netinet/sctp_header.h @@ -1,630 +1,624 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_header.h,v 1.14 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_header_h__ #define __sctp_header_h__ #include #include #include #define SCTP_PACKED __attribute__((packed)) /* * Parameter structures */ struct sctp_ipv4addr_param { struct sctp_paramhdr ph;/* type=SCTP_IPV4_PARAM_TYPE, len=8 */ uint32_t addr; /* IPV4 address */ } SCTP_PACKED; #define SCTP_V6_ADDR_BYTES 16 struct sctp_ipv6addr_param { struct sctp_paramhdr ph;/* type=SCTP_IPV6_PARAM_TYPE, len=20 */ uint8_t addr[SCTP_V6_ADDR_BYTES]; /* IPV6 address */ } SCTP_PACKED; /* Cookie Preservative */ struct sctp_cookie_perserve_param { struct sctp_paramhdr ph;/* type=SCTP_COOKIE_PRESERVE, len=8 */ uint32_t time; /* time in ms to extend cookie */ } SCTP_PACKED; #define SCTP_ARRAY_MIN_LEN 1 /* Host Name Address */ struct sctp_host_name_param { struct sctp_paramhdr ph;/* type=SCTP_HOSTNAME_ADDRESS */ char name[SCTP_ARRAY_MIN_LEN]; /* host name */ } SCTP_PACKED; /* * This is the maximum padded size of a s-a-p * so paramheadr + 3 address types (6 bytes) + 2 byte pad = 12 */ #define SCTP_MAX_ADDR_PARAMS_SIZE 12 /* supported address type */ struct sctp_supported_addr_param { struct sctp_paramhdr ph;/* type=SCTP_SUPPORTED_ADDRTYPE */ uint16_t addr_type[SCTP_ARRAY_MIN_LEN]; /* array of supported address * types */ } SCTP_PACKED; /* ECN parameter */ struct sctp_ecn_supported_param { struct sctp_paramhdr ph;/* type=SCTP_ECN_CAPABLE */ } SCTP_PACKED; /* heartbeat info parameter */ struct sctp_heartbeat_info_param { struct sctp_paramhdr ph; uint32_t time_value_1; uint32_t time_value_2; uint32_t random_value1; uint32_t random_value2; uint16_t user_req; uint8_t addr_family; uint8_t addr_len; char address[SCTP_ADDRMAX]; } SCTP_PACKED; /* draft-ietf-tsvwg-prsctp */ /* PR-SCTP supported parameter */ struct sctp_prsctp_supported_param { struct sctp_paramhdr ph; } SCTP_PACKED; /* draft-ietf-tsvwg-addip-sctp */ struct sctp_asconf_paramhdr { /* an ASCONF "parameter" */ struct sctp_paramhdr ph;/* a SCTP parameter header */ uint32_t correlation_id;/* correlation id for this param */ } SCTP_PACKED; struct sctp_asconf_addr_param { /* an ASCONF address parameter */ struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ struct sctp_ipv6addr_param addrp; /* max storage size */ } SCTP_PACKED; struct sctp_asconf_tag_param { /* an ASCONF NAT-Vtag parameter */ struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ uint32_t local_vtag; uint32_t remote_vtag; } SCTP_PACKED; struct sctp_asconf_addrv4_param { /* an ASCONF address (v4) parameter */ struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ struct sctp_ipv4addr_param addrp; /* max storage size */ } SCTP_PACKED; #define SCTP_MAX_SUPPORTED_EXT 256 struct sctp_supported_chunk_types_param { struct sctp_paramhdr ph;/* type = 0x8008 len = x */ uint8_t chunk_types[]; } SCTP_PACKED; -/* ECN Nonce: draft-ladha-sctp-ecn-nonce */ -struct sctp_ecn_nonce_supported_param { - struct sctp_paramhdr ph;/* type = 0x8001 len = 4 */ -} SCTP_PACKED; - - /* * Structures for DATA chunks */ struct sctp_data { uint32_t tsn; uint16_t stream_id; uint16_t stream_sequence; uint32_t protocol_id; /* user data follows */ } SCTP_PACKED; struct sctp_data_chunk { struct sctp_chunkhdr ch; struct sctp_data dp; } SCTP_PACKED; /* * Structures for the control chunks */ /* Initiate (INIT)/Initiate Ack (INIT ACK) */ struct sctp_init { uint32_t initiate_tag; /* initiate tag */ uint32_t a_rwnd; /* a_rwnd */ uint16_t num_outbound_streams; /* OS */ uint16_t num_inbound_streams; /* MIS */ uint32_t initial_tsn; /* I-TSN */ /* optional param's follow */ } SCTP_PACKED; #define SCTP_IDENTIFICATION_SIZE 16 #define SCTP_ADDRESS_SIZE 4 #define SCTP_RESERVE_SPACE 6 /* state cookie header */ struct sctp_state_cookie { /* this is our definition... */ uint8_t identification[SCTP_IDENTIFICATION_SIZE]; /* id of who we are */ struct timeval time_entered; /* the time I built cookie */ uint32_t cookie_life; /* life I will award this cookie */ uint32_t tie_tag_my_vtag; /* my tag in old association */ uint32_t tie_tag_peer_vtag; /* peers tag in old association */ uint32_t peers_vtag; /* peers tag in INIT (for quick ref) */ uint32_t my_vtag; /* my tag in INIT-ACK (for quick ref) */ uint32_t address[SCTP_ADDRESS_SIZE]; /* 4 ints/128 bits */ uint32_t addr_type; /* address type */ uint32_t laddress[SCTP_ADDRESS_SIZE]; /* my local from address */ uint32_t laddr_type; /* my local from address type */ uint32_t scope_id; /* v6 scope id for link-locals */ uint16_t peerport; /* port address of the peer in the INIT */ uint16_t myport; /* my port address used in the INIT */ uint8_t ipv4_addr_legal;/* Are V4 addr legal? */ uint8_t ipv6_addr_legal;/* Are V6 addr legal? */ uint8_t local_scope; /* IPv6 local scope flag */ uint8_t site_scope; /* IPv6 site scope flag */ uint8_t ipv4_scope; /* IPv4 private addr scope */ uint8_t loopback_scope; /* loopback scope information */ uint8_t reserved[SCTP_RESERVE_SPACE]; /* Align to 64 bits */ /* * at the end is tacked on the INIT chunk and the INIT-ACK chunk * (minus the cookie). */ } SCTP_PACKED; /* Used for NAT state error cause */ struct sctp_missing_nat_state { uint16_t cause; uint16_t length; uint8_t data[]; } SCTP_PACKED; struct sctp_inv_mandatory_param { uint16_t cause; uint16_t length; uint32_t num_param; uint16_t param; /* * We include this to 0 it since only a missing cookie will cause * this error. */ uint16_t resv; } SCTP_PACKED; struct sctp_unresolv_addr { uint16_t cause; uint16_t length; uint16_t addr_type; uint16_t reserved; /* Only one invalid addr type */ } SCTP_PACKED; /* state cookie parameter */ struct sctp_state_cookie_param { struct sctp_paramhdr ph; struct sctp_state_cookie cookie; } SCTP_PACKED; struct sctp_init_chunk { struct sctp_chunkhdr ch; struct sctp_init init; } SCTP_PACKED; struct sctp_init_msg { struct sctphdr sh; struct sctp_init_chunk msg; } SCTP_PACKED; /* ... used for both INIT and INIT ACK */ #define sctp_init_ack sctp_init #define sctp_init_ack_chunk sctp_init_chunk #define sctp_init_ack_msg sctp_init_msg /* Selective Ack (SACK) */ struct sctp_gap_ack_block { uint16_t start; /* Gap Ack block start */ uint16_t end; /* Gap Ack block end */ } SCTP_PACKED; struct sctp_sack { uint32_t cum_tsn_ack; /* cumulative TSN Ack */ uint32_t a_rwnd; /* updated a_rwnd of sender */ uint16_t num_gap_ack_blks; /* number of Gap Ack blocks */ uint16_t num_dup_tsns; /* number of duplicate TSNs */ /* struct sctp_gap_ack_block's follow */ /* uint32_t duplicate_tsn's follow */ } SCTP_PACKED; struct sctp_sack_chunk { struct sctp_chunkhdr ch; struct sctp_sack sack; } SCTP_PACKED; struct sctp_nr_sack { uint32_t cum_tsn_ack; /* cumulative TSN Ack */ uint32_t a_rwnd; /* updated a_rwnd of sender */ uint16_t num_gap_ack_blks; /* number of Gap Ack blocks */ uint16_t num_nr_gap_ack_blks; /* number of NR Gap Ack blocks */ uint16_t num_dup_tsns; /* number of duplicate TSNs */ uint16_t reserved; /* not currently used */ /* struct sctp_gap_ack_block's follow */ /* uint32_t duplicate_tsn's follow */ } SCTP_PACKED; struct sctp_nr_sack_chunk { struct sctp_chunkhdr ch; struct sctp_nr_sack nr_sack; } SCTP_PACKED; /* Heartbeat Request (HEARTBEAT) */ struct sctp_heartbeat { struct sctp_heartbeat_info_param hb_info; } SCTP_PACKED; struct sctp_heartbeat_chunk { struct sctp_chunkhdr ch; struct sctp_heartbeat heartbeat; } SCTP_PACKED; /* ... used for Heartbeat Ack (HEARTBEAT ACK) */ #define sctp_heartbeat_ack sctp_heartbeat #define sctp_heartbeat_ack_chunk sctp_heartbeat_chunk /* Abort Asssociation (ABORT) */ struct sctp_abort_chunk { struct sctp_chunkhdr ch; /* optional error cause may follow */ } SCTP_PACKED; struct sctp_abort_msg { struct sctphdr sh; struct sctp_abort_chunk msg; } SCTP_PACKED; /* Shutdown Association (SHUTDOWN) */ struct sctp_shutdown_chunk { struct sctp_chunkhdr ch; uint32_t cumulative_tsn_ack; } SCTP_PACKED; /* Shutdown Acknowledgment (SHUTDOWN ACK) */ struct sctp_shutdown_ack_chunk { struct sctp_chunkhdr ch; } SCTP_PACKED; /* Operation Error (ERROR) */ struct sctp_error_chunk { struct sctp_chunkhdr ch; /* optional error causes follow */ } SCTP_PACKED; /* Cookie Echo (COOKIE ECHO) */ struct sctp_cookie_echo_chunk { struct sctp_chunkhdr ch; struct sctp_state_cookie cookie; } SCTP_PACKED; /* Cookie Acknowledgment (COOKIE ACK) */ struct sctp_cookie_ack_chunk { struct sctp_chunkhdr ch; } SCTP_PACKED; /* Explicit Congestion Notification Echo (ECNE) */ struct old_sctp_ecne_chunk { struct sctp_chunkhdr ch; uint32_t tsn; } SCTP_PACKED; struct sctp_ecne_chunk { struct sctp_chunkhdr ch; uint32_t tsn; uint32_t num_pkts_since_cwr; } SCTP_PACKED; /* Congestion Window Reduced (CWR) */ struct sctp_cwr_chunk { struct sctp_chunkhdr ch; uint32_t tsn; } SCTP_PACKED; /* Shutdown Complete (SHUTDOWN COMPLETE) */ struct sctp_shutdown_complete_chunk { struct sctp_chunkhdr ch; } SCTP_PACKED; /* Oper error holding a stale cookie */ struct sctp_stale_cookie_msg { struct sctp_paramhdr ph;/* really an error cause */ uint32_t time_usec; } SCTP_PACKED; struct sctp_adaptation_layer_indication { struct sctp_paramhdr ph; uint32_t indication; } SCTP_PACKED; struct sctp_cookie_while_shutting_down { struct sctphdr sh; struct sctp_chunkhdr ch; struct sctp_paramhdr ph;/* really an error cause */ } SCTP_PACKED; struct sctp_shutdown_complete_msg { struct sctphdr sh; struct sctp_shutdown_complete_chunk shut_cmp; } SCTP_PACKED; /* * draft-ietf-tsvwg-addip-sctp */ /* Address/Stream Configuration Change (ASCONF) */ struct sctp_asconf_chunk { struct sctp_chunkhdr ch; uint32_t serial_number; /* lookup address parameter (mandatory) */ /* asconf parameters follow */ } SCTP_PACKED; /* Address/Stream Configuration Acknowledge (ASCONF ACK) */ struct sctp_asconf_ack_chunk { struct sctp_chunkhdr ch; uint32_t serial_number; /* asconf parameters follow */ } SCTP_PACKED; /* draft-ietf-tsvwg-prsctp */ /* Forward Cumulative TSN (FORWARD TSN) */ struct sctp_forward_tsn_chunk { struct sctp_chunkhdr ch; uint32_t new_cumulative_tsn; /* stream/sequence pairs (sctp_strseq) follow */ } SCTP_PACKED; struct sctp_strseq { uint16_t stream; uint16_t sequence; } SCTP_PACKED; struct sctp_forward_tsn_msg { struct sctphdr sh; struct sctp_forward_tsn_chunk msg; } SCTP_PACKED; /* should be a multiple of 4 - 1 aka 3/7/11 etc. */ #define SCTP_NUM_DB_TO_VERIFY 31 struct sctp_chunk_desc { uint8_t chunk_type; uint8_t data_bytes[SCTP_NUM_DB_TO_VERIFY]; uint32_t tsn_ifany; } SCTP_PACKED; struct sctp_pktdrop_chunk { struct sctp_chunkhdr ch; uint32_t bottle_bw; uint32_t current_onq; uint16_t trunc_len; uint16_t reserved; uint8_t data[]; } SCTP_PACKED; /**********STREAM RESET STUFF ******************/ struct sctp_stream_reset_out_request { struct sctp_paramhdr ph; uint32_t request_seq; /* monotonically increasing seq no */ uint32_t response_seq; /* if a response, the resp seq no */ uint32_t send_reset_at_tsn; /* last TSN I assigned outbound */ uint16_t list_of_streams[]; /* if not all list of streams */ } SCTP_PACKED; struct sctp_stream_reset_in_request { struct sctp_paramhdr ph; uint32_t request_seq; uint16_t list_of_streams[]; /* if not all list of streams */ } SCTP_PACKED; struct sctp_stream_reset_tsn_request { struct sctp_paramhdr ph; uint32_t request_seq; } SCTP_PACKED; struct sctp_stream_reset_response { struct sctp_paramhdr ph; uint32_t response_seq; /* if a response, the resp seq no */ uint32_t result; } SCTP_PACKED; struct sctp_stream_reset_response_tsn { struct sctp_paramhdr ph; uint32_t response_seq; /* if a response, the resp seq no */ uint32_t result; uint32_t senders_next_tsn; uint32_t receivers_next_tsn; } SCTP_PACKED; struct sctp_stream_reset_add_strm { struct sctp_paramhdr ph; uint32_t request_seq; uint16_t number_of_streams; uint16_t reserved; } SCTP_PACKED; #define SCTP_STREAM_RESET_NOTHING 0x00000000 /* Nothing for me to do */ #define SCTP_STREAM_RESET_PERFORMED 0x00000001 /* Did it */ #define SCTP_STREAM_RESET_DENIED 0x00000002 /* refused to do it */ #define SCTP_STREAM_RESET_ERROR_STR 0x00000003 /* bad Stream no */ #define SCTP_STREAM_RESET_TRY_LATER 0x00000004 /* collision, try again */ #define SCTP_STREAM_RESET_BAD_SEQNO 0x00000005 /* bad str-reset seq no */ /* * convience structures, note that if you are making a request for specific * streams then the request will need to be an overlay structure. */ struct sctp_stream_reset_out_req { struct sctp_chunkhdr ch; struct sctp_stream_reset_out_request sr_req; } SCTP_PACKED; struct sctp_stream_reset_in_req { struct sctp_chunkhdr ch; struct sctp_stream_reset_in_request sr_req; } SCTP_PACKED; struct sctp_stream_reset_tsn_req { struct sctp_chunkhdr ch; struct sctp_stream_reset_tsn_request sr_req; } SCTP_PACKED; struct sctp_stream_reset_resp { struct sctp_chunkhdr ch; struct sctp_stream_reset_response sr_resp; } SCTP_PACKED; /* respone only valid with a TSN request */ struct sctp_stream_reset_resp_tsn { struct sctp_chunkhdr ch; struct sctp_stream_reset_response_tsn sr_resp; } SCTP_PACKED; /****************************************************/ /* * Authenticated chunks support draft-ietf-tsvwg-sctp-auth */ /* Should we make the max be 32? */ #define SCTP_RANDOM_MAX_SIZE 256 struct sctp_auth_random { struct sctp_paramhdr ph;/* type = 0x8002 */ uint8_t random_data[]; } SCTP_PACKED; struct sctp_auth_chunk_list { struct sctp_paramhdr ph;/* type = 0x8003 */ uint8_t chunk_types[]; } SCTP_PACKED; struct sctp_auth_hmac_algo { struct sctp_paramhdr ph;/* type = 0x8004 */ uint16_t hmac_ids[]; } SCTP_PACKED; struct sctp_auth_chunk { struct sctp_chunkhdr ch; uint16_t shared_key_id; uint16_t hmac_id; uint8_t hmac[]; } SCTP_PACKED; struct sctp_auth_invalid_hmac { struct sctp_paramhdr ph; uint16_t hmac_id; uint16_t padding; } SCTP_PACKED; /* * we pre-reserve enough room for a ECNE or CWR AND a SACK with no missing * pieces. If ENCE is missing we could have a couple of blocks. This way we * optimize so we MOST likely can bundle a SACK/ECN with the smallest size * data chunk I will split into. We could increase throughput slightly by * taking out these two but the 24-sack/8-CWR i.e. 32 bytes I pre-reserve I * feel is worth it for now. */ #ifndef SCTP_MAX_OVERHEAD #ifdef INET6 #define SCTP_MAX_OVERHEAD (sizeof(struct sctp_data_chunk) + \ sizeof(struct sctphdr) + \ sizeof(struct sctp_ecne_chunk) + \ sizeof(struct sctp_sack_chunk) + \ sizeof(struct ip6_hdr)) #define SCTP_MED_OVERHEAD (sizeof(struct sctp_data_chunk) + \ sizeof(struct sctphdr) + \ sizeof(struct ip6_hdr)) #define SCTP_MIN_OVERHEAD (sizeof(struct ip6_hdr) + \ sizeof(struct sctphdr)) #else #define SCTP_MAX_OVERHEAD (sizeof(struct sctp_data_chunk) + \ sizeof(struct sctphdr) + \ sizeof(struct sctp_ecne_chunk) + \ sizeof(struct sctp_sack_chunk) + \ sizeof(struct ip)) #define SCTP_MED_OVERHEAD (sizeof(struct sctp_data_chunk) + \ sizeof(struct sctphdr) + \ sizeof(struct ip)) #define SCTP_MIN_OVERHEAD (sizeof(struct ip) + \ sizeof(struct sctphdr)) #endif /* INET6 */ #endif /* !SCTP_MAX_OVERHEAD */ #define SCTP_MED_V4_OVERHEAD (sizeof(struct sctp_data_chunk) + \ sizeof(struct sctphdr) + \ sizeof(struct ip)) #define SCTP_MIN_V4_OVERHEAD (sizeof(struct ip) + \ sizeof(struct sctphdr)) #undef SCTP_PACKED #endif /* !__sctp_header_h__ */ diff --git a/sys/netinet/sctp_indata.c b/sys/netinet/sctp_indata.c index fce913ef63b9..a6f8d250d5cb 100644 --- a/sys/netinet/sctp_indata.c +++ b/sys/netinet/sctp_indata.c @@ -1,5656 +1,5479 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_indata.c,v 1.36 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include /* * NOTES: On the outbound side of things I need to check the sack timer to * see if I should generate a sack into the chunk queue (if I have data to * send that is and will be sending it .. for bundling. * * The callback in sctp_usrreq.c will get called when the socket is read from. * This will cause sctp_service_queues() to get called on the top entry in * the list. */ void sctp_set_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc) { asoc->my_rwnd = sctp_calc_rwnd(stcb, asoc); } /* Calculate what the rwnd would be */ uint32_t sctp_calc_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc) { uint32_t calc = 0; /* * This is really set wrong with respect to a 1-2-m socket. Since * the sb_cc is the count that everyone as put up. When we re-write * sctp_soreceive then we will fix this so that ONLY this * associations data is taken into account. */ if (stcb->sctp_socket == NULL) return (calc); if (stcb->asoc.sb_cc == 0 && asoc->size_on_reasm_queue == 0 && asoc->size_on_all_streams == 0) { /* Full rwnd granted */ calc = max(SCTP_SB_LIMIT_RCV(stcb->sctp_socket), SCTP_MINIMAL_RWND); return (calc); } /* get actual space */ calc = (uint32_t) sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv); /* * take out what has NOT been put on socket queue and we yet hold * for putting up. */ calc = sctp_sbspace_sub(calc, (uint32_t) (asoc->size_on_reasm_queue + asoc->cnt_on_reasm_queue * MSIZE)); calc = sctp_sbspace_sub(calc, (uint32_t) (asoc->size_on_all_streams + asoc->cnt_on_all_streams * MSIZE)); if (calc == 0) { /* out of space */ return (calc); } /* what is the overhead of all these rwnd's */ calc = sctp_sbspace_sub(calc, stcb->asoc.my_rwnd_control_len); /* * If the window gets too small due to ctrl-stuff, reduce it to 1, * even it is 0. SWS engaged */ if (calc < stcb->asoc.my_rwnd_control_len) { calc = 1; } return (calc); } /* * Build out our readq entry based on the incoming packet. */ struct sctp_queued_to_read * sctp_build_readq_entry(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t tsn, uint32_t ppid, uint32_t context, uint16_t stream_no, uint16_t stream_seq, uint8_t flags, struct mbuf *dm) { struct sctp_queued_to_read *read_queue_e = NULL; sctp_alloc_a_readq(stcb, read_queue_e); if (read_queue_e == NULL) { goto failed_build; } read_queue_e->sinfo_stream = stream_no; read_queue_e->sinfo_ssn = stream_seq; read_queue_e->sinfo_flags = (flags << 8); read_queue_e->sinfo_ppid = ppid; read_queue_e->sinfo_context = stcb->asoc.context; read_queue_e->sinfo_timetolive = 0; read_queue_e->sinfo_tsn = tsn; read_queue_e->sinfo_cumtsn = tsn; read_queue_e->sinfo_assoc_id = sctp_get_associd(stcb); read_queue_e->whoFrom = net; read_queue_e->length = 0; atomic_add_int(&net->ref_count, 1); read_queue_e->data = dm; read_queue_e->spec_flags = 0; read_queue_e->tail_mbuf = NULL; read_queue_e->aux_data = NULL; read_queue_e->stcb = stcb; read_queue_e->port_from = stcb->rport; read_queue_e->do_not_ref_stcb = 0; read_queue_e->end_added = 0; read_queue_e->some_taken = 0; read_queue_e->pdapi_aborted = 0; failed_build: return (read_queue_e); } /* * Build out our readq entry based on the incoming packet. */ static struct sctp_queued_to_read * sctp_build_readq_entry_chk(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk) { struct sctp_queued_to_read *read_queue_e = NULL; sctp_alloc_a_readq(stcb, read_queue_e); if (read_queue_e == NULL) { goto failed_build; } read_queue_e->sinfo_stream = chk->rec.data.stream_number; read_queue_e->sinfo_ssn = chk->rec.data.stream_seq; read_queue_e->sinfo_flags = (chk->rec.data.rcv_flags << 8); read_queue_e->sinfo_ppid = chk->rec.data.payloadtype; read_queue_e->sinfo_context = stcb->asoc.context; read_queue_e->sinfo_timetolive = 0; read_queue_e->sinfo_tsn = chk->rec.data.TSN_seq; read_queue_e->sinfo_cumtsn = chk->rec.data.TSN_seq; read_queue_e->sinfo_assoc_id = sctp_get_associd(stcb); read_queue_e->whoFrom = chk->whoTo; read_queue_e->aux_data = NULL; read_queue_e->length = 0; atomic_add_int(&chk->whoTo->ref_count, 1); read_queue_e->data = chk->data; read_queue_e->tail_mbuf = NULL; read_queue_e->stcb = stcb; read_queue_e->port_from = stcb->rport; read_queue_e->spec_flags = 0; read_queue_e->do_not_ref_stcb = 0; read_queue_e->end_added = 0; read_queue_e->some_taken = 0; read_queue_e->pdapi_aborted = 0; failed_build: return (read_queue_e); } struct mbuf * sctp_build_ctl_nchunk(struct sctp_inpcb *inp, struct sctp_sndrcvinfo *sinfo) { struct sctp_sndrcvinfo *outinfo; struct cmsghdr *cmh; struct mbuf *ret; int len; int use_extended = 0; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) { /* user does not want the sndrcv ctl */ return (NULL); } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) { use_extended = 1; len = CMSG_LEN(sizeof(struct sctp_extrcvinfo)); } else { len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo)); } ret = sctp_get_mbuf_for_msg(len, 0, M_DONTWAIT, 1, MT_DATA); if (ret == NULL) { /* No space */ return (ret); } /* We need a CMSG header followed by the struct */ cmh = mtod(ret, struct cmsghdr *); outinfo = (struct sctp_sndrcvinfo *)CMSG_DATA(cmh); cmh->cmsg_level = IPPROTO_SCTP; if (use_extended) { cmh->cmsg_type = SCTP_EXTRCV; cmh->cmsg_len = len; memcpy(outinfo, sinfo, len); } else { cmh->cmsg_type = SCTP_SNDRCV; cmh->cmsg_len = len; *outinfo = *sinfo; } SCTP_BUF_LEN(ret) = cmh->cmsg_len; return (ret); } char * sctp_build_ctl_cchunk(struct sctp_inpcb *inp, int *control_len, struct sctp_sndrcvinfo *sinfo) { struct sctp_sndrcvinfo *outinfo; struct cmsghdr *cmh; char *buf; int len; int use_extended = 0; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) { /* user does not want the sndrcv ctl */ return (NULL); } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) { use_extended = 1; len = CMSG_LEN(sizeof(struct sctp_extrcvinfo)); } else { len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo)); } SCTP_MALLOC(buf, char *, len, SCTP_M_CMSG); if (buf == NULL) { /* No space */ return (buf); } /* We need a CMSG header followed by the struct */ cmh = (struct cmsghdr *)buf; outinfo = (struct sctp_sndrcvinfo *)CMSG_DATA(cmh); cmh->cmsg_level = IPPROTO_SCTP; if (use_extended) { cmh->cmsg_type = SCTP_EXTRCV; cmh->cmsg_len = len; memcpy(outinfo, sinfo, len); } else { cmh->cmsg_type = SCTP_SNDRCV; cmh->cmsg_len = len; *outinfo = *sinfo; } *control_len = len; return (buf); } static void sctp_mark_non_revokable(struct sctp_association *asoc, uint32_t tsn) { uint32_t gap, i, cumackp1; int fnd = 0; if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { return; } cumackp1 = asoc->cumulative_tsn + 1; if (SCTP_TSN_GT(cumackp1, tsn)) { /* * this tsn is behind the cum ack and thus we don't need to * worry about it being moved from one to the other. */ return; } SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn); if (!SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { printf("gap:%x tsn:%x\n", gap, tsn); sctp_print_mapping_array(asoc); #ifdef INVARIANTS panic("Things are really messed up now!!"); #endif } SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = tsn; } if (tsn == asoc->highest_tsn_inside_map) { /* We must back down to see what the new highest is */ for (i = tsn - 1; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { asoc->highest_tsn_inside_map = i; fnd = 1; break; } } if (!fnd) { asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; } } } /* * We are delivering currently from the reassembly queue. We must continue to * deliver until we either: 1) run out of space. 2) run out of sequential * TSN's 3) hit the SCTP_DATA_LAST_FRAG flag. */ static void sctp_service_reassembly(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk, *nchk; uint16_t nxt_todel; uint16_t stream_no; int end = 0; int cntDel; struct sctp_queued_to_read *control, *ctl, *nctl; if (stcb == NULL) return; cntDel = stream_no = 0; if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { /* socket above is long gone or going.. */ abandon: asoc->fragmented_delivery_inprogress = 0; TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); asoc->size_on_reasm_queue -= chk->send_size; sctp_ucount_decr(asoc->cnt_on_reasm_queue); /* * Lose the data pointer, since its in the socket * buffer */ if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } /* Now free the address and data */ sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ } return; } SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { if (chk->rec.data.TSN_seq != (asoc->tsn_last_delivered + 1)) { /* Can't deliver more :< */ return; } stream_no = chk->rec.data.stream_number; nxt_todel = asoc->strmin[stream_no].last_sequence_delivered + 1; if (nxt_todel != chk->rec.data.stream_seq && (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) { /* * Not the next sequence to deliver in its stream OR * unordered */ return; } if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { control = sctp_build_readq_entry_chk(stcb, chk); if (control == NULL) { /* out of memory? */ return; } /* save it off for our future deliveries */ stcb->asoc.control_pdapi = control; if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) end = 1; else end = 0; sctp_mark_non_revokable(asoc, chk->rec.data.TSN_seq); sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, end, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); cntDel++; } else { if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) end = 1; else end = 0; sctp_mark_non_revokable(asoc, chk->rec.data.TSN_seq); if (sctp_append_to_readq(stcb->sctp_ep, stcb, stcb->asoc.control_pdapi, chk->data, end, chk->rec.data.TSN_seq, &stcb->sctp_socket->so_rcv)) { /* * something is very wrong, either * control_pdapi is NULL, or the tail_mbuf * is corrupt, or there is a EOM already on * the mbuf chain. */ if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto abandon; } else { #ifdef INVARIANTS if ((stcb->asoc.control_pdapi == NULL) || (stcb->asoc.control_pdapi->tail_mbuf == NULL)) { panic("This should not happen control_pdapi NULL?"); } /* if we did not panic, it was a EOM */ panic("Bad chunking ??"); #else if ((stcb->asoc.control_pdapi == NULL) || (stcb->asoc.control_pdapi->tail_mbuf == NULL)) { SCTP_PRINTF("This should not happen control_pdapi NULL?\n"); } SCTP_PRINTF("Bad chunking ??\n"); SCTP_PRINTF("Dumping re-assembly queue this will probably hose the association\n"); #endif goto abandon; } } cntDel++; } /* pull it we did it */ TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { asoc->fragmented_delivery_inprogress = 0; if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) { asoc->strmin[stream_no].last_sequence_delivered++; } if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0) { SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs); } } else if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { /* * turn the flag back on since we just delivered * yet another one. */ asoc->fragmented_delivery_inprogress = 1; } asoc->tsn_of_pdapi_last_delivered = chk->rec.data.TSN_seq; asoc->last_flags_delivered = chk->rec.data.rcv_flags; asoc->last_strm_seq_delivered = chk->rec.data.stream_seq; asoc->last_strm_no_delivered = chk->rec.data.stream_number; asoc->tsn_last_delivered = chk->rec.data.TSN_seq; asoc->size_on_reasm_queue -= chk->send_size; sctp_ucount_decr(asoc->cnt_on_reasm_queue); /* free up the chk */ chk->data = NULL; sctp_free_a_chunk(stcb, chk); if (asoc->fragmented_delivery_inprogress == 0) { /* * Now lets see if we can deliver the next one on * the stream */ struct sctp_stream_in *strm; strm = &asoc->strmin[stream_no]; nxt_todel = strm->last_sequence_delivered + 1; TAILQ_FOREACH_SAFE(ctl, &strm->inqueue, next, nctl) { /* Deliver more if we can. */ if (nxt_todel == ctl->sinfo_ssn) { TAILQ_REMOVE(&strm->inqueue, ctl, next); asoc->size_on_all_streams -= ctl->length; sctp_ucount_decr(asoc->cnt_on_all_streams); strm->last_sequence_delivered++; sctp_mark_non_revokable(asoc, ctl->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, ctl, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } else { break; } nxt_todel = strm->last_sequence_delivered + 1; } break; } } } /* * Queue the chunk either right into the socket buffer if it is the next one * to go OR put it in the correct place in the delivery queue. If we do * append to the so_buf, keep doing so until we are out of order. One big * question still remains, what to do when the socket buffer is FULL?? */ static void sctp_queue_data_to_stream(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_queued_to_read *control, int *abort_flag) { /* * FIX-ME maybe? What happens when the ssn wraps? If we are getting * all the data in one stream this could happen quite rapidly. One * could use the TSN to keep track of things, but this scheme breaks * down in the other type of stream useage that could occur. Send a * single msg to stream 0, send 4Billion messages to stream 1, now * send a message to stream 0. You have a situation where the TSN * has wrapped but not in the stream. Is this worth worrying about * or should we just change our queue sort at the bottom to be by * TSN. * * Could it also be legal for a peer to send ssn 1 with TSN 2 and ssn 2 * with TSN 1? If the peer is doing some sort of funky TSN/SSN * assignment this could happen... and I don't see how this would be * a violation. So for now I am undecided an will leave the sort by * SSN alone. Maybe a hybred approach is the answer * */ struct sctp_stream_in *strm; struct sctp_queued_to_read *at; int queue_needed; uint16_t nxt_todel; struct mbuf *oper; queue_needed = 1; asoc->size_on_all_streams += control->length; sctp_ucount_incr(asoc->cnt_on_all_streams); strm = &asoc->strmin[control->sinfo_stream]; nxt_todel = strm->last_sequence_delivered + 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_INTO_STRD); } SCTPDBG(SCTP_DEBUG_INDATA1, "queue to stream called for ssn:%u lastdel:%u nxt:%u\n", (uint32_t) control->sinfo_stream, (uint32_t) strm->last_sequence_delivered, (uint32_t) nxt_todel); if (SCTP_SSN_GE(strm->last_sequence_delivered, control->sinfo_ssn)) { /* The incoming sseq is behind where we last delivered? */ SCTPDBG(SCTP_DEBUG_INDATA1, "Duplicate S-SEQ:%d delivered:%d from peer, Abort association\n", control->sinfo_ssn, strm->last_sequence_delivered); protocol_error: /* * throw it in the stream so it gets cleaned up in * association destruction */ TAILQ_INSERT_HEAD(&strm->inqueue, control, next); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (sizeof(uint32_t) * 3); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_1); ippp++; *ippp = control->sinfo_tsn; ippp++; *ippp = ((control->sinfo_stream << 16) | control->sinfo_ssn); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_1; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } if (nxt_todel == control->sinfo_ssn) { /* can be delivered right away? */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_IMMED_DEL); } /* EY it wont be queued if it could be delivered directly */ queue_needed = 0; asoc->size_on_all_streams -= control->length; sctp_ucount_decr(asoc->cnt_on_all_streams); strm->last_sequence_delivered++; sctp_mark_non_revokable(asoc, control->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); TAILQ_FOREACH_SAFE(control, &strm->inqueue, next, at) { /* all delivered */ nxt_todel = strm->last_sequence_delivered + 1; if (nxt_todel == control->sinfo_ssn) { TAILQ_REMOVE(&strm->inqueue, control, next); asoc->size_on_all_streams -= control->length; sctp_ucount_decr(asoc->cnt_on_all_streams); strm->last_sequence_delivered++; /* * We ignore the return of deliver_data here * since we always can hold the chunk on the * d-queue. And we have a finite number that * can be delivered from the strq. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_IMMED_DEL); } sctp_mark_non_revokable(asoc, control->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); continue; } break; } } if (queue_needed) { /* * Ok, we did not deliver this guy, find the correct place * to put it on the queue. */ if (SCTP_TSN_GE(asoc->cumulative_tsn, control->sinfo_tsn)) { goto protocol_error; } if (TAILQ_EMPTY(&strm->inqueue)) { /* Empty queue */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_INSERT_HD); } TAILQ_INSERT_HEAD(&strm->inqueue, control, next); } else { TAILQ_FOREACH(at, &strm->inqueue, next) { if (SCTP_SSN_GT(at->sinfo_ssn, control->sinfo_ssn)) { /* * one in queue is bigger than the * new one, insert before this one */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, at, SCTP_STR_LOG_FROM_INSERT_MD); } TAILQ_INSERT_BEFORE(at, control, next); break; } else if (at->sinfo_ssn == control->sinfo_ssn) { /* * Gak, He sent me a duplicate str * seq number */ /* * foo bar, I guess I will just free * this new guy, should we abort * too? FIX ME MAYBE? Or it COULD be * that the SSN's have wrapped. * Maybe I should compare to TSN * somehow... sigh for now just blow * away the chunk! */ if (control->data) sctp_m_freem(control->data); control->data = NULL; asoc->size_on_all_streams -= control->length; sctp_ucount_decr(asoc->cnt_on_all_streams); if (control->whoFrom) { sctp_free_remote_addr(control->whoFrom); control->whoFrom = NULL; } sctp_free_a_readq(stcb, control); return; } else { if (TAILQ_NEXT(at, next) == NULL) { /* * We are at the end, insert * it after this one */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del(control, at, SCTP_STR_LOG_FROM_INSERT_TL); } TAILQ_INSERT_AFTER(&strm->inqueue, at, control, next); break; } } } } } } /* * Returns two things: You get the total size of the deliverable parts of the * first fragmented message on the reassembly queue. And you get a 1 back if * all of the message is ready or a 0 back if the message is still incomplete */ static int sctp_is_all_msg_on_reasm(struct sctp_association *asoc, uint32_t * t_size) { struct sctp_tmit_chunk *chk; uint32_t tsn; *t_size = 0; chk = TAILQ_FIRST(&asoc->reasmqueue); if (chk == NULL) { /* nothing on the queue */ return (0); } if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0) { /* Not a first on the queue */ return (0); } tsn = chk->rec.data.TSN_seq; TAILQ_FOREACH(chk, &asoc->reasmqueue, sctp_next) { if (tsn != chk->rec.data.TSN_seq) { return (0); } *t_size += chk->send_size; if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { return (1); } tsn++; } return (0); } static void sctp_deliver_reasm_check(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; uint16_t nxt_todel; uint32_t tsize, pd_point; doit_again: chk = TAILQ_FIRST(&asoc->reasmqueue); if (chk == NULL) { /* Huh? */ asoc->size_on_reasm_queue = 0; asoc->cnt_on_reasm_queue = 0; return; } if (asoc->fragmented_delivery_inprogress == 0) { nxt_todel = asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered + 1; if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) && (nxt_todel == chk->rec.data.stream_seq || (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED))) { /* * Yep the first one is here and its ok to deliver * but should we? */ if (stcb->sctp_socket) { pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket), stcb->sctp_ep->partial_delivery_point); } else { pd_point = stcb->sctp_ep->partial_delivery_point; } if (sctp_is_all_msg_on_reasm(asoc, &tsize) || (tsize >= pd_point)) { /* * Yes, we setup to start reception, by * backing down the TSN just in case we * can't deliver. If we */ asoc->fragmented_delivery_inprogress = 1; asoc->tsn_last_delivered = chk->rec.data.TSN_seq - 1; asoc->str_of_pdapi = chk->rec.data.stream_number; asoc->ssn_of_pdapi = chk->rec.data.stream_seq; asoc->pdapi_ppid = chk->rec.data.payloadtype; asoc->fragment_flags = chk->rec.data.rcv_flags; sctp_service_reassembly(stcb, asoc); } } } else { /* * Service re-assembly will deliver stream data queued at * the end of fragmented delivery.. but it wont know to go * back and call itself again... we do that here with the * got doit_again */ sctp_service_reassembly(stcb, asoc); if (asoc->fragmented_delivery_inprogress == 0) { /* * finished our Fragmented delivery, could be more * waiting? */ goto doit_again; } } } /* * Dump onto the re-assembly queue, in its proper place. After dumping on the * queue, see if anthing can be delivered. If so pull it off (or as much as * we can. If we run out of space then we must dump what we can and set the * appropriate flag to say we queued what we could. */ static void sctp_queue_data_for_reasm(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_tmit_chunk *chk, int *abort_flag) { struct mbuf *oper; uint32_t cum_ackp1, last_tsn, prev_tsn, post_tsn; u_char last_flags; struct sctp_tmit_chunk *at, *prev, *next; prev = next = NULL; cum_ackp1 = asoc->tsn_last_delivered + 1; if (TAILQ_EMPTY(&asoc->reasmqueue)) { /* This is the first one on the queue */ TAILQ_INSERT_HEAD(&asoc->reasmqueue, chk, sctp_next); /* * we do not check for delivery of anything when only one * fragment is here */ asoc->size_on_reasm_queue = chk->send_size; sctp_ucount_incr(asoc->cnt_on_reasm_queue); if (chk->rec.data.TSN_seq == cum_ackp1) { if (asoc->fragmented_delivery_inprogress == 0 && (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) != SCTP_DATA_FIRST_FRAG) { /* * An empty queue, no delivery inprogress, * we hit the next one and it does NOT have * a FIRST fragment mark. */ SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, its not first, no fragmented delivery in progress\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (sizeof(uint32_t) * 3); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_2); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_2; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; } else if (asoc->fragmented_delivery_inprogress && (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == SCTP_DATA_FIRST_FRAG) { /* * We are doing a partial delivery and the * NEXT chunk MUST be either the LAST or * MIDDLE fragment NOT a FIRST */ SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS a first and fragmented delivery in progress\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_3); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_3; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; } else if (asoc->fragmented_delivery_inprogress) { /* * Here we are ok with a MIDDLE or LAST * piece */ if (chk->rec.data.stream_number != asoc->str_of_pdapi) { /* Got to be the right STR No */ SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS not same stream number %d vs %d\n", chk->rec.data.stream_number, asoc->str_of_pdapi); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (sizeof(uint32_t) * 3); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_4); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_4; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; } else if ((asoc->fragment_flags & SCTP_DATA_UNORDERED) != SCTP_DATA_UNORDERED && chk->rec.data.stream_seq != asoc->ssn_of_pdapi) { /* Got to be the right STR Seq */ SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS not same stream seq %d vs %d\n", chk->rec.data.stream_seq, asoc->ssn_of_pdapi); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_5); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_5; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; } } } return; } /* Find its place */ TAILQ_FOREACH(at, &asoc->reasmqueue, sctp_next) { if (SCTP_TSN_GT(at->rec.data.TSN_seq, chk->rec.data.TSN_seq)) { /* * one in queue is bigger than the new one, insert * before this one */ /* A check */ asoc->size_on_reasm_queue += chk->send_size; sctp_ucount_incr(asoc->cnt_on_reasm_queue); next = at; TAILQ_INSERT_BEFORE(at, chk, sctp_next); break; } else if (at->rec.data.TSN_seq == chk->rec.data.TSN_seq) { /* Gak, He sent me a duplicate str seq number */ /* * foo bar, I guess I will just free this new guy, * should we abort too? FIX ME MAYBE? Or it COULD be * that the SSN's have wrapped. Maybe I should * compare to TSN somehow... sigh for now just blow * away the chunk! */ if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); return; } else { last_flags = at->rec.data.rcv_flags; last_tsn = at->rec.data.TSN_seq; prev = at; if (TAILQ_NEXT(at, sctp_next) == NULL) { /* * We are at the end, insert it after this * one */ /* check it first */ asoc->size_on_reasm_queue += chk->send_size; sctp_ucount_incr(asoc->cnt_on_reasm_queue); TAILQ_INSERT_AFTER(&asoc->reasmqueue, at, chk, sctp_next); break; } } } /* Now the audits */ if (prev) { prev_tsn = chk->rec.data.TSN_seq - 1; if (prev_tsn == prev->rec.data.TSN_seq) { /* * Ok the one I am dropping onto the end is the * NEXT. A bit of valdiation here. */ if ((prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_FIRST_FRAG || (prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_MIDDLE_FRAG) { /* * Insert chk MUST be a MIDDLE or LAST * fragment */ if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_FIRST_FRAG) { SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - It can be a midlle or last but not a first\n"); SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it's a FIRST!\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_6); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_6; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } if (chk->rec.data.stream_number != prev->rec.data.stream_number) { /* * Huh, need the correct STR here, * they must be the same. */ SCTP_PRINTF("Prev check - Gak, Evil plot, ssn:%d not the same as at:%d\n", chk->rec.data.stream_number, prev->rec.data.stream_number); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_7); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_7; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } if ((prev->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0 && chk->rec.data.stream_seq != prev->rec.data.stream_seq) { /* * Huh, need the correct STR here, * they must be the same. */ SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - Gak, Evil plot, sseq:%d not the same as at:%d\n", chk->rec.data.stream_seq, prev->rec.data.stream_seq); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_8); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_8; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } } else if ((prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_LAST_FRAG) { /* Insert chk MUST be a FIRST */ if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) != SCTP_DATA_FIRST_FRAG) { SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - Gak, evil plot, its not FIRST and it must be!\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_9); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_9; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } } } } if (next) { post_tsn = chk->rec.data.TSN_seq + 1; if (post_tsn == next->rec.data.TSN_seq) { /* * Ok the one I am inserting ahead of is my NEXT * one. A bit of valdiation here. */ if (next->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { /* Insert chk MUST be a last fragment */ if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) != SCTP_DATA_LAST_FRAG) { SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Next is FIRST, we must be LAST\n"); SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, its not a last!\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_10); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_10; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } } else if ((next->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_MIDDLE_FRAG || (next->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_LAST_FRAG) { /* * Insert chk CAN be MIDDLE or FIRST NOT * LAST */ if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) == SCTP_DATA_LAST_FRAG) { SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Next is a MIDDLE/LAST\n"); SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, new prev chunk is a LAST\n"); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_11); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_11; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } if (chk->rec.data.stream_number != next->rec.data.stream_number) { /* * Huh, need the correct STR here, * they must be the same. */ SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Gak, Evil plot, ssn:%d not the same as at:%d\n", chk->rec.data.stream_number, next->rec.data.stream_number); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_12); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_12; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } if ((next->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0 && chk->rec.data.stream_seq != next->rec.data.stream_seq) { /* * Huh, need the correct STR here, * they must be the same. */ SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Gak, Evil plot, sseq:%d not the same as at:%d\n", chk->rec.data.stream_seq, next->rec.data.stream_seq); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_13); ippp++; *ippp = chk->rec.data.TSN_seq; ippp++; *ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_13; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return; } } } } /* Do we need to do some delivery? check */ sctp_deliver_reasm_check(stcb, asoc); } /* * This is an unfortunate routine. It checks to make sure a evil guy is not * stuffing us full of bad packet fragments. A broken peer could also do this * but this is doubtful. It is to bad I must worry about evil crackers sigh * :< more cycles. */ static int sctp_does_tsn_belong_to_reasm(struct sctp_association *asoc, uint32_t TSN_seq) { struct sctp_tmit_chunk *at; uint32_t tsn_est; TAILQ_FOREACH(at, &asoc->reasmqueue, sctp_next) { if (SCTP_TSN_GT(TSN_seq, at->rec.data.TSN_seq)) { /* is it one bigger? */ tsn_est = at->rec.data.TSN_seq + 1; if (tsn_est == TSN_seq) { /* yep. It better be a last then */ if ((at->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) != SCTP_DATA_LAST_FRAG) { /* * Ok this guy belongs next to a guy * that is NOT last, it should be a * middle/last, not a complete * chunk. */ return (1); } else { /* * This guy is ok since its a LAST * and the new chunk is a fully * self- contained one. */ return (0); } } } else if (TSN_seq == at->rec.data.TSN_seq) { /* Software error since I have a dup? */ return (1); } else { /* * Ok, 'at' is larger than new chunk but does it * need to be right before it. */ tsn_est = TSN_seq + 1; if (tsn_est == at->rec.data.TSN_seq) { /* Yep, It better be a first */ if ((at->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) != SCTP_DATA_FIRST_FRAG) { return (1); } else { return (0); } } } } return (0); } static int sctp_process_a_data_chunk(struct sctp_tcb *stcb, struct sctp_association *asoc, struct mbuf **m, int offset, struct sctp_data_chunk *ch, int chk_length, struct sctp_nets *net, uint32_t * high_tsn, int *abort_flag, int *break_flag, int last_chunk) { /* Process a data chunk */ /* struct sctp_tmit_chunk *chk; */ struct sctp_tmit_chunk *chk; uint32_t tsn, gap; struct mbuf *dmbuf; int indx, the_len; int need_reasm_check = 0; uint16_t strmno, strmseq; struct mbuf *oper; struct sctp_queued_to_read *control; int ordered; uint32_t protocol_id; uint8_t chunk_flags; struct sctp_stream_reset_list *liste; chk = NULL; tsn = ntohl(ch->dp.tsn); chunk_flags = ch->ch.chunk_flags; if ((chunk_flags & SCTP_DATA_SACK_IMMEDIATELY) == SCTP_DATA_SACK_IMMEDIATELY) { asoc->send_sack = 1; } protocol_id = ch->dp.protocol_id; ordered = ((chunk_flags & SCTP_DATA_UNORDERED) == 0); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, SCTP_MAP_TSN_ENTERS); } if (stcb == NULL) { return (0); } SCTP_LTRACE_CHK(stcb->sctp_ep, stcb, ch->ch.chunk_type, tsn); if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) { /* It is a duplicate */ SCTP_STAT_INCR(sctps_recvdupdata); if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) { /* Record a dup for the next outbound sack */ asoc->dup_tsns[asoc->numduptsns] = tsn; asoc->numduptsns++; } asoc->send_sack = 1; return (0); } /* Calculate the number of TSN's between the base and this TSN */ SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn); if (gap >= (SCTP_MAPPING_ARRAY << 3)) { /* Can't hold the bit in the mapping at max array, toss it */ return (0); } if (gap >= (uint32_t) (asoc->mapping_array_size << 3)) { SCTP_TCB_LOCK_ASSERT(stcb); if (sctp_expand_mapping_array(asoc, gap)) { /* Can't expand, drop it */ return (0); } } if (SCTP_TSN_GT(tsn, *high_tsn)) { *high_tsn = tsn; } /* See if we have received this one already */ if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap) || SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, gap)) { SCTP_STAT_INCR(sctps_recvdupdata); if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) { /* Record a dup for the next outbound sack */ asoc->dup_tsns[asoc->numduptsns] = tsn; asoc->numduptsns++; } asoc->send_sack = 1; return (0); } /* * Check to see about the GONE flag, duplicates would cause a sack * to be sent up above */ if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) ) { /* * wait a minute, this guy is gone, there is no longer a * receiver. Send peer an ABORT! */ struct mbuf *op_err; op_err = sctp_generate_invmanparam(SCTP_CAUSE_OUT_OF_RESC); sctp_abort_an_association(stcb->sctp_ep, stcb, 0, op_err, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return (0); } /* * Now before going further we see if there is room. If NOT then we * MAY let one through only IF this TSN is the one we are waiting * for on a partial delivery API. */ /* now do the tests */ if (((asoc->cnt_on_all_streams + asoc->cnt_on_reasm_queue + asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) || (((int)asoc->my_rwnd) <= 0)) { /* * When we have NO room in the rwnd we check to make sure * the reader is doing its job... */ if (stcb->sctp_socket->so_rcv.sb_cc) { /* some to read, wake-up */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return (0); } #endif sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } /* now is it in the mapping array of what we have accepted? */ if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map) && SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { /* Nope not in the valid range dump it */ sctp_set_rwnd(stcb, asoc); if ((asoc->cnt_on_all_streams + asoc->cnt_on_reasm_queue + asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) { SCTP_STAT_INCR(sctps_datadropchklmt); } else { SCTP_STAT_INCR(sctps_datadroprwnd); } indx = *break_flag; *break_flag = 1; return (0); } } strmno = ntohs(ch->dp.stream_id); if (strmno >= asoc->streamincnt) { struct sctp_paramhdr *phdr; struct mbuf *mb; mb = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) * 2), 0, M_DONTWAIT, 1, MT_DATA); if (mb != NULL) { /* add some space up front so prepend will work well */ SCTP_BUF_RESV_UF(mb, sizeof(struct sctp_chunkhdr)); phdr = mtod(mb, struct sctp_paramhdr *); /* * Error causes are just param's and this one has * two back to back phdr, one with the error type * and size, the other with the streamid and a rsvd */ SCTP_BUF_LEN(mb) = (sizeof(struct sctp_paramhdr) * 2); phdr->param_type = htons(SCTP_CAUSE_INVALID_STREAM); phdr->param_length = htons(sizeof(struct sctp_paramhdr) * 2); phdr++; /* We insert the stream in the type field */ phdr->param_type = ch->dp.stream_id; /* And set the length to 0 for the rsvd field */ phdr->param_length = 0; sctp_queue_op_err(stcb, mb); } SCTP_STAT_INCR(sctps_badsid); SCTP_TCB_LOCK_ASSERT(stcb); SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = tsn; } if (tsn == (asoc->cumulative_tsn + 1)) { /* Update cum-ack */ asoc->cumulative_tsn = tsn; } return (0); } /* * Before we continue lets validate that we are not being fooled by * an evil attacker. We can only have 4k chunks based on our TSN * spread allowed by the mapping array 512 * 8 bits, so there is no * way our stream sequence numbers could have wrapped. We of course * only validate the FIRST fragment so the bit must be set. */ strmseq = ntohs(ch->dp.stream_sequence); #ifdef SCTP_ASOCLOG_OF_TSNS SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->tsn_in_at >= SCTP_TSN_LOG_SIZE) { asoc->tsn_in_at = 0; asoc->tsn_in_wrapped = 1; } asoc->in_tsnlog[asoc->tsn_in_at].tsn = tsn; asoc->in_tsnlog[asoc->tsn_in_at].strm = strmno; asoc->in_tsnlog[asoc->tsn_in_at].seq = strmseq; asoc->in_tsnlog[asoc->tsn_in_at].sz = chk_length; asoc->in_tsnlog[asoc->tsn_in_at].flgs = chunk_flags; asoc->in_tsnlog[asoc->tsn_in_at].stcb = (void *)stcb; asoc->in_tsnlog[asoc->tsn_in_at].in_pos = asoc->tsn_in_at; asoc->in_tsnlog[asoc->tsn_in_at].in_out = 1; asoc->tsn_in_at++; #endif if ((chunk_flags & SCTP_DATA_FIRST_FRAG) && (TAILQ_EMPTY(&asoc->resetHead)) && (chunk_flags & SCTP_DATA_UNORDERED) == 0 && SCTP_SSN_GE(asoc->strmin[strmno].last_sequence_delivered, strmseq)) { /* The incoming sseq is behind where we last delivered? */ SCTPDBG(SCTP_DEBUG_INDATA1, "EVIL/Broken-Dup S-SEQ:%d delivered:%d from peer, Abort!\n", strmseq, asoc->strmin[strmno].last_sequence_delivered); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_14); ippp++; *ippp = tsn; ippp++; *ippp = ((strmno << 16) | strmseq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_14; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return (0); } /************************************ * From here down we may find ch-> invalid * so its a good idea NOT to use it. *************************************/ the_len = (chk_length - sizeof(struct sctp_data_chunk)); if (last_chunk == 0) { dmbuf = SCTP_M_COPYM(*m, (offset + sizeof(struct sctp_data_chunk)), the_len, M_DONTWAIT); #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = dmbuf; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif } else { /* We can steal the last chunk */ int l_len; dmbuf = *m; /* lop off the top part */ m_adj(dmbuf, (offset + sizeof(struct sctp_data_chunk))); if (SCTP_BUF_NEXT(dmbuf) == NULL) { l_len = SCTP_BUF_LEN(dmbuf); } else { /* * need to count up the size hopefully does not hit * this to often :-0 */ struct mbuf *lat; l_len = 0; lat = dmbuf; while (lat) { l_len += SCTP_BUF_LEN(lat); lat = SCTP_BUF_NEXT(lat); } } if (l_len > the_len) { /* Trim the end round bytes off too */ m_adj(dmbuf, -(l_len - the_len)); } } if (dmbuf == NULL) { SCTP_STAT_INCR(sctps_nomem); return (0); } if ((chunk_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG && asoc->fragmented_delivery_inprogress == 0 && TAILQ_EMPTY(&asoc->resetHead) && ((ordered == 0) || ((uint16_t) (asoc->strmin[strmno].last_sequence_delivered + 1) == strmseq && TAILQ_EMPTY(&asoc->strmin[strmno].inqueue)))) { /* Candidate for express delivery */ /* * Its not fragmented, No PD-API is up, Nothing in the * delivery queue, Its un-ordered OR ordered and the next to * deliver AND nothing else is stuck on the stream queue, * And there is room for it in the socket buffer. Lets just * stuff it up the buffer.... */ /* It would be nice to avoid this copy if we could :< */ sctp_alloc_a_readq(stcb, control); sctp_build_readq_entry_mac(control, stcb, asoc->context, net, tsn, protocol_id, stcb->asoc.context, strmno, strmseq, chunk_flags, dmbuf); if (control == NULL) { goto failed_express_del; } SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = tsn; } sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); if ((chunk_flags & SCTP_DATA_UNORDERED) == 0) { /* for ordered, bump what we delivered */ asoc->strmin[strmno].last_sequence_delivered++; } SCTP_STAT_INCR(sctps_recvexpress); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del_alt(stcb, tsn, strmseq, strmno, SCTP_STR_LOG_FROM_EXPRS_DEL); } control = NULL; goto finish_express_del; } failed_express_del: /* If we reach here this is a new chunk */ chk = NULL; control = NULL; /* Express for fragmented delivery? */ if ((asoc->fragmented_delivery_inprogress) && (stcb->asoc.control_pdapi) && (asoc->str_of_pdapi == strmno) && (asoc->ssn_of_pdapi == strmseq) ) { control = stcb->asoc.control_pdapi; if ((chunk_flags & SCTP_DATA_FIRST_FRAG) == SCTP_DATA_FIRST_FRAG) { /* Can't be another first? */ goto failed_pdapi_express_del; } if (tsn == (control->sinfo_tsn + 1)) { /* Yep, we can add it on */ int end = 0; uint32_t cumack; if (chunk_flags & SCTP_DATA_LAST_FRAG) { end = 1; } cumack = asoc->cumulative_tsn; if ((cumack + 1) == tsn) cumack = tsn; if (sctp_append_to_readq(stcb->sctp_ep, stcb, control, dmbuf, end, tsn, &stcb->sctp_socket->so_rcv)) { SCTP_PRINTF("Append fails end:%d\n", end); goto failed_pdapi_express_del; } SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = tsn; } SCTP_STAT_INCR(sctps_recvexpressm); control->sinfo_tsn = tsn; asoc->tsn_last_delivered = tsn; asoc->fragment_flags = chunk_flags; asoc->tsn_of_pdapi_last_delivered = tsn; asoc->last_flags_delivered = chunk_flags; asoc->last_strm_seq_delivered = strmseq; asoc->last_strm_no_delivered = strmno; if (end) { /* clean up the flags and such */ asoc->fragmented_delivery_inprogress = 0; if ((chunk_flags & SCTP_DATA_UNORDERED) == 0) { asoc->strmin[strmno].last_sequence_delivered++; } stcb->asoc.control_pdapi = NULL; if (TAILQ_EMPTY(&asoc->reasmqueue) == 0) { /* * There could be another message * ready */ need_reasm_check = 1; } } control = NULL; goto finish_express_del; } } failed_pdapi_express_del: control = NULL; if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = tsn; } } else { SCTP_SET_TSN_PRESENT(asoc->mapping_array, gap); if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map)) { asoc->highest_tsn_inside_map = tsn; } } if ((chunk_flags & SCTP_DATA_NOT_FRAG) != SCTP_DATA_NOT_FRAG) { sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* No memory so we drop the chunk */ SCTP_STAT_INCR(sctps_nomem); if (last_chunk == 0) { /* we copied it, free the copy */ sctp_m_freem(dmbuf); } return (0); } chk->rec.data.TSN_seq = tsn; chk->no_fr_allowed = 0; chk->rec.data.stream_seq = strmseq; chk->rec.data.stream_number = strmno; chk->rec.data.payloadtype = protocol_id; chk->rec.data.context = stcb->asoc.context; chk->rec.data.doing_fast_retransmit = 0; chk->rec.data.rcv_flags = chunk_flags; chk->asoc = asoc; chk->send_size = the_len; chk->whoTo = net; atomic_add_int(&net->ref_count, 1); chk->data = dmbuf; } else { sctp_alloc_a_readq(stcb, control); sctp_build_readq_entry_mac(control, stcb, asoc->context, net, tsn, protocol_id, stcb->asoc.context, strmno, strmseq, chunk_flags, dmbuf); if (control == NULL) { /* No memory so we drop the chunk */ SCTP_STAT_INCR(sctps_nomem); if (last_chunk == 0) { /* we copied it, free the copy */ sctp_m_freem(dmbuf); } return (0); } control->length = the_len; } /* Mark it as received */ /* Now queue it where it belongs */ if (control != NULL) { /* First a sanity check */ if (asoc->fragmented_delivery_inprogress) { /* * Ok, we have a fragmented delivery in progress if * this chunk is next to deliver OR belongs in our * view to the reassembly, the peer is evil or * broken. */ uint32_t estimate_tsn; estimate_tsn = asoc->tsn_last_delivered + 1; if (TAILQ_EMPTY(&asoc->reasmqueue) && (estimate_tsn == control->sinfo_tsn)) { /* Evil/Broke peer */ sctp_m_freem(control->data); control->data = NULL; if (control->whoFrom) { sctp_free_remote_addr(control->whoFrom); control->whoFrom = NULL; } sctp_free_a_readq(stcb, control); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_15); ippp++; *ippp = tsn; ippp++; *ippp = ((strmno << 16) | strmseq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_15; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return (0); } else { if (sctp_does_tsn_belong_to_reasm(asoc, control->sinfo_tsn)) { sctp_m_freem(control->data); control->data = NULL; if (control->whoFrom) { sctp_free_remote_addr(control->whoFrom); control->whoFrom = NULL; } sctp_free_a_readq(stcb, control); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_16); ippp++; *ippp = tsn; ippp++; *ippp = ((strmno << 16) | strmseq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_16; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return (0); } } } else { /* No PDAPI running */ if (!TAILQ_EMPTY(&asoc->reasmqueue)) { /* * Reassembly queue is NOT empty validate * that this tsn does not need to be in * reasembly queue. If it does then our peer * is broken or evil. */ if (sctp_does_tsn_belong_to_reasm(asoc, control->sinfo_tsn)) { sctp_m_freem(control->data); control->data = NULL; if (control->whoFrom) { sctp_free_remote_addr(control->whoFrom); control->whoFrom = NULL; } sctp_free_a_readq(stcb, control); oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (3 * sizeof(uint32_t)); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_17); ippp++; *ippp = tsn; ippp++; *ippp = ((strmno << 16) | strmseq); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_17; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); *abort_flag = 1; return (0); } } } /* ok, if we reach here we have passed the sanity checks */ if (chunk_flags & SCTP_DATA_UNORDERED) { /* queue directly into socket buffer */ sctp_mark_non_revokable(asoc, control->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } else { /* * Special check for when streams are resetting. We * could be more smart about this and check the * actual stream to see if it is not being reset.. * that way we would not create a HOLB when amongst * streams being reset and those not being reset. * * We take complete messages that have a stream reset * intervening (aka the TSN is after where our * cum-ack needs to be) off and put them on a * pending_reply_queue. The reassembly ones we do * not have to worry about since they are all sorted * and proceessed by TSN order. It is only the * singletons I must worry about. */ if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) && SCTP_TSN_GT(tsn, liste->tsn)) { /* * yep its past where we need to reset... go * ahead and queue it. */ if (TAILQ_EMPTY(&asoc->pending_reply_queue)) { /* first one on */ TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next); } else { struct sctp_queued_to_read *ctlOn, *nctlOn; unsigned char inserted = 0; TAILQ_FOREACH_SAFE(ctlOn, &asoc->pending_reply_queue, next, nctlOn) { if (SCTP_TSN_GT(control->sinfo_tsn, ctlOn->sinfo_tsn)) { continue; } else { /* found it */ TAILQ_INSERT_BEFORE(ctlOn, control, next); inserted = 1; break; } } if (inserted == 0) { /* * must be put at end, use * prevP (all setup from * loop) to setup nextP. */ TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next); } } } else { sctp_queue_data_to_stream(stcb, asoc, control, abort_flag); if (*abort_flag) { return (0); } } } } else { /* Into the re-assembly queue */ sctp_queue_data_for_reasm(stcb, asoc, chk, abort_flag); if (*abort_flag) { /* * the assoc is now gone and chk was put onto the * reasm queue, which has all been freed. */ *m = NULL; return (0); } } finish_express_del: if (tsn == (asoc->cumulative_tsn + 1)) { /* Update cum-ack */ asoc->cumulative_tsn = tsn; } if (last_chunk) { *m = NULL; } if (ordered) { SCTP_STAT_INCR_COUNTER64(sctps_inorderchunks); } else { SCTP_STAT_INCR_COUNTER64(sctps_inunorderchunks); } SCTP_STAT_INCR(sctps_recvdata); /* Set it present please */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { sctp_log_strm_del_alt(stcb, tsn, strmseq, strmno, SCTP_STR_LOG_FROM_MARK_TSN); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(asoc->mapping_array_base_tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, SCTP_MAP_PREPARE_SLIDE); } /* check the special flag for stream resets */ if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) && SCTP_TSN_GE(asoc->cumulative_tsn, liste->tsn)) { /* * we have finished working through the backlogged TSN's now * time to reset streams. 1: call reset function. 2: free * pending_reply space 3: distribute any chunks in * pending_reply_queue. */ struct sctp_queued_to_read *ctl, *nctl; sctp_reset_in_stream(stcb, liste->number_entries, liste->req.list_of_streams); TAILQ_REMOVE(&asoc->resetHead, liste, next_resp); SCTP_FREE(liste, SCTP_M_STRESET); /* sa_ignore FREED_MEMORY */ liste = TAILQ_FIRST(&asoc->resetHead); if (TAILQ_EMPTY(&asoc->resetHead)) { /* All can be removed */ TAILQ_FOREACH_SAFE(ctl, &asoc->pending_reply_queue, next, nctl) { TAILQ_REMOVE(&asoc->pending_reply_queue, ctl, next); sctp_queue_data_to_stream(stcb, asoc, ctl, abort_flag); if (*abort_flag) { return (0); } } } else { TAILQ_FOREACH_SAFE(ctl, &asoc->pending_reply_queue, next, nctl) { if (SCTP_TSN_GT(ctl->sinfo_tsn, liste->tsn)) { break; } /* * if ctl->sinfo_tsn is <= liste->tsn we can * process it which is the NOT of * ctl->sinfo_tsn > liste->tsn */ TAILQ_REMOVE(&asoc->pending_reply_queue, ctl, next); sctp_queue_data_to_stream(stcb, asoc, ctl, abort_flag); if (*abort_flag) { return (0); } } } /* * Now service re-assembly to pick up anything that has been * held on reassembly queue? */ sctp_deliver_reasm_check(stcb, asoc); need_reasm_check = 0; } if (need_reasm_check) { /* Another one waits ? */ sctp_deliver_reasm_check(stcb, asoc); } return (1); } int8_t sctp_map_lookup_tab[256] = { 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 6, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 7, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 6, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 8 }; void sctp_slide_mapping_arrays(struct sctp_tcb *stcb) { /* * Now we also need to check the mapping array in a couple of ways. * 1) Did we move the cum-ack point? * * When you first glance at this you might think that all entries that * make up the postion of the cum-ack would be in the nr-mapping * array only.. i.e. things up to the cum-ack are always * deliverable. Thats true with one exception, when its a fragmented * message we may not deliver the data until some threshold (or all * of it) is in place. So we must OR the nr_mapping_array and * mapping_array to get a true picture of the cum-ack. */ struct sctp_association *asoc; int at; uint8_t val; int slide_from, slide_end, lgap, distance; uint32_t old_cumack, old_base, old_highest, highest_tsn; asoc = &stcb->asoc; at = 0; old_cumack = asoc->cumulative_tsn; old_base = asoc->mapping_array_base_tsn; old_highest = asoc->highest_tsn_inside_map; /* * We could probably improve this a small bit by calculating the * offset of the current cum-ack as the starting point. */ at = 0; for (slide_from = 0; slide_from < stcb->asoc.mapping_array_size; slide_from++) { val = asoc->nr_mapping_array[slide_from] | asoc->mapping_array[slide_from]; if (val == 0xff) { at += 8; } else { /* there is a 0 bit */ at += sctp_map_lookup_tab[val]; break; } } asoc->cumulative_tsn = asoc->mapping_array_base_tsn + (at - 1); if (SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_map) && SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_nr_map)) { #ifdef INVARIANTS panic("huh, cumack 0x%x greater than high-tsn 0x%x in map", asoc->cumulative_tsn, asoc->highest_tsn_inside_map); #else SCTP_PRINTF("huh, cumack 0x%x greater than high-tsn 0x%x in map - should panic?\n", asoc->cumulative_tsn, asoc->highest_tsn_inside_map); sctp_print_mapping_array(asoc); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 6, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } asoc->highest_tsn_inside_map = asoc->cumulative_tsn; asoc->highest_tsn_inside_nr_map = asoc->cumulative_tsn; #endif } if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { highest_tsn = asoc->highest_tsn_inside_nr_map; } else { highest_tsn = asoc->highest_tsn_inside_map; } if ((asoc->cumulative_tsn == highest_tsn) && (at >= 8)) { /* The complete array was completed by a single FR */ /* highest becomes the cum-ack */ int clr; #ifdef INVARIANTS unsigned int i; #endif /* clear the array */ clr = ((at + 7) >> 3); if (clr > asoc->mapping_array_size) { clr = asoc->mapping_array_size; } memset(asoc->mapping_array, 0, clr); memset(asoc->nr_mapping_array, 0, clr); #ifdef INVARIANTS for (i = 0; i < asoc->mapping_array_size; i++) { if ((asoc->mapping_array[i]) || (asoc->nr_mapping_array[i])) { printf("Error Mapping array's not clean at clear\n"); sctp_print_mapping_array(asoc); } } #endif asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1; asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map = asoc->cumulative_tsn; } else if (at >= 8) { /* we can slide the mapping array down */ /* slide_from holds where we hit the first NON 0xff byte */ /* * now calculate the ceiling of the move using our highest * TSN value */ SCTP_CALC_TSN_TO_GAP(lgap, highest_tsn, asoc->mapping_array_base_tsn); slide_end = (lgap >> 3); if (slide_end < slide_from) { sctp_print_mapping_array(asoc); #ifdef INVARIANTS panic("impossible slide"); #else printf("impossible slide lgap:%x slide_end:%x slide_from:%x? at:%d\n", lgap, slide_end, slide_from, at); return; #endif } if (slide_end > asoc->mapping_array_size) { #ifdef INVARIANTS panic("would overrun buffer"); #else printf("Gak, would have overrun map end:%d slide_end:%d\n", asoc->mapping_array_size, slide_end); slide_end = asoc->mapping_array_size; #endif } distance = (slide_end - slide_from) + 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(old_base, old_cumack, old_highest, SCTP_MAP_PREPARE_SLIDE); sctp_log_map((uint32_t) slide_from, (uint32_t) slide_end, (uint32_t) lgap, SCTP_MAP_SLIDE_FROM); } if (distance + slide_from > asoc->mapping_array_size || distance < 0) { /* * Here we do NOT slide forward the array so that * hopefully when more data comes in to fill it up * we will be able to slide it forward. Really I * don't think this should happen :-0 */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map((uint32_t) distance, (uint32_t) slide_from, (uint32_t) asoc->mapping_array_size, SCTP_MAP_SLIDE_NONE); } } else { int ii; for (ii = 0; ii < distance; ii++) { asoc->mapping_array[ii] = asoc->mapping_array[slide_from + ii]; asoc->nr_mapping_array[ii] = asoc->nr_mapping_array[slide_from + ii]; } for (ii = distance; ii < asoc->mapping_array_size; ii++) { asoc->mapping_array[ii] = 0; asoc->nr_mapping_array[ii] = 0; } if (asoc->highest_tsn_inside_map + 1 == asoc->mapping_array_base_tsn) { asoc->highest_tsn_inside_map += (slide_from << 3); } if (asoc->highest_tsn_inside_nr_map + 1 == asoc->mapping_array_base_tsn) { asoc->highest_tsn_inside_nr_map += (slide_from << 3); } asoc->mapping_array_base_tsn += (slide_from << 3); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(asoc->mapping_array_base_tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } } } } - void sctp_sack_check(struct sctp_tcb *stcb, int was_a_gap, int *abort_flag) { struct sctp_association *asoc; uint32_t highest_tsn; asoc = &stcb->asoc; if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { highest_tsn = asoc->highest_tsn_inside_nr_map; } else { highest_tsn = asoc->highest_tsn_inside_map; } /* * Now we need to see if we need to queue a sack or just start the * timer (if allowed). */ if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) { /* * Ok special case, in SHUTDOWN-SENT case. here we maker * sure SACK timer is off and instead send a SHUTDOWN and a * SACK */ if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_INDATA + SCTP_LOC_18); } sctp_send_shutdown(stcb, stcb->asoc.primary_destination); sctp_send_sack(stcb); } else { int is_a_gap; /* is there a gap now ? */ is_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); /* * CMT DAC algorithm: increase number of packets received * since last ack */ stcb->asoc.cmt_dac_pkts_rcvd++; if ((stcb->asoc.send_sack == 1) || /* We need to send a * SACK */ ((was_a_gap) && (is_a_gap == 0)) || /* was a gap, but no * longer is one */ (stcb->asoc.numduptsns) || /* we have dup's */ (is_a_gap) || /* is still a gap */ (stcb->asoc.delayed_ack == 0) || /* Delayed sack disabled */ (stcb->asoc.data_pkts_seen >= stcb->asoc.sack_freq) /* hit limit of pkts */ ) { if ((stcb->asoc.sctp_cmt_on_off > 0) && (SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) && (stcb->asoc.send_sack == 0) && (stcb->asoc.numduptsns == 0) && (stcb->asoc.delayed_ack) && (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer))) { /* * CMT DAC algorithm: With CMT, delay acks * even in the face of * * reordering. Therefore, if acks that do not * have to be sent because of the above * reasons, will be delayed. That is, acks * that would have been sent due to gap * reports will be delayed with DAC. Start * the delayed ack timer. */ sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } else { /* * Ok we must build a SACK since the timer * is pending, we got our first packet OR * there are gaps or duplicates. */ (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); sctp_send_sack(stcb); } } else { if (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } } } } void sctp_service_queues(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; uint32_t tsize, pd_point; uint16_t nxt_todel; if (asoc->fragmented_delivery_inprogress) { sctp_service_reassembly(stcb, asoc); } /* Can we proceed further, i.e. the PD-API is complete */ if (asoc->fragmented_delivery_inprogress) { /* no */ return; } /* * Now is there some other chunk I can deliver from the reassembly * queue. */ doit_again: chk = TAILQ_FIRST(&asoc->reasmqueue); if (chk == NULL) { asoc->size_on_reasm_queue = 0; asoc->cnt_on_reasm_queue = 0; return; } nxt_todel = asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered + 1; if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) && ((nxt_todel == chk->rec.data.stream_seq) || (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED))) { /* * Yep the first one is here. We setup to start reception, * by backing down the TSN just in case we can't deliver. */ /* * Before we start though either all of the message should * be here or the socket buffer max or nothing on the * delivery queue and something can be delivered. */ if (stcb->sctp_socket) { pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket), stcb->sctp_ep->partial_delivery_point); } else { pd_point = stcb->sctp_ep->partial_delivery_point; } if (sctp_is_all_msg_on_reasm(asoc, &tsize) || (tsize >= pd_point)) { asoc->fragmented_delivery_inprogress = 1; asoc->tsn_last_delivered = chk->rec.data.TSN_seq - 1; asoc->str_of_pdapi = chk->rec.data.stream_number; asoc->ssn_of_pdapi = chk->rec.data.stream_seq; asoc->pdapi_ppid = chk->rec.data.payloadtype; asoc->fragment_flags = chk->rec.data.rcv_flags; sctp_service_reassembly(stcb, asoc); if (asoc->fragmented_delivery_inprogress == 0) { goto doit_again; } } } } int sctp_process_data(struct mbuf **mm, int iphlen, int *offset, int length, struct sctphdr *sh, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t * high_tsn) { struct sctp_data_chunk *ch, chunk_buf; struct sctp_association *asoc; int num_chunks = 0; /* number of control chunks processed */ int stop_proc = 0; int chk_length, break_flag, last_chunk; int abort_flag = 0, was_a_gap; struct mbuf *m; uint32_t highest_tsn; /* set the rwnd */ sctp_set_rwnd(stcb, &stcb->asoc); m = *mm; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { highest_tsn = asoc->highest_tsn_inside_nr_map; } else { highest_tsn = asoc->highest_tsn_inside_map; } was_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); /* * setup where we got the last DATA packet from for any SACK that * may need to go out. Don't bump the net. This is done ONLY when a * chunk is assigned. */ asoc->last_data_chunk_from = net; /*- * Now before we proceed we must figure out if this is a wasted * cluster... i.e. it is a small packet sent in and yet the driver * underneath allocated a full cluster for it. If so we must copy it * to a smaller mbuf and free up the cluster mbuf. This will help * with cluster starvation. Note for __Panda__ we don't do this * since it has clusters all the way down to 64 bytes. */ if (SCTP_BUF_LEN(m) < (long)MLEN && SCTP_BUF_NEXT(m) == NULL) { /* we only handle mbufs that are singletons.. not chains */ m = sctp_get_mbuf_for_msg(SCTP_BUF_LEN(m), 0, M_DONTWAIT, 1, MT_DATA); if (m) { /* ok lets see if we can copy the data up */ caddr_t *from, *to; /* get the pointers and copy */ to = mtod(m, caddr_t *); from = mtod((*mm), caddr_t *); memcpy(to, from, SCTP_BUF_LEN((*mm))); /* copy the length and free up the old */ SCTP_BUF_LEN(m) = SCTP_BUF_LEN((*mm)); sctp_m_freem(*mm); /* sucess, back copy */ *mm = m; } else { /* We are in trouble in the mbuf world .. yikes */ m = *mm; } } /* get pointer to the first chunk header */ ch = (struct sctp_data_chunk *)sctp_m_getptr(m, *offset, sizeof(struct sctp_data_chunk), (uint8_t *) & chunk_buf); if (ch == NULL) { return (1); } /* * process all DATA chunks... */ *high_tsn = asoc->cumulative_tsn; break_flag = 0; asoc->data_pkts_seen++; while (stop_proc == 0) { /* validate chunk length */ chk_length = ntohs(ch->ch.chunk_length); if (length - *offset < chk_length) { /* all done, mutulated chunk */ stop_proc = 1; break; } if (ch->ch.chunk_type == SCTP_DATA) { if ((size_t)chk_length < sizeof(struct sctp_data_chunk) + 1) { /* * Need to send an abort since we had a * invalid data chunk. */ struct mbuf *op_err; op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 2 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr) + (2 * sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_19); ippp++; *ippp = asoc->cumulative_tsn; } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_19; sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, 0, net->port); return (2); } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xB1, 0); #endif if (SCTP_SIZE32(chk_length) == (length - *offset)) { last_chunk = 1; } else { last_chunk = 0; } if (sctp_process_a_data_chunk(stcb, asoc, mm, *offset, ch, chk_length, net, high_tsn, &abort_flag, &break_flag, last_chunk)) { num_chunks++; } if (abort_flag) return (2); if (break_flag) { /* * Set because of out of rwnd space and no * drop rep space left. */ stop_proc = 1; break; } } else { /* not a data chunk in the data region */ switch (ch->ch.chunk_type) { case SCTP_INITIATION: case SCTP_INITIATION_ACK: case SCTP_SELECTIVE_ACK: case SCTP_NR_SELECTIVE_ACK: /* EY */ case SCTP_HEARTBEAT_REQUEST: case SCTP_HEARTBEAT_ACK: case SCTP_ABORT_ASSOCIATION: case SCTP_SHUTDOWN: case SCTP_SHUTDOWN_ACK: case SCTP_OPERATION_ERROR: case SCTP_COOKIE_ECHO: case SCTP_COOKIE_ACK: case SCTP_ECN_ECHO: case SCTP_ECN_CWR: case SCTP_SHUTDOWN_COMPLETE: case SCTP_AUTHENTICATION: case SCTP_ASCONF_ACK: case SCTP_PACKET_DROPPED: case SCTP_STREAM_RESET: case SCTP_FORWARD_CUM_TSN: case SCTP_ASCONF: /* * Now, what do we do with KNOWN chunks that * are NOT in the right place? * * For now, I do nothing but ignore them. We * may later want to add sysctl stuff to * switch out and do either an ABORT() or * possibly process them. */ if (SCTP_BASE_SYSCTL(sctp_strict_data_order)) { struct mbuf *op_err; op_err = sctp_generate_invmanparam(SCTP_CAUSE_PROTOCOL_VIOLATION); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, 0, net->port); return (2); } break; default: /* unknown chunk type, use bit rules */ if (ch->ch.chunk_type & 0x40) { /* Add a error report to the queue */ struct mbuf *merr; struct sctp_paramhdr *phd; merr = sctp_get_mbuf_for_msg(sizeof(*phd), 0, M_DONTWAIT, 1, MT_DATA); if (merr) { phd = mtod(merr, struct sctp_paramhdr *); /* * We cheat and use param * type since we did not * bother to define a error * cause struct. They are * the same basic format * with different names. */ phd->param_type = htons(SCTP_CAUSE_UNRECOG_CHUNK); phd->param_length = htons(chk_length + sizeof(*phd)); SCTP_BUF_LEN(merr) = sizeof(*phd); SCTP_BUF_NEXT(merr) = SCTP_M_COPYM(m, *offset, SCTP_SIZE32(chk_length), M_DONTWAIT); if (SCTP_BUF_NEXT(merr)) { sctp_queue_op_err(stcb, merr); } else { sctp_m_freem(merr); } } } if ((ch->ch.chunk_type & 0x80) == 0) { /* discard the rest of this packet */ stop_proc = 1; } /* else skip this bad chunk and * continue... */ break; }; /* switch of chunk type */ } *offset += SCTP_SIZE32(chk_length); if ((*offset >= length) || stop_proc) { /* no more data left in the mbuf chain */ stop_proc = 1; continue; } ch = (struct sctp_data_chunk *)sctp_m_getptr(m, *offset, sizeof(struct sctp_data_chunk), (uint8_t *) & chunk_buf); if (ch == NULL) { *offset = length; stop_proc = 1; break; } } /* while */ if (break_flag) { /* * we need to report rwnd overrun drops. */ sctp_send_packet_dropped(stcb, net, *mm, iphlen, 0); } if (num_chunks) { /* * Did we get data, if so update the time for auto-close and * give peer credit for being alive. */ SCTP_STAT_INCR(sctps_recvpktwithdata); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INDATA, __LINE__); } stcb->asoc.overall_error_count = 0; (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_last_rcvd); } /* now service all of the reassm queue if needed */ if (!(TAILQ_EMPTY(&asoc->reasmqueue))) sctp_service_queues(stcb, asoc); if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) { /* Assure that we ack right away */ stcb->asoc.send_sack = 1; } /* Start a sack timer or QUEUE a SACK for sending */ sctp_sack_check(stcb, was_a_gap, &abort_flag); if (abort_flag) return (2); return (0); } static int sctp_process_segment_range(struct sctp_tcb *stcb, struct sctp_tmit_chunk **p_tp1, uint32_t last_tsn, uint16_t frag_strt, uint16_t frag_end, int nr_sacking, int *num_frs, uint32_t * biggest_newly_acked_tsn, uint32_t * this_sack_lowest_newack, int *ecn_seg_sums) { struct sctp_tmit_chunk *tp1; unsigned int theTSN; int j, wake_him = 0, circled = 0; /* Recover the tp1 we last saw */ tp1 = *p_tp1; if (tp1 == NULL) { tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); } for (j = frag_strt; j <= frag_end; j++) { theTSN = j + last_tsn; while (tp1) { if (tp1->rec.data.doing_fast_retransmit) (*num_frs) += 1; /*- * CMT: CUCv2 algorithm. For each TSN being * processed from the sent queue, track the * next expected pseudo-cumack, or * rtx_pseudo_cumack, if required. Separate * cumack trackers for first transmissions, * and retransmissions. */ if ((tp1->whoTo->find_pseudo_cumack == 1) && (tp1->sent < SCTP_DATAGRAM_RESEND) && (tp1->snd_count == 1)) { tp1->whoTo->pseudo_cumack = tp1->rec.data.TSN_seq; tp1->whoTo->find_pseudo_cumack = 0; } if ((tp1->whoTo->find_rtx_pseudo_cumack == 1) && (tp1->sent < SCTP_DATAGRAM_RESEND) && (tp1->snd_count > 1)) { tp1->whoTo->rtx_pseudo_cumack = tp1->rec.data.TSN_seq; tp1->whoTo->find_rtx_pseudo_cumack = 0; } if (tp1->rec.data.TSN_seq == theTSN) { if (tp1->sent != SCTP_DATAGRAM_UNSENT) { /*- * must be held until * cum-ack passes */ - /*- - * ECN Nonce: Add the nonce - * value to the sender's - * nonce sum - */ if (tp1->sent < SCTP_DATAGRAM_RESEND) { /*- * If it is less than RESEND, it is * now no-longer in flight. * Higher values may already be set * via previous Gap Ack Blocks... * i.e. ACKED or RESEND. */ if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, *biggest_newly_acked_tsn)) { *biggest_newly_acked_tsn = tp1->rec.data.TSN_seq; } /*- * CMT: SFR algo (and HTNA) - set * saw_newack to 1 for dest being * newly acked. update * this_sack_highest_newack if * appropriate. */ if (tp1->rec.data.chunk_was_revoked == 0) tp1->whoTo->saw_newack = 1; if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, tp1->whoTo->this_sack_highest_newack)) { tp1->whoTo->this_sack_highest_newack = tp1->rec.data.TSN_seq; } /*- * CMT DAC algo: also update * this_sack_lowest_newack */ if (*this_sack_lowest_newack == 0) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(*this_sack_lowest_newack, last_tsn, tp1->rec.data.TSN_seq, 0, 0, SCTP_LOG_TSN_ACKED); } *this_sack_lowest_newack = tp1->rec.data.TSN_seq; } /*- * CMT: CUCv2 algorithm. If (rtx-)pseudo-cumack for corresp * dest is being acked, then we have a new (rtx-)pseudo-cumack. Set * new_(rtx_)pseudo_cumack to TRUE so that the cwnd for this dest can be * updated. Also trigger search for the next expected (rtx-)pseudo-cumack. * Separate pseudo_cumack trackers for first transmissions and * retransmissions. */ if (tp1->rec.data.TSN_seq == tp1->whoTo->pseudo_cumack) { if (tp1->rec.data.chunk_was_revoked == 0) { tp1->whoTo->new_pseudo_cumack = 1; } tp1->whoTo->find_pseudo_cumack = 1; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK); } if (tp1->rec.data.TSN_seq == tp1->whoTo->rtx_pseudo_cumack) { if (tp1->rec.data.chunk_was_revoked == 0) { tp1->whoTo->new_pseudo_cumack = 1; } tp1->whoTo->find_rtx_pseudo_cumack = 1; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(*biggest_newly_acked_tsn, last_tsn, tp1->rec.data.TSN_seq, frag_strt, frag_end, SCTP_LOG_TSN_ACKED); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_GAP, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } sctp_flight_size_decrease(tp1); sctp_total_flight_decrease(stcb, tp1); tp1->whoTo->net_ack += tp1->send_size; if (tp1->snd_count < 2) { /*- * True non-retransmited chunk */ tp1->whoTo->net_ack2 += tp1->send_size; /*- * update RTO too ? */ if (tp1->do_rtt) { tp1->whoTo->RTO = sctp_calculate_rto(stcb, &stcb->asoc, tp1->whoTo, &tp1->sent_rcv_time, sctp_align_safe_nocopy); tp1->do_rtt = 0; } } } if (tp1->sent <= SCTP_DATAGRAM_RESEND) { - (*ecn_seg_sums) += tp1->rec.data.ect_nonce; - (*ecn_seg_sums) &= SCTP_SACK_NONCE_SUM; if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, stcb->asoc.this_sack_highest_gap)) { stcb->asoc.this_sack_highest_gap = tp1->rec.data.TSN_seq; } if (tp1->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_decr(stcb->asoc.sent_queue_retran_cnt); #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xB2, (stcb->asoc.sent_queue_retran_cnt & 0x000000ff)); #endif } } /*- * All chunks NOT UNSENT fall through here and are marked * (leave PR-SCTP ones that are to skip alone though) */ if (tp1->sent != SCTP_FORWARD_TSN_SKIP) tp1->sent = SCTP_DATAGRAM_MARKED; if (tp1->rec.data.chunk_was_revoked) { /* deflate the cwnd */ tp1->whoTo->cwnd -= tp1->book_size; tp1->rec.data.chunk_was_revoked = 0; } /* NR Sack code here */ if (nr_sacking) { if (tp1->data) { /* * sa_ignore * NO_NULL_CHK */ sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); sctp_m_freem(tp1->data); tp1->data = NULL; } wake_him++; } } break; } /* if (tp1->TSN_seq == theTSN) */ if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, theTSN)) { break; } tp1 = TAILQ_NEXT(tp1, sctp_next); if ((tp1 == NULL) && (circled == 0)) { circled++; tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); } } /* end while (tp1) */ if (tp1 == NULL) { circled = 0; tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); } /* In case the fragments were not in order we must reset */ } /* end for (j = fragStart */ *p_tp1 = tp1; return (wake_him); /* Return value only used for nr-sack */ } static int sctp_handle_segments(struct mbuf *m, int *offset, struct sctp_tcb *stcb, struct sctp_association *asoc, uint32_t last_tsn, uint32_t * biggest_tsn_acked, uint32_t * biggest_newly_acked_tsn, uint32_t * this_sack_lowest_newack, int num_seg, int num_nr_seg, int *ecn_seg_sums) { struct sctp_gap_ack_block *frag, block; struct sctp_tmit_chunk *tp1; int i; int num_frs = 0; int chunk_freed; int non_revocable; uint16_t frag_strt, frag_end, prev_frag_end; tp1 = TAILQ_FIRST(&asoc->sent_queue); prev_frag_end = 0; chunk_freed = 0; for (i = 0; i < (num_seg + num_nr_seg); i++) { if (i == num_seg) { prev_frag_end = 0; tp1 = TAILQ_FIRST(&asoc->sent_queue); } frag = (struct sctp_gap_ack_block *)sctp_m_getptr(m, *offset, sizeof(struct sctp_gap_ack_block), (uint8_t *) & block); *offset += sizeof(block); if (frag == NULL) { return (chunk_freed); } frag_strt = ntohs(frag->start); frag_end = ntohs(frag->end); if (frag_strt > frag_end) { /* This gap report is malformed, skip it. */ continue; } if (frag_strt <= prev_frag_end) { /* This gap report is not in order, so restart. */ tp1 = TAILQ_FIRST(&asoc->sent_queue); } if (SCTP_TSN_GT((last_tsn + frag_end), *biggest_tsn_acked)) { *biggest_tsn_acked = last_tsn + frag_end; } if (i < num_seg) { non_revocable = 0; } else { non_revocable = 1; } if (sctp_process_segment_range(stcb, &tp1, last_tsn, frag_strt, frag_end, non_revocable, &num_frs, biggest_newly_acked_tsn, this_sack_lowest_newack, ecn_seg_sums)) { chunk_freed = 1; } prev_frag_end = frag_end; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { if (num_frs) sctp_log_fr(*biggest_tsn_acked, *biggest_newly_acked_tsn, last_tsn, SCTP_FR_LOG_BIGGEST_TSNS); } return (chunk_freed); } static void sctp_check_for_revoked(struct sctp_tcb *stcb, struct sctp_association *asoc, uint32_t cumack, uint32_t biggest_tsn_acked) { struct sctp_tmit_chunk *tp1; int tot_revoked = 0; TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, cumack)) { /* * ok this guy is either ACK or MARKED. If it is * ACKED it has been previously acked but not this * time i.e. revoked. If it is MARKED it was ACK'ed * again. */ if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, biggest_tsn_acked)) { break; } if (tp1->sent == SCTP_DATAGRAM_ACKED) { /* it has been revoked */ tp1->sent = SCTP_DATAGRAM_SENT; tp1->rec.data.chunk_was_revoked = 1; /* * We must add this stuff back in to assure * timers and such get started. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } sctp_flight_size_increase(tp1); sctp_total_flight_increase(stcb, tp1); /* * We inflate the cwnd to compensate for our * artificial inflation of the flight_size. */ tp1->whoTo->cwnd += tp1->book_size; tot_revoked++; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(asoc->last_acked_seq, cumack, tp1->rec.data.TSN_seq, 0, 0, SCTP_LOG_TSN_REVOKED); } } else if (tp1->sent == SCTP_DATAGRAM_MARKED) { /* it has been re-acked in this SACK */ tp1->sent = SCTP_DATAGRAM_ACKED; } } if (tp1->sent == SCTP_DATAGRAM_UNSENT) break; } - if (tot_revoked > 0) { - /* - * Setup the ecn nonce re-sync point. We do this since once - * data is revoked we begin to retransmit things, which do - * NOT have the ECN bits set. This means we are now out of - * sync and must wait until we get back in sync with the - * peer to check ECN bits. - */ - tp1 = TAILQ_FIRST(&asoc->send_queue); - if (tp1 == NULL) { - asoc->nonce_resync_tsn = asoc->sending_seq; - } else { - asoc->nonce_resync_tsn = tp1->rec.data.TSN_seq; - } - asoc->nonce_wait_for_ecne = 0; - asoc->nonce_sum_check = 0; - } } static void sctp_strike_gap_ack_chunks(struct sctp_tcb *stcb, struct sctp_association *asoc, uint32_t biggest_tsn_acked, uint32_t biggest_tsn_newly_acked, uint32_t this_sack_lowest_newack, int accum_moved) { struct sctp_tmit_chunk *tp1; int strike_flag = 0; struct timeval now; int tot_retrans = 0; uint32_t sending_seq; struct sctp_nets *net; int num_dests_sacked = 0; /* * select the sending_seq, this is either the next thing ready to be * sent but not transmitted, OR, the next seq we assign. */ tp1 = TAILQ_FIRST(&stcb->asoc.send_queue); if (tp1 == NULL) { sending_seq = asoc->sending_seq; } else { sending_seq = tp1->rec.data.TSN_seq; } /* CMT DAC algo: finding out if SACK is a mixed SACK */ if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (net->saw_newack) num_dests_sacked++; } } if (stcb->asoc.peer_supports_prsctp) { (void)SCTP_GETTIME_TIMEVAL(&now); } TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { strike_flag = 0; if (tp1->no_fr_allowed) { /* this one had a timeout or something */ continue; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { if (tp1->sent < SCTP_DATAGRAM_RESEND) sctp_log_fr(biggest_tsn_newly_acked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_CHECK_STRIKE); } if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, biggest_tsn_acked) || tp1->sent == SCTP_DATAGRAM_UNSENT) { /* done */ break; } if (stcb->asoc.peer_supports_prsctp) { if ((PR_SCTP_TTL_ENABLED(tp1->flags)) && tp1->sent < SCTP_DATAGRAM_ACKED) { /* Is it expired? */ if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) { /* Yes so drop it */ if (tp1->data != NULL) { (void)sctp_release_pr_sctp_chunk(stcb, tp1, (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT), SCTP_SO_NOT_LOCKED); } continue; } } } if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, asoc->this_sack_highest_gap)) { /* we are beyond the tsn in the sack */ break; } if (tp1->sent >= SCTP_DATAGRAM_RESEND) { /* either a RESEND, ACKED, or MARKED */ /* skip */ if (tp1->sent == SCTP_FORWARD_TSN_SKIP) { /* Continue strikin FWD-TSN chunks */ tp1->rec.data.fwd_tsn_cnt++; } continue; } /* * CMT : SFR algo (covers part of DAC and HTNA as well) */ if (tp1->whoTo && tp1->whoTo->saw_newack == 0) { /* * No new acks were receieved for data sent to this * dest. Therefore, according to the SFR algo for * CMT, no data sent to this dest can be marked for * FR using this SACK. */ continue; } else if (tp1->whoTo && SCTP_TSN_GT(tp1->rec.data.TSN_seq, tp1->whoTo->this_sack_highest_newack)) { /* * CMT: New acks were receieved for data sent to * this dest. But no new acks were seen for data * sent after tp1. Therefore, according to the SFR * algo for CMT, tp1 cannot be marked for FR using * this SACK. This step covers part of the DAC algo * and the HTNA algo as well. */ continue; } /* * Here we check to see if we were have already done a FR * and if so we see if the biggest TSN we saw in the sack is * smaller than the recovery point. If so we don't strike * the tsn... otherwise we CAN strike the TSN. */ /* * @@@ JRI: Check for CMT if (accum_moved && * asoc->fast_retran_loss_recovery && (sctp_cmt_on_off == * 0)) { */ if (accum_moved && asoc->fast_retran_loss_recovery) { /* * Strike the TSN if in fast-recovery and cum-ack * moved. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(biggest_tsn_newly_acked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } if (tp1->sent < SCTP_DATAGRAM_RESEND) { tp1->sent++; } if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { /* * CMT DAC algorithm: If SACK flag is set to * 0, then lowest_newack test will not pass * because it would have been set to the * cumack earlier. If not already to be * rtx'd, If not a mixed sack and if tp1 is * not between two sacked TSNs, then mark by * one more. NOTE that we are marking by one * additional time since the SACK DAC flag * indicates that two packets have been * received after this missing TSN. */ if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) && SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.TSN_seq)) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(16 + num_dests_sacked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } tp1->sent++; } } } else if ((tp1->rec.data.doing_fast_retransmit) && (asoc->sctp_cmt_on_off == 0)) { /* * For those that have done a FR we must take * special consideration if we strike. I.e the * biggest_newly_acked must be higher than the * sending_seq at the time we did the FR. */ if ( #ifdef SCTP_FR_TO_ALTERNATE /* * If FR's go to new networks, then we must only do * this for singly homed asoc's. However if the FR's * go to the same network (Armando's work) then its * ok to FR multiple times. */ (asoc->numnets < 2) #else (1) #endif ) { if (SCTP_TSN_GE(biggest_tsn_newly_acked, tp1->rec.data.fast_retran_tsn)) { /* * Strike the TSN, since this ack is * beyond where things were when we * did a FR. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(biggest_tsn_newly_acked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } if (tp1->sent < SCTP_DATAGRAM_RESEND) { tp1->sent++; } strike_flag = 1; if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { /* * CMT DAC algorithm: If * SACK flag is set to 0, * then lowest_newack test * will not pass because it * would have been set to * the cumack earlier. If * not already to be rtx'd, * If not a mixed sack and * if tp1 is not between two * sacked TSNs, then mark by * one more. NOTE that we * are marking by one * additional time since the * SACK DAC flag indicates * that two packets have * been received after this * missing TSN. */ if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) && SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.TSN_seq)) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(32 + num_dests_sacked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } if (tp1->sent < SCTP_DATAGRAM_RESEND) { tp1->sent++; } } } } } /* * JRI: TODO: remove code for HTNA algo. CMT's SFR * algo covers HTNA. */ } else if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, biggest_tsn_newly_acked)) { /* * We don't strike these: This is the HTNA * algorithm i.e. we don't strike If our TSN is * larger than the Highest TSN Newly Acked. */ ; } else { /* Strike the TSN */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(biggest_tsn_newly_acked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } if (tp1->sent < SCTP_DATAGRAM_RESEND) { tp1->sent++; } if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { /* * CMT DAC algorithm: If SACK flag is set to * 0, then lowest_newack test will not pass * because it would have been set to the * cumack earlier. If not already to be * rtx'd, If not a mixed sack and if tp1 is * not between two sacked TSNs, then mark by * one more. NOTE that we are marking by one * additional time since the SACK DAC flag * indicates that two packets have been * received after this missing TSN. */ if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) && SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.TSN_seq)) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(48 + num_dests_sacked, tp1->rec.data.TSN_seq, tp1->sent, SCTP_FR_LOG_STRIKE_CHUNK); } tp1->sent++; } } } if (tp1->sent == SCTP_DATAGRAM_RESEND) { struct sctp_nets *alt; /* fix counts and things */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND, (tp1->whoTo ? (tp1->whoTo->flight_size) : 0), tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } if (tp1->whoTo) { tp1->whoTo->net_ack++; sctp_flight_size_decrease(tp1); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd(SCTP_INCREASE_PEER_RWND, asoc->peers_rwnd, tp1->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); } /* add back to the rwnd */ asoc->peers_rwnd += (tp1->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); /* remove from the total flight */ sctp_total_flight_decrease(stcb, tp1); if ((stcb->asoc.peer_supports_prsctp) && (PR_SCTP_RTX_ENABLED(tp1->flags))) { /* * Has it been retransmitted tv_sec times? - * we store the retran count there. */ if (tp1->snd_count > tp1->rec.data.timetodrop.tv_sec) { /* Yes, so drop it */ if (tp1->data != NULL) { (void)sctp_release_pr_sctp_chunk(stcb, tp1, (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT), SCTP_SO_NOT_LOCKED); } /* Make sure to flag we had a FR */ tp1->whoTo->net_ack++; continue; } } /* printf("OK, we are now ready to FR this guy\n"); */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(tp1->rec.data.TSN_seq, tp1->snd_count, 0, SCTP_FR_MARKED); } if (strike_flag) { /* This is a subsequent FR */ SCTP_STAT_INCR(sctps_sendmultfastretrans); } sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); if (asoc->sctp_cmt_on_off > 0) { /* * CMT: Using RTX_SSTHRESH policy for CMT. * If CMT is being used, then pick dest with * largest ssthresh for any retransmission. */ tp1->no_fr_allowed = 1; alt = tp1->whoTo; /* sa_ignore NO_NULL_CHK */ if (asoc->sctp_cmt_pf > 0) { /* * JRS 5/18/07 - If CMT PF is on, * use the PF version of * find_alt_net() */ alt = sctp_find_alternate_net(stcb, alt, 2); } else { /* * JRS 5/18/07 - If only CMT is on, * use the CMT version of * find_alt_net() */ /* sa_ignore NO_NULL_CHK */ alt = sctp_find_alternate_net(stcb, alt, 1); } if (alt == NULL) { alt = tp1->whoTo; } /* * CUCv2: If a different dest is picked for * the retransmission, then new * (rtx-)pseudo_cumack needs to be tracked * for orig dest. Let CUCv2 track new (rtx-) * pseudo-cumack always. */ if (tp1->whoTo) { tp1->whoTo->find_pseudo_cumack = 1; tp1->whoTo->find_rtx_pseudo_cumack = 1; } } else {/* CMT is OFF */ #ifdef SCTP_FR_TO_ALTERNATE /* Can we find an alternate? */ alt = sctp_find_alternate_net(stcb, tp1->whoTo, 0); #else /* * default behavior is to NOT retransmit * FR's to an alternate. Armando Caro's * paper details why. */ alt = tp1->whoTo; #endif } tp1->rec.data.doing_fast_retransmit = 1; tot_retrans++; /* mark the sending seq for possible subsequent FR's */ /* * printf("Marking TSN for FR new value %x\n", * (uint32_t)tpi->rec.data.TSN_seq); */ if (TAILQ_EMPTY(&asoc->send_queue)) { /* * If the queue of send is empty then its * the next sequence number that will be * assigned so we subtract one from this to * get the one we last sent. */ tp1->rec.data.fast_retran_tsn = sending_seq; } else { /* * If there are chunks on the send queue * (unsent data that has made it from the * stream queues but not out the door, we * take the first one (which will have the * lowest TSN) and subtract one to get the * one we last sent. */ struct sctp_tmit_chunk *ttt; ttt = TAILQ_FIRST(&asoc->send_queue); tp1->rec.data.fast_retran_tsn = ttt->rec.data.TSN_seq; } if (tp1->do_rtt) { /* * this guy had a RTO calculation pending on * it, cancel it */ tp1->do_rtt = 0; } if (alt != tp1->whoTo) { /* yes, there is an alternate. */ sctp_free_remote_addr(tp1->whoTo); /* sa_ignore FREED_MEMORY */ tp1->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } } } - - if (tot_retrans > 0) { - /* - * Setup the ecn nonce re-sync point. We do this since once - * we go to FR something we introduce a Karn's rule scenario - * and won't know the totals for the ECN bits. - */ - asoc->nonce_resync_tsn = sending_seq; - asoc->nonce_wait_for_ecne = 0; - asoc->nonce_sum_check = 0; - } } struct sctp_tmit_chunk * sctp_try_advance_peer_ack_point(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *tp1, *tp2, *a_adv = NULL; struct timeval now; int now_filled = 0; if (asoc->peer_supports_prsctp == 0) { return (NULL); } TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { if (tp1->sent != SCTP_FORWARD_TSN_SKIP && tp1->sent != SCTP_DATAGRAM_RESEND) { /* no chance to advance, out of here */ break; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { if (tp1->sent == SCTP_FORWARD_TSN_SKIP) { sctp_misc_ints(SCTP_FWD_TSN_CHECK, asoc->advanced_peer_ack_point, tp1->rec.data.TSN_seq, 0, 0); } } if (!PR_SCTP_ENABLED(tp1->flags)) { /* * We can't fwd-tsn past any that are reliable aka * retransmitted until the asoc fails. */ break; } if (!now_filled) { (void)SCTP_GETTIME_TIMEVAL(&now); now_filled = 1; } /* * now we got a chunk which is marked for another * retransmission to a PR-stream but has run out its chances * already maybe OR has been marked to skip now. Can we skip * it if its a resend? */ if (tp1->sent == SCTP_DATAGRAM_RESEND && (PR_SCTP_TTL_ENABLED(tp1->flags))) { /* * Now is this one marked for resend and its time is * now up? */ if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) { /* Yes so drop it */ if (tp1->data) { (void)sctp_release_pr_sctp_chunk(stcb, tp1, (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT), SCTP_SO_NOT_LOCKED); } } else { /* * No, we are done when hit one for resend * whos time as not expired. */ break; } } /* * Ok now if this chunk is marked to drop it we can clean up * the chunk, advance our peer ack point and we can check * the next chunk. */ if (tp1->sent == SCTP_FORWARD_TSN_SKIP) { /* advance PeerAckPoint goes forward */ if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, asoc->advanced_peer_ack_point)) { asoc->advanced_peer_ack_point = tp1->rec.data.TSN_seq; a_adv = tp1; } else if (tp1->rec.data.TSN_seq == asoc->advanced_peer_ack_point) { /* No update but we do save the chk */ a_adv = tp1; } } else { /* * If it is still in RESEND we can advance no * further */ break; } } return (a_adv); } static int sctp_fs_audit(struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; int inflight = 0, resend = 0, inbetween = 0, acked = 0, above = 0; int entry_flight, entry_cnt, ret; entry_flight = asoc->total_flight; entry_cnt = asoc->total_flight_count; ret = 0; if (asoc->pr_sctp_cnt >= asoc->sent_queue_cnt) return (0); TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->sent < SCTP_DATAGRAM_RESEND) { printf("Chk TSN:%u size:%d inflight cnt:%d\n", chk->rec.data.TSN_seq, chk->send_size, chk->snd_count ); inflight++; } else if (chk->sent == SCTP_DATAGRAM_RESEND) { resend++; } else if (chk->sent < SCTP_DATAGRAM_ACKED) { inbetween++; } else if (chk->sent > SCTP_DATAGRAM_ACKED) { above++; } else { acked++; } } if ((inflight > 0) || (inbetween > 0)) { #ifdef INVARIANTS panic("Flight size-express incorrect? \n"); #else printf("asoc->total_flight:%d cnt:%d\n", entry_flight, entry_cnt); SCTP_PRINTF("Flight size-express incorrect F:%d I:%d R:%d Ab:%d ACK:%d\n", inflight, inbetween, resend, above, acked); ret = 1; #endif } return (ret); } static void sctp_window_probe_recovery(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_nets *net, struct sctp_tmit_chunk *tp1) { tp1->window_probe = 0; if ((tp1->sent >= SCTP_DATAGRAM_ACKED) || (tp1->data == NULL)) { /* TSN's skipped we do NOT move back. */ sctp_misc_ints(SCTP_FLIGHT_LOG_DWN_WP_FWD, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); return; } /* First setup this by shrinking flight */ sctp_flight_size_decrease(tp1); sctp_total_flight_decrease(stcb, tp1); /* Now mark for resend */ tp1->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(asoc->sent_queue_retran_cnt); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_WP, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } } void sctp_express_handle_sack(struct sctp_tcb *stcb, uint32_t cumack, - uint32_t rwnd, int nonce_sum_flag, int *abort_now) + uint32_t rwnd, int *abort_now) { struct sctp_nets *net; struct sctp_association *asoc; struct sctp_tmit_chunk *tp1, *tp2; uint32_t old_rwnd; int win_probe_recovery = 0; int win_probe_recovered = 0; int j, done_once = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) { sctp_misc_ints(SCTP_SACK_LOG_EXPRESS, cumack, rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd); } SCTP_TCB_LOCK_ASSERT(stcb); #ifdef SCTP_ASOCLOG_OF_TSNS stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cumack; stcb->asoc.cumack_log_at++; if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) { stcb->asoc.cumack_log_at = 0; } #endif asoc = &stcb->asoc; old_rwnd = asoc->peers_rwnd; if (SCTP_TSN_GT(asoc->last_acked_seq, cumack)) { /* old ack */ return; } else if (asoc->last_acked_seq == cumack) { /* Window update sack */ asoc->peers_rwnd = sctp_sbspace_sub(rwnd, (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } if (asoc->peers_rwnd > old_rwnd) { goto again; } return; } /* First setup for CC stuff */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (SCTP_TSN_GT(cumack, net->cwr_window_tsn)) { /* Drag along the window_tsn for cwr's */ net->cwr_window_tsn = cumack; } net->prev_cwnd = net->cwnd; net->net_ack = 0; net->net_ack2 = 0; /* * CMT: Reset CUC and Fast recovery algo variables before * SACK processing */ net->new_pseudo_cumack = 0; net->will_exit_fast_recovery = 0; } if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) { uint32_t send_s; if (!TAILQ_EMPTY(&asoc->sent_queue)) { tp1 = TAILQ_LAST(&asoc->sent_queue, sctpchunk_listhead); send_s = tp1->rec.data.TSN_seq + 1; } else { send_s = asoc->sending_seq; } if (SCTP_TSN_GE(cumack, send_s)) { #ifndef INVARIANTS struct mbuf *oper; #endif #ifdef INVARIANTS panic("Impossible sack 1"); #else *abort_now = 1; /* XXX */ oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_25); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_25; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); return; #endif } } asoc->this_sack_highest_gap = cumack; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INDATA, __LINE__); } stcb->asoc.overall_error_count = 0; if (SCTP_TSN_GT(cumack, asoc->last_acked_seq)) { /* process the new consecutive TSN first */ TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { if (SCTP_TSN_GE(cumack, tp1->rec.data.TSN_seq)) { if (tp1->sent == SCTP_DATAGRAM_UNSENT) { printf("Warning, an unsent is now acked?\n"); } - /* - * ECN Nonce: Add the nonce to the sender's - * nonce sum - */ - asoc->nonce_sum_expect_base += tp1->rec.data.ect_nonce; if (tp1->sent < SCTP_DATAGRAM_ACKED) { /* * If it is less than ACKED, it is * now no-longer in flight. Higher * values may occur during marking */ if (tp1->sent < SCTP_DATAGRAM_RESEND) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } sctp_flight_size_decrease(tp1); /* sa_ignore NO_NULL_CHK */ sctp_total_flight_decrease(stcb, tp1); } tp1->whoTo->net_ack += tp1->send_size; if (tp1->snd_count < 2) { /* * True non-retransmited * chunk */ tp1->whoTo->net_ack2 += tp1->send_size; /* update RTO too? */ if (tp1->do_rtt) { tp1->whoTo->RTO = /* * sa_ignore * NO_NULL_CHK */ sctp_calculate_rto(stcb, asoc, tp1->whoTo, &tp1->sent_rcv_time, sctp_align_safe_nocopy); tp1->do_rtt = 0; } } /* * CMT: CUCv2 algorithm. From the * cumack'd TSNs, for each TSN being * acked for the first time, set the * following variables for the * corresp destination. * new_pseudo_cumack will trigger a * cwnd update. * find_(rtx_)pseudo_cumack will * trigger search for the next * expected (rtx-)pseudo-cumack. */ tp1->whoTo->new_pseudo_cumack = 1; tp1->whoTo->find_pseudo_cumack = 1; tp1->whoTo->find_rtx_pseudo_cumack = 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { /* sa_ignore NO_NULL_CHK */ sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK); } } if (tp1->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_decr(asoc->sent_queue_retran_cnt); } if (tp1->rec.data.chunk_was_revoked) { /* deflate the cwnd */ tp1->whoTo->cwnd -= tp1->book_size; tp1->rec.data.chunk_was_revoked = 0; } tp1->sent = SCTP_DATAGRAM_ACKED; TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next); if (tp1->data) { /* sa_ignore NO_NULL_CHK */ sctp_free_bufspace(stcb, asoc, tp1, 1); sctp_m_freem(tp1->data); tp1->data = NULL; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(asoc->last_acked_seq, cumack, tp1->rec.data.TSN_seq, 0, 0, SCTP_LOG_FREE_SENT); } asoc->sent_queue_cnt--; sctp_free_a_chunk(stcb, tp1); } else { break; } } } /* sa_ignore NO_NULL_CHK */ if (stcb->sctp_socket) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif SOCKBUF_LOCK(&stcb->sctp_socket->so_snd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { /* sa_ignore NO_NULL_CHK */ sctp_wakeup_log(stcb, cumack, 1, SCTP_WAKESND_FROM_SACK); } #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return; } #endif sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { sctp_wakeup_log(stcb, cumack, 1, SCTP_NOWAKE_FROM_SACK); } } /* JRS - Use the congestion control given in the CC module */ if (asoc->last_acked_seq != cumack) asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, 1, 0, 0); asoc->last_acked_seq = cumack; if (TAILQ_EMPTY(&asoc->sent_queue)) { /* nothing left in-flight */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { net->flight_size = 0; net->partial_bytes_acked = 0; } asoc->total_flight = 0; asoc->total_flight_count = 0; } - /* ECN Nonce updates */ - if (asoc->ecn_nonce_allowed) { - if (asoc->nonce_sum_check) { - if (nonce_sum_flag != ((asoc->nonce_sum_expect_base) & SCTP_SACK_NONCE_SUM)) { - if (asoc->nonce_wait_for_ecne == 0) { - struct sctp_tmit_chunk *lchk; - - lchk = TAILQ_FIRST(&asoc->send_queue); - asoc->nonce_wait_for_ecne = 1; - if (lchk) { - asoc->nonce_wait_tsn = lchk->rec.data.TSN_seq; - } else { - asoc->nonce_wait_tsn = asoc->sending_seq; - } - } else { - if (SCTP_TSN_GE(asoc->last_acked_seq, asoc->nonce_wait_tsn)) { - /* - * Misbehaving peer. We need - * to react to this guy - */ - asoc->ecn_allowed = 0; - asoc->ecn_nonce_allowed = 0; - } - } - } - } else { - /* See if Resynchronization Possible */ - if (SCTP_TSN_GT(asoc->last_acked_seq, asoc->nonce_resync_tsn)) { - asoc->nonce_sum_check = 1; - /* - * Now we must calculate what the base is. - * We do this based on two things, we know - * the total's for all the segments - * gap-acked in the SACK (none). We also - * know the SACK's nonce sum, its in - * nonce_sum_flag. So we can build a truth - * table to back-calculate the new value of - * asoc->nonce_sum_expect_base: - * - * SACK-flag-Value Seg-Sums Base 0 0 0 - * 1 0 1 0 1 1 1 - * 1 0 - */ - asoc->nonce_sum_expect_base = (0 ^ nonce_sum_flag) & SCTP_SACK_NONCE_SUM; - } - } - } /* RWND update */ asoc->peers_rwnd = sctp_sbspace_sub(rwnd, (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } if (asoc->peers_rwnd > old_rwnd) { win_probe_recovery = 1; } /* Now assure a timer where data is queued at */ again: j = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { int to_ticks; if (win_probe_recovery && (net->window_probe)) { win_probe_recovered = 1; /* * Find first chunk that was used with window probe * and clear the sent */ /* sa_ignore FREED_MEMORY */ TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (tp1->window_probe) { /* move back to data send queue */ sctp_window_probe_recovery(stcb, asoc, net, tp1); break; } } } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } if (net->flight_size) { j++; (void)SCTP_OS_TIMER_START(&net->rxt_timer.timer, to_ticks, sctp_timeout_handler, &net->rxt_timer); if (net->window_probe) { net->window_probe = 0; } } else { if (net->window_probe) { /* * In window probes we must assure a timer * is still running there */ net->window_probe = 0; if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { SCTP_OS_TIMER_START(&net->rxt_timer.timer, to_ticks, sctp_timeout_handler, &net->rxt_timer); } } else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_22); } if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck4); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_23); } } } } if ((j == 0) && (!TAILQ_EMPTY(&asoc->sent_queue)) && (asoc->sent_queue_retran_cnt == 0) && (win_probe_recovered == 0) && (done_once == 0)) { /* * huh, this should not happen unless all packets are * PR-SCTP and marked to skip of course. */ if (sctp_fs_audit(asoc)) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { net->flight_size = 0; } asoc->total_flight = 0; asoc->total_flight_count = 0; asoc->sent_queue_retran_cnt = 0; TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (tp1->sent < SCTP_DATAGRAM_RESEND) { sctp_flight_size_increase(tp1); sctp_total_flight_increase(stcb, tp1); } else if (tp1->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(asoc->sent_queue_retran_cnt); } } } done_once = 1; goto again; } /**********************************/ /* Now what about shutdown issues */ /**********************************/ if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) { /* nothing left on sendqueue.. consider done */ /* clean up */ if ((asoc->stream_queue_cnt == 1) && ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED)) && (asoc->locked_on_sending) ) { struct sctp_stream_queue_pending *sp; /* * I may be in a state where we got all across.. but * cannot write more due to a shutdown... we abort * since the user did not indicate EOR in this case. * The sp will be cleaned during free of the asoc. */ sp = TAILQ_LAST(&((asoc->locked_on_sending)->outqueue), sctp_streamhead); if ((sp) && (sp->length == 0)) { /* Let cleanup code purge it */ if (sp->msg_is_complete) { asoc->stream_queue_cnt--; } else { asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; asoc->locked_on_sending = NULL; asoc->stream_queue_cnt--; } } } if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) && (asoc->stream_queue_cnt == 0)) { if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) { /* Need to abort here */ struct mbuf *oper; abort_out_now: *abort_now = 1; /* XXX */ oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_24); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_24; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, oper, SCTP_SO_NOT_LOCKED); } else { if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(stcb); sctp_send_shutdown(stcb, stcb->asoc.primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } else if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) && (asoc->stream_queue_cnt == 0)) { if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) { goto abort_out_now; } SCTP_STAT_DECR_GAUGE32(sctps_currestab); SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_ACK_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); sctp_stop_timers_for_shutdown(stcb); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, stcb->sctp_ep, stcb, asoc->primary_destination); } } /*********************************************/ /* Here we perform PR-SCTP procedures */ /* (section 4.2) */ /*********************************************/ /* C1. update advancedPeerAckPoint */ if (SCTP_TSN_GT(cumack, asoc->advanced_peer_ack_point)) { asoc->advanced_peer_ack_point = cumack; } /* PR-Sctp issues need to be addressed too */ if ((asoc->peer_supports_prsctp) && (asoc->pr_sctp_cnt > 0)) { struct sctp_tmit_chunk *lchk; uint32_t old_adv_peer_ack_point; old_adv_peer_ack_point = asoc->advanced_peer_ack_point; lchk = sctp_try_advance_peer_ack_point(stcb, asoc); /* C3. See if we need to send a Fwd-TSN */ if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cumack)) { /* - * ISSUE with ECN, see FWD-TSN processing for notes - * on issues that will occur when the ECN NONCE - * stuff is put into SCTP for cross checking. + * ISSUE with ECN, see FWD-TSN processing. */ if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) { send_forward_tsn(stcb, asoc); - /* - * ECN Nonce: Disable Nonce Sum check when - * FWD TSN is sent and store resync tsn - */ - asoc->nonce_sum_check = 0; - asoc->nonce_resync_tsn = asoc->advanced_peer_ack_point; } else if (lchk) { /* try to FR fwd-tsn's that get lost too */ if (lchk->rec.data.fwd_tsn_cnt >= 3) { send_forward_tsn(stcb, asoc); } } } if (lchk) { /* Assure a timer is up */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo); } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SACK_RWND_UPDATE, rwnd, stcb->asoc.peers_rwnd, stcb->asoc.total_flight, stcb->asoc.total_output_queue_size); } } void sctp_handle_sack(struct mbuf *m, int offset_seg, int offset_dup, struct sctp_tcb *stcb, struct sctp_nets *net_from, uint16_t num_seg, uint16_t num_nr_seg, uint16_t num_dup, int *abort_now, uint8_t flags, uint32_t cum_ack, uint32_t rwnd) { struct sctp_association *asoc; struct sctp_tmit_chunk *tp1, *tp2; uint32_t last_tsn, biggest_tsn_acked, biggest_tsn_newly_acked, this_sack_lowest_newack; uint32_t sav_cum_ack; uint16_t wake_him = 0; uint32_t send_s = 0; long j; int accum_moved = 0; int will_exit_fast_recovery = 0; uint32_t a_rwnd, old_rwnd; int win_probe_recovery = 0; int win_probe_recovered = 0; struct sctp_nets *net = NULL; - int nonce_sum_flag, ecn_seg_sums = 0; + int ecn_seg_sums = 0; int done_once; uint8_t reneged_all = 0; uint8_t cmt_dac_flag; /* * we take any chance we can to service our queues since we cannot * get awoken when the socket is read from :< */ /* * Now perform the actual SACK handling: 1) Verify that it is not an * old sack, if so discard. 2) If there is nothing left in the send * queue (cum-ack is equal to last acked) then you have a duplicate * too, update any rwnd change and verify no timers are running. * then return. 3) Process any new consequtive data i.e. cum-ack * moved process these first and note that it moved. 4) Process any * sack blocks. 5) Drop any acked from the queue. 6) Check for any * revoked blocks and mark. 7) Update the cwnd. 8) Nothing left, * sync up flightsizes and things, stop all timers and also check * for shutdown_pending state. If so then go ahead and send off the * shutdown. If in shutdown recv, send off the shutdown-ack and * start that timer, Ret. 9) Strike any non-acked things and do FR * procedure if needed being sure to set the FR flag. 10) Do pr-sctp * procedures. 11) Apply any FR penalties. 12) Assure we will SACK * if in shutdown_recv state. */ SCTP_TCB_LOCK_ASSERT(stcb); /* CMT DAC algo */ this_sack_lowest_newack = 0; j = 0; SCTP_STAT_INCR(sctps_slowpath_sack); last_tsn = cum_ack; - nonce_sum_flag = flags & SCTP_SACK_NONCE_SUM; cmt_dac_flag = flags & SCTP_SACK_CMT_DAC; #ifdef SCTP_ASOCLOG_OF_TSNS stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cum_ack; stcb->asoc.cumack_log_at++; if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) { stcb->asoc.cumack_log_at = 0; } #endif a_rwnd = rwnd; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) { sctp_misc_ints(SCTP_SACK_LOG_NORMAL, cum_ack, rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd); } old_rwnd = stcb->asoc.peers_rwnd; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INDATA, __LINE__); } stcb->asoc.overall_error_count = 0; asoc = &stcb->asoc; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(asoc->last_acked_seq, cum_ack, 0, num_seg, num_dup, SCTP_LOG_NEW_SACK); } if ((num_dup) && (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_FR_LOGGING_ENABLE | SCTP_EARLYFR_LOGGING_ENABLE))) { uint16_t i; uint32_t *dupdata, dblock; for (i = 0; i < num_dup; i++) { dupdata = (uint32_t *) sctp_m_getptr(m, offset_dup + i * sizeof(uint32_t), sizeof(uint32_t), (uint8_t *) & dblock); if (dupdata == NULL) { break; } sctp_log_fr(*dupdata, 0, 0, SCTP_FR_DUPED); } } if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) { /* reality check */ if (!TAILQ_EMPTY(&asoc->sent_queue)) { tp1 = TAILQ_LAST(&asoc->sent_queue, sctpchunk_listhead); send_s = tp1->rec.data.TSN_seq + 1; } else { tp1 = NULL; send_s = asoc->sending_seq; } if (SCTP_TSN_GE(cum_ack, send_s)) { struct mbuf *oper; /* * no way, we have not even sent this TSN out yet. * Peer is hopelessly messed up with us. */ printf("NEW cum_ack:%x send_s:%x is smaller or equal\n", cum_ack, send_s); if (tp1) { printf("Got send_s from tsn:%x + 1 of tp1:%p\n", tp1->rec.data.TSN_seq, tp1); } hopeless_peer: *abort_now = 1; /* XXX */ oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_25); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_25; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); return; } } /**********************/ /* 1) check the range */ /**********************/ if (SCTP_TSN_GT(asoc->last_acked_seq, last_tsn)) { /* acking something behind */ return; } sav_cum_ack = asoc->last_acked_seq; /* update the Rwnd of the peer */ if (TAILQ_EMPTY(&asoc->sent_queue) && TAILQ_EMPTY(&asoc->send_queue) && (asoc->stream_queue_cnt == 0)) { /* nothing left on send/sent and strmq */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, asoc->peers_rwnd, 0, 0, a_rwnd); } asoc->peers_rwnd = a_rwnd; if (asoc->sent_queue_retran_cnt) { asoc->sent_queue_retran_cnt = 0; } if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } /* stop any timers */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_26); if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck1); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_26); } } net->partial_bytes_acked = 0; net->flight_size = 0; } asoc->total_flight = 0; asoc->total_flight_count = 0; return; } /* * We init netAckSz and netAckSz2 to 0. These are used to track 2 * things. The total byte count acked is tracked in netAckSz AND * netAck2 is used to track the total bytes acked that are un- * amibguious and were never retransmitted. We track these on a per * destination address basis. */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (SCTP_TSN_GT(cum_ack, net->cwr_window_tsn)) { /* Drag along the window_tsn for cwr's */ net->cwr_window_tsn = cum_ack; } net->prev_cwnd = net->cwnd; net->net_ack = 0; net->net_ack2 = 0; /* * CMT: Reset CUC and Fast recovery algo variables before * SACK processing */ net->new_pseudo_cumack = 0; net->will_exit_fast_recovery = 0; } /* process the new consecutive TSN first */ TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (SCTP_TSN_GE(last_tsn, tp1->rec.data.TSN_seq)) { if (tp1->sent != SCTP_DATAGRAM_UNSENT) { - /* - * ECN Nonce: Add the nonce to the sender's - * nonce sum - */ - asoc->nonce_sum_expect_base += tp1->rec.data.ect_nonce; accum_moved = 1; if (tp1->sent < SCTP_DATAGRAM_ACKED) { /* * If it is less than ACKED, it is * now no-longer in flight. Higher * values may occur during marking */ if ((tp1->whoTo->dest_state & SCTP_ADDR_UNCONFIRMED) && (tp1->snd_count < 2)) { /* * If there was no retran * and the address is * un-confirmed and we sent * there and are now * sacked.. its confirmed, * mark it so. */ tp1->whoTo->dest_state &= ~SCTP_ADDR_UNCONFIRMED; } if (tp1->sent < SCTP_DATAGRAM_RESEND) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } sctp_flight_size_decrease(tp1); sctp_total_flight_decrease(stcb, tp1); } tp1->whoTo->net_ack += tp1->send_size; /* CMT SFR and DAC algos */ this_sack_lowest_newack = tp1->rec.data.TSN_seq; tp1->whoTo->saw_newack = 1; if (tp1->snd_count < 2) { /* * True non-retransmited * chunk */ tp1->whoTo->net_ack2 += tp1->send_size; /* update RTO too? */ if (tp1->do_rtt) { tp1->whoTo->RTO = sctp_calculate_rto(stcb, asoc, tp1->whoTo, &tp1->sent_rcv_time, sctp_align_safe_nocopy); tp1->do_rtt = 0; } } /* * CMT: CUCv2 algorithm. From the * cumack'd TSNs, for each TSN being * acked for the first time, set the * following variables for the * corresp destination. * new_pseudo_cumack will trigger a * cwnd update. * find_(rtx_)pseudo_cumack will * trigger search for the next * expected (rtx-)pseudo-cumack. */ tp1->whoTo->new_pseudo_cumack = 1; tp1->whoTo->find_pseudo_cumack = 1; tp1->whoTo->find_rtx_pseudo_cumack = 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(asoc->last_acked_seq, cum_ack, tp1->rec.data.TSN_seq, 0, 0, SCTP_LOG_TSN_ACKED); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK); } } if (tp1->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_decr(asoc->sent_queue_retran_cnt); #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xB3, (asoc->sent_queue_retran_cnt & 0x000000ff)); #endif } if (tp1->rec.data.chunk_was_revoked) { /* deflate the cwnd */ tp1->whoTo->cwnd -= tp1->book_size; tp1->rec.data.chunk_was_revoked = 0; } tp1->sent = SCTP_DATAGRAM_ACKED; } } else { break; } } biggest_tsn_newly_acked = biggest_tsn_acked = last_tsn; /* always set this up to cum-ack */ asoc->this_sack_highest_gap = last_tsn; if ((num_seg > 0) || (num_nr_seg > 0)) { /* * CMT: SFR algo (and HTNA) - this_sack_highest_newack has * to be greater than the cumack. Also reset saw_newack to 0 * for all dests. */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { net->saw_newack = 0; net->this_sack_highest_newack = last_tsn; } /* * thisSackHighestGap will increase while handling NEW * segments this_sack_highest_newack will increase while * handling NEWLY ACKED chunks. this_sack_lowest_newack is * used for CMT DAC algo. saw_newack will also change. */ if (sctp_handle_segments(m, &offset_seg, stcb, asoc, last_tsn, &biggest_tsn_acked, &biggest_tsn_newly_acked, &this_sack_lowest_newack, num_seg, num_nr_seg, &ecn_seg_sums)) { wake_him++; } if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) { /* * validate the biggest_tsn_acked in the gap acks if * strict adherence is wanted. */ if (SCTP_TSN_GE(biggest_tsn_acked, send_s)) { /* * peer is either confused or we are under * attack. We must abort. */ printf("Hopeless peer! biggest_tsn_acked:%x largest seq:%x\n", biggest_tsn_acked, send_s); goto hopeless_peer; } } } /*******************************************/ /* cancel ALL T3-send timer if accum moved */ /*******************************************/ if (asoc->sctp_cmt_on_off > 0) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (net->new_pseudo_cumack) sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_27); } } else { if (accum_moved) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_28); } } } /********************************************/ /* drop the acked chunks from the sentqueue */ /********************************************/ asoc->last_acked_seq = cum_ack; TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, cum_ack)) { break; } if (tp1->sent == SCTP_DATAGRAM_UNSENT) { /* no more sent on list */ printf("Warning, tp1->sent == %d and its now acked?\n", tp1->sent); } TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next); if (tp1->pr_sctp_on) { if (asoc->pr_sctp_cnt != 0) asoc->pr_sctp_cnt--; } asoc->sent_queue_cnt--; if (tp1->data) { /* sa_ignore NO_NULL_CHK */ sctp_free_bufspace(stcb, asoc, tp1, 1); sctp_m_freem(tp1->data); tp1->data = NULL; if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(tp1->flags)) { asoc->sent_queue_cnt_removeable--; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { sctp_log_sack(asoc->last_acked_seq, cum_ack, tp1->rec.data.TSN_seq, 0, 0, SCTP_LOG_FREE_SENT); } sctp_free_a_chunk(stcb, tp1); wake_him++; } if (TAILQ_EMPTY(&asoc->sent_queue) && (asoc->total_flight > 0)) { #ifdef INVARIANTS panic("Warning flight size is postive and should be 0"); #else SCTP_PRINTF("Warning flight size incorrect should be 0 is %d\n", asoc->total_flight); #endif asoc->total_flight = 0; } /* sa_ignore NO_NULL_CHK */ if ((wake_him) && (stcb->sctp_socket)) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif SOCKBUF_LOCK(&stcb->sctp_socket->so_snd); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { sctp_wakeup_log(stcb, cum_ack, wake_him, SCTP_WAKESND_FROM_SACK); } #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return; } #endif sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { sctp_wakeup_log(stcb, cum_ack, wake_him, SCTP_NOWAKE_FROM_SACK); } } if (asoc->fast_retran_loss_recovery && accum_moved) { if (SCTP_TSN_GE(asoc->last_acked_seq, asoc->fast_recovery_tsn)) { /* Setup so we will exit RFC2582 fast recovery */ will_exit_fast_recovery = 1; } } /* * Check for revoked fragments: * * if Previous sack - Had no frags then we can't have any revoked if * Previous sack - Had frag's then - If we now have frags aka * num_seg > 0 call sctp_check_for_revoked() to tell if peer revoked * some of them. else - The peer revoked all ACKED fragments, since * we had some before and now we have NONE. */ if (num_seg) { sctp_check_for_revoked(stcb, asoc, cum_ack, biggest_tsn_acked); asoc->saw_sack_with_frags = 1; } else if (asoc->saw_sack_with_frags) { int cnt_revoked = 0; /* Peer revoked all dg's marked or acked */ TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (tp1->sent == SCTP_DATAGRAM_ACKED) { tp1->sent = SCTP_DATAGRAM_SENT; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) tp1->whoTo, tp1->rec.data.TSN_seq); } sctp_flight_size_increase(tp1); sctp_total_flight_increase(stcb, tp1); tp1->rec.data.chunk_was_revoked = 1; /* * To ensure that this increase in * flightsize, which is artificial, does not * throttle the sender, we also increase the * cwnd artificially. */ tp1->whoTo->cwnd += tp1->book_size; cnt_revoked++; } } if (cnt_revoked) { reneged_all = 1; } asoc->saw_sack_with_frags = 0; } if (num_nr_seg > 0) asoc->saw_sack_with_nr_frags = 1; else asoc->saw_sack_with_nr_frags = 0; /* JRS - Use the congestion control given in the CC module */ asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, accum_moved, reneged_all, will_exit_fast_recovery); if (TAILQ_EMPTY(&asoc->sent_queue)) { /* nothing left in-flight */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { /* stop all timers */ if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck4); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_29); } } sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_30); net->flight_size = 0; net->partial_bytes_acked = 0; } asoc->total_flight = 0; asoc->total_flight_count = 0; } /**********************************/ /* Now what about shutdown issues */ /**********************************/ if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) { /* nothing left on sendqueue.. consider done */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, asoc->peers_rwnd, 0, 0, a_rwnd); } asoc->peers_rwnd = a_rwnd; if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } /* clean up */ if ((asoc->stream_queue_cnt == 1) && ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED)) && (asoc->locked_on_sending) ) { struct sctp_stream_queue_pending *sp; /* * I may be in a state where we got all across.. but * cannot write more due to a shutdown... we abort * since the user did not indicate EOR in this case. */ sp = TAILQ_LAST(&((asoc->locked_on_sending)->outqueue), sctp_streamhead); if ((sp) && (sp->length == 0)) { asoc->locked_on_sending = NULL; if (sp->msg_is_complete) { asoc->stream_queue_cnt--; } else { asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; asoc->stream_queue_cnt--; } } } if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) && (asoc->stream_queue_cnt == 0)) { if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) { /* Need to abort here */ struct mbuf *oper; abort_out_now: *abort_now = 1; /* XXX */ oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_31); } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_31; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, oper, SCTP_SO_NOT_LOCKED); return; } else { if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(stcb); sctp_send_shutdown(stcb, stcb->asoc.primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } return; } else if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) && (asoc->stream_queue_cnt == 0)) { if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) { goto abort_out_now; } SCTP_STAT_DECR_GAUGE32(sctps_currestab); SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_ACK_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); sctp_stop_timers_for_shutdown(stcb); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, stcb->sctp_ep, stcb, asoc->primary_destination); return; } } /* * Now here we are going to recycle net_ack for a different use... * HEADS UP. */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { net->net_ack = 0; } /* * CMT DAC algorithm: If SACK DAC flag was 0, then no extra marking * to be done. Setting this_sack_lowest_newack to the cum_ack will * automatically ensure that. */ if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac) && (cmt_dac_flag == 0)) { this_sack_lowest_newack = cum_ack; } if ((num_seg > 0) || (num_nr_seg > 0)) { sctp_strike_gap_ack_chunks(stcb, asoc, biggest_tsn_acked, biggest_tsn_newly_acked, this_sack_lowest_newack, accum_moved); } /* JRS - Use the congestion control given in the CC module */ asoc->cc_functions.sctp_cwnd_update_after_fr(stcb, asoc); - /****************************************************************** - * Here we do the stuff with ECN Nonce checking. - * We basically check to see if the nonce sum flag was incorrect - * or if resynchronization needs to be done. Also if we catch a - * misbehaving receiver we give him the kick. - ******************************************************************/ - - if (asoc->ecn_nonce_allowed) { - if (asoc->nonce_sum_check) { - if (nonce_sum_flag != ((asoc->nonce_sum_expect_base + ecn_seg_sums) & SCTP_SACK_NONCE_SUM)) { - if (asoc->nonce_wait_for_ecne == 0) { - struct sctp_tmit_chunk *lchk; - - lchk = TAILQ_FIRST(&asoc->send_queue); - asoc->nonce_wait_for_ecne = 1; - if (lchk) { - asoc->nonce_wait_tsn = lchk->rec.data.TSN_seq; - } else { - asoc->nonce_wait_tsn = asoc->sending_seq; - } - } else { - if (SCTP_TSN_GE(asoc->last_acked_seq, asoc->nonce_wait_tsn)) { - /* - * Misbehaving peer. We need - * to react to this guy - */ - asoc->ecn_allowed = 0; - asoc->ecn_nonce_allowed = 0; - } - } - } - } else { - /* See if Resynchronization Possible */ - if (SCTP_TSN_GT(asoc->last_acked_seq, asoc->nonce_resync_tsn)) { - asoc->nonce_sum_check = 1; - /* - * now we must calculate what the base is. - * We do this based on two things, we know - * the total's for all the segments - * gap-acked in the SACK, its stored in - * ecn_seg_sums. We also know the SACK's - * nonce sum, its in nonce_sum_flag. So we - * can build a truth table to back-calculate - * the new value of - * asoc->nonce_sum_expect_base: - * - * SACK-flag-Value Seg-Sums Base 0 0 0 - * 1 0 1 0 1 1 1 - * 1 0 - */ - asoc->nonce_sum_expect_base = (ecn_seg_sums ^ nonce_sum_flag) & SCTP_SACK_NONCE_SUM; - } - } - } /* Now are we exiting loss recovery ? */ if (will_exit_fast_recovery) { /* Ok, we must exit fast recovery */ asoc->fast_retran_loss_recovery = 0; } if ((asoc->sat_t3_loss_recovery) && SCTP_TSN_GE(asoc->last_acked_seq, asoc->sat_t3_recovery_tsn)) { /* end satellite t3 loss recovery */ asoc->sat_t3_loss_recovery = 0; } /* * CMT Fast recovery */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (net->will_exit_fast_recovery) { /* Ok, we must exit fast recovery */ net->fast_retran_loss_recovery = 0; } } /* Adjust and set the new rwnd value */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, asoc->peers_rwnd, asoc->total_flight, (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)), a_rwnd); } asoc->peers_rwnd = sctp_sbspace_sub(a_rwnd, (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } if (asoc->peers_rwnd > old_rwnd) { win_probe_recovery = 1; } /* * Now we must setup so we have a timer up for anyone with * outstanding data. */ done_once = 0; again: j = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (win_probe_recovery && (net->window_probe)) { win_probe_recovered = 1; /*- * Find first chunk that was used with * window probe and clear the event. Put * it back into the send queue as if has * not been sent. */ TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (tp1->window_probe) { sctp_window_probe_recovery(stcb, asoc, net, tp1); break; } } } if (net->flight_size) { j++; if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); } if (net->window_probe) { net->window_probe = 0; } } else { if (net->window_probe) { /* * In window probes we must assure a timer * is still running there */ if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); } } else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_22); } if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpidsck4); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_23); } } } } if ((j == 0) && (!TAILQ_EMPTY(&asoc->sent_queue)) && (asoc->sent_queue_retran_cnt == 0) && (win_probe_recovered == 0) && (done_once == 0)) { /* * huh, this should not happen unless all packets are * PR-SCTP and marked to skip of course. */ if (sctp_fs_audit(asoc)) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { net->flight_size = 0; } asoc->total_flight = 0; asoc->total_flight_count = 0; asoc->sent_queue_retran_cnt = 0; TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { if (tp1->sent < SCTP_DATAGRAM_RESEND) { sctp_flight_size_increase(tp1); sctp_total_flight_increase(stcb, tp1); } else if (tp1->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(asoc->sent_queue_retran_cnt); } } } done_once = 1; goto again; } /*********************************************/ /* Here we perform PR-SCTP procedures */ /* (section 4.2) */ /*********************************************/ /* C1. update advancedPeerAckPoint */ if (SCTP_TSN_GT(cum_ack, asoc->advanced_peer_ack_point)) { asoc->advanced_peer_ack_point = cum_ack; } /* C2. try to further move advancedPeerAckPoint ahead */ if ((asoc->peer_supports_prsctp) && (asoc->pr_sctp_cnt > 0)) { struct sctp_tmit_chunk *lchk; uint32_t old_adv_peer_ack_point; old_adv_peer_ack_point = asoc->advanced_peer_ack_point; lchk = sctp_try_advance_peer_ack_point(stcb, asoc); /* C3. See if we need to send a Fwd-TSN */ if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cum_ack)) { /* - * ISSUE with ECN, see FWD-TSN processing for notes - * on issues that will occur when the ECN NONCE - * stuff is put into SCTP for cross checking. + * ISSUE with ECN, see FWD-TSN processing. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { sctp_misc_ints(SCTP_FWD_TSN_CHECK, 0xee, cum_ack, asoc->advanced_peer_ack_point, old_adv_peer_ack_point); } if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) { - send_forward_tsn(stcb, asoc); - /* - * ECN Nonce: Disable Nonce Sum check when - * FWD TSN is sent and store resync tsn - */ - asoc->nonce_sum_check = 0; - asoc->nonce_resync_tsn = asoc->advanced_peer_ack_point; } else if (lchk) { /* try to FR fwd-tsn's that get lost too */ if (lchk->rec.data.fwd_tsn_cnt >= 3) { send_forward_tsn(stcb, asoc); } } } if (lchk) { /* Assure a timer is up */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo); } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SACK_RWND_UPDATE, a_rwnd, stcb->asoc.peers_rwnd, stcb->asoc.total_flight, stcb->asoc.total_output_queue_size); } } void sctp_update_acked(struct sctp_tcb *stcb, struct sctp_shutdown_chunk *cp, struct sctp_nets *netp, int *abort_flag) { /* Copy cum-ack */ uint32_t cum_ack, a_rwnd; cum_ack = ntohl(cp->cumulative_tsn_ack); /* Arrange so a_rwnd does NOT change */ a_rwnd = stcb->asoc.peers_rwnd + stcb->asoc.total_flight; /* Now call the express sack handling */ - sctp_express_handle_sack(stcb, cum_ack, a_rwnd, 0, abort_flag); + sctp_express_handle_sack(stcb, cum_ack, a_rwnd, abort_flag); } static void sctp_kick_prsctp_reorder_queue(struct sctp_tcb *stcb, struct sctp_stream_in *strmin) { struct sctp_queued_to_read *ctl, *nctl; struct sctp_association *asoc; uint16_t tt; asoc = &stcb->asoc; tt = strmin->last_sequence_delivered; /* * First deliver anything prior to and including the stream no that * came in */ TAILQ_FOREACH_SAFE(ctl, &strmin->inqueue, next, nctl) { if (SCTP_SSN_GE(tt, ctl->sinfo_ssn)) { /* this is deliverable now */ TAILQ_REMOVE(&strmin->inqueue, ctl, next); /* subtract pending on streams */ asoc->size_on_all_streams -= ctl->length; sctp_ucount_decr(asoc->cnt_on_all_streams); /* deliver it to at least the delivery-q */ if (stcb->sctp_socket) { sctp_mark_non_revokable(asoc, ctl->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, ctl, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_HELD, SCTP_SO_NOT_LOCKED); } } else { /* no more delivery now. */ break; } } /* * now we must deliver things in queue the normal way if any are * now ready. */ tt = strmin->last_sequence_delivered + 1; TAILQ_FOREACH_SAFE(ctl, &strmin->inqueue, next, nctl) { if (tt == ctl->sinfo_ssn) { /* this is deliverable now */ TAILQ_REMOVE(&strmin->inqueue, ctl, next); /* subtract pending on streams */ asoc->size_on_all_streams -= ctl->length; sctp_ucount_decr(asoc->cnt_on_all_streams); /* deliver it to at least the delivery-q */ strmin->last_sequence_delivered = ctl->sinfo_ssn; if (stcb->sctp_socket) { sctp_mark_non_revokable(asoc, ctl->sinfo_tsn); sctp_add_to_readq(stcb->sctp_ep, stcb, ctl, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_HELD, SCTP_SO_NOT_LOCKED); } tt = strmin->last_sequence_delivered + 1; } else { break; } } } static void sctp_flush_reassm_for_str_seq(struct sctp_tcb *stcb, struct sctp_association *asoc, uint16_t stream, uint16_t seq) { struct sctp_tmit_chunk *chk, *nchk; /* For each one on here see if we need to toss it */ /* * For now large messages held on the reasmqueue that are complete * will be tossed too. We could in theory do more work to spin * through and stop after dumping one msg aka seeing the start of a * new msg at the head, and call the delivery function... to see if * it can be delivered... But for now we just dump everything on the * queue. */ TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { /* * Do not toss it if on a different stream or marked for * unordered delivery in which case the stream sequence * number has no meaning. */ if ((chk->rec.data.stream_number != stream) || ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == SCTP_DATA_UNORDERED)) { continue; } if (chk->rec.data.stream_seq == seq) { /* It needs to be tossed */ TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (SCTP_TSN_GT(chk->rec.data.TSN_seq, asoc->tsn_last_delivered)) { asoc->tsn_last_delivered = chk->rec.data.TSN_seq; asoc->str_of_pdapi = chk->rec.data.stream_number; asoc->ssn_of_pdapi = chk->rec.data.stream_seq; asoc->fragment_flags = chk->rec.data.rcv_flags; } asoc->size_on_reasm_queue -= chk->send_size; sctp_ucount_decr(asoc->cnt_on_reasm_queue); /* Clear up any stream problem */ if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) != SCTP_DATA_UNORDERED && SCTP_SSN_GT(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered)) { /* * We must dump forward this streams * sequence number if the chunk is not * unordered that is being skipped. There is * a chance that if the peer does not * include the last fragment in its FWD-TSN * we WILL have a problem here since you * would have a partial chunk in queue that * may not be deliverable. Also if a Partial * delivery API as started the user may get * a partial chunk. The next read returning * a new chunk... really ugly but I see no * way around it! Maybe a notify?? */ asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered = chk->rec.data.stream_seq; } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); } else if (SCTP_SSN_GT(chk->rec.data.stream_seq, seq)) { /* * If the stream_seq is > than the purging one, we * are done */ break; } } } void sctp_handle_forward_tsn(struct sctp_tcb *stcb, struct sctp_forward_tsn_chunk *fwd, int *abort_flag, struct mbuf *m, int offset) { - /* - * ISSUES that MUST be fixed for ECN! When we are the sender of the - * forward TSN, when the SACK comes back that acknowledges the - * FWD-TSN we must reset the NONCE sum to match correctly. This will - * get quite tricky since we may have sent more data interveneing - * and must carefully account for what the SACK says on the nonce - * and any gaps that are reported. This work will NOT be done here, - * but I note it here since it is really related to PR-SCTP and - * FWD-TSN's - */ - /* The pr-sctp fwd tsn */ /* * here we will perform all the data receiver side steps for * processing FwdTSN, as required in by pr-sctp draft: * * Assume we get FwdTSN(x): * * 1) update local cumTSN to x 2) try to further advance cumTSN to x + * others we have 3) examine and update re-ordering queue on * pr-in-streams 4) clean up re-assembly queue 5) Send a sack to * report where we are. */ struct sctp_association *asoc; uint32_t new_cum_tsn, gap; unsigned int i, fwd_sz, cumack_set_flag, m_size; uint32_t str_seq; struct sctp_stream_in *strm; struct sctp_tmit_chunk *chk, *nchk; struct sctp_queued_to_read *ctl, *sv; cumack_set_flag = 0; asoc = &stcb->asoc; if ((fwd_sz = ntohs(fwd->ch.chunk_length)) < sizeof(struct sctp_forward_tsn_chunk)) { SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size too small/big fwd-tsn\n"); return; } m_size = (stcb->asoc.mapping_array_size << 3); /*************************************************************/ /* 1. Here we update local cumTSN and shift the bitmap array */ /*************************************************************/ new_cum_tsn = ntohl(fwd->new_cumulative_tsn); if (SCTP_TSN_GE(asoc->cumulative_tsn, new_cum_tsn)) { /* Already got there ... */ return; } /* * now we know the new TSN is more advanced, let's find the actual * gap */ SCTP_CALC_TSN_TO_GAP(gap, new_cum_tsn, asoc->mapping_array_base_tsn); asoc->cumulative_tsn = new_cum_tsn; if (gap >= m_size) { if ((long)gap > sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv)) { struct mbuf *oper; /* * out of range (of single byte chunks in the rwnd I * give out). This must be an attacker. */ *abort_flag = 1; oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + (sizeof(uint32_t) * 3); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_33); ippp++; *ippp = asoc->highest_tsn_inside_map; ippp++; *ippp = new_cum_tsn; } stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_33; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_PEER_FAULTY, oper, SCTP_SO_NOT_LOCKED); return; } SCTP_STAT_INCR(sctps_fwdtsn_map_over); memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size); asoc->mapping_array_base_tsn = new_cum_tsn + 1; asoc->highest_tsn_inside_map = new_cum_tsn; memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size); asoc->highest_tsn_inside_nr_map = new_cum_tsn; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 3, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } - asoc->last_echo_tsn = asoc->highest_tsn_inside_map; } else { SCTP_TCB_LOCK_ASSERT(stcb); for (i = 0; i <= gap; i++) { if (!SCTP_IS_TSN_PRESENT(asoc->mapping_array, i) && !SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, i)) { SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, i); if (SCTP_TSN_GT(asoc->mapping_array_base_tsn + i, asoc->highest_tsn_inside_nr_map)) { asoc->highest_tsn_inside_nr_map = asoc->mapping_array_base_tsn + i; } } } } /*************************************************************/ /* 2. Clear up re-assembly queue */ /*************************************************************/ /* * First service it if pd-api is up, just in case we can progress it * forward */ if (asoc->fragmented_delivery_inprogress) { sctp_service_reassembly(stcb, asoc); } /* For each one on here see if we need to toss it */ /* * For now large messages held on the reasmqueue that are complete * will be tossed too. We could in theory do more work to spin * through and stop after dumping one msg aka seeing the start of a * new msg at the head, and call the delivery function... to see if * it can be delivered... But for now we just dump everything on the * queue. */ TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { if (SCTP_TSN_GE(new_cum_tsn, chk->rec.data.TSN_seq)) { /* It needs to be tossed */ TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (SCTP_TSN_GT(chk->rec.data.TSN_seq, asoc->tsn_last_delivered)) { asoc->tsn_last_delivered = chk->rec.data.TSN_seq; asoc->str_of_pdapi = chk->rec.data.stream_number; asoc->ssn_of_pdapi = chk->rec.data.stream_seq; asoc->fragment_flags = chk->rec.data.rcv_flags; } asoc->size_on_reasm_queue -= chk->send_size; sctp_ucount_decr(asoc->cnt_on_reasm_queue); /* Clear up any stream problem */ if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) != SCTP_DATA_UNORDERED && SCTP_SSN_GT(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered)) { /* * We must dump forward this streams * sequence number if the chunk is not * unordered that is being skipped. There is * a chance that if the peer does not * include the last fragment in its FWD-TSN * we WILL have a problem here since you * would have a partial chunk in queue that * may not be deliverable. Also if a Partial * delivery API as started the user may get * a partial chunk. The next read returning * a new chunk... really ugly but I see no * way around it! Maybe a notify?? */ asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered = chk->rec.data.stream_seq; } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); } else { /* * Ok we have gone beyond the end of the fwd-tsn's * mark. */ break; } } /*******************************************************/ /* 3. Update the PR-stream re-ordering queues and fix */ /* delivery issues as needed. */ /*******************************************************/ fwd_sz -= sizeof(*fwd); if (m && fwd_sz) { /* New method. */ unsigned int num_str; struct sctp_strseq *stseq, strseqbuf; offset += sizeof(*fwd); SCTP_INP_READ_LOCK(stcb->sctp_ep); num_str = fwd_sz / sizeof(struct sctp_strseq); for (i = 0; i < num_str; i++) { uint16_t st; stseq = (struct sctp_strseq *)sctp_m_getptr(m, offset, sizeof(struct sctp_strseq), (uint8_t *) & strseqbuf); offset += sizeof(struct sctp_strseq); if (stseq == NULL) { break; } /* Convert */ st = ntohs(stseq->stream); stseq->stream = st; st = ntohs(stseq->sequence); stseq->sequence = st; /* now process */ /* * Ok we now look for the stream/seq on the read * queue where its not all delivered. If we find it * we transmute the read entry into a PDI_ABORTED. */ if (stseq->stream >= asoc->streamincnt) { /* screwed up streams, stop! */ break; } if ((asoc->str_of_pdapi == stseq->stream) && (asoc->ssn_of_pdapi == stseq->sequence)) { /* * If this is the one we were partially * delivering now then we no longer are. * Note this will change with the reassembly * re-write. */ asoc->fragmented_delivery_inprogress = 0; } sctp_flush_reassm_for_str_seq(stcb, asoc, stseq->stream, stseq->sequence); TAILQ_FOREACH(ctl, &stcb->sctp_ep->read_queue, next) { if ((ctl->sinfo_stream == stseq->stream) && (ctl->sinfo_ssn == stseq->sequence)) { str_seq = (stseq->stream << 16) | stseq->sequence; ctl->end_added = 1; ctl->pdapi_aborted = 1; sv = stcb->asoc.control_pdapi; stcb->asoc.control_pdapi = ctl; sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, stcb, SCTP_PARTIAL_DELIVERY_ABORTED, (void *)&str_seq, SCTP_SO_NOT_LOCKED); stcb->asoc.control_pdapi = sv; break; } else if ((ctl->sinfo_stream == stseq->stream) && SCTP_SSN_GT(ctl->sinfo_ssn, stseq->sequence)) { /* We are past our victim SSN */ break; } } strm = &asoc->strmin[stseq->stream]; if (SCTP_SSN_GT(stseq->sequence, strm->last_sequence_delivered)) { /* Update the sequence number */ strm->last_sequence_delivered = stseq->sequence; } /* now kick the stream the new way */ /* sa_ignore NO_NULL_CHK */ sctp_kick_prsctp_reorder_queue(stcb, strm); } SCTP_INP_READ_UNLOCK(stcb->sctp_ep); } /* * Now slide thing forward. */ sctp_slide_mapping_arrays(stcb); if (!TAILQ_EMPTY(&asoc->reasmqueue)) { /* now lets kick out and check for more fragmented delivery */ /* sa_ignore NO_NULL_CHK */ sctp_deliver_reasm_check(stcb, &stcb->asoc); } } diff --git a/sys/netinet/sctp_indata.h b/sys/netinet/sctp_indata.h index 79978a5c9921..7f62ccaca304 100644 --- a/sys/netinet/sctp_indata.h +++ b/sys/netinet/sctp_indata.h @@ -1,129 +1,129 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_indata.h,v 1.9 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_indata_h__ #define __sctp_indata_h__ #if defined(_KERNEL) || defined(__Userspace__) struct sctp_queued_to_read * sctp_build_readq_entry(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t tsn, uint32_t ppid, uint32_t context, uint16_t stream_no, uint16_t stream_seq, uint8_t flags, struct mbuf *dm); #define sctp_build_readq_entry_mac(_ctl, in_it, a, net, tsn, ppid, context, stream_no, stream_seq, flags, dm) do { \ if (_ctl) { \ atomic_add_int(&((net)->ref_count), 1); \ (_ctl)->sinfo_stream = stream_no; \ (_ctl)->sinfo_ssn = stream_seq; \ (_ctl)->sinfo_flags = (flags << 8); \ (_ctl)->sinfo_ppid = ppid; \ (_ctl)->sinfo_context = a; \ (_ctl)->sinfo_timetolive = 0; \ (_ctl)->sinfo_tsn = tsn; \ (_ctl)->sinfo_cumtsn = tsn; \ (_ctl)->sinfo_assoc_id = sctp_get_associd((in_it)); \ (_ctl)->length = 0; \ (_ctl)->held_length = 0; \ (_ctl)->whoFrom = net; \ (_ctl)->data = dm; \ (_ctl)->tail_mbuf = NULL; \ (_ctl)->aux_data = NULL; \ (_ctl)->stcb = (in_it); \ (_ctl)->port_from = (in_it)->rport; \ (_ctl)->spec_flags = 0; \ (_ctl)->do_not_ref_stcb = 0; \ (_ctl)->end_added = 0; \ (_ctl)->pdapi_aborted = 0; \ (_ctl)->some_taken = 0; \ } \ } while (0) struct mbuf * sctp_build_ctl_nchunk(struct sctp_inpcb *inp, struct sctp_sndrcvinfo *sinfo); char * sctp_build_ctl_cchunk(struct sctp_inpcb *inp, int *control_len, struct sctp_sndrcvinfo *sinfo); void sctp_set_rwnd(struct sctp_tcb *, struct sctp_association *); uint32_t sctp_calc_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc); void sctp_express_handle_sack(struct sctp_tcb *stcb, uint32_t cumack, - uint32_t rwnd, int nonce_sum_flag, int *abort_now); + uint32_t rwnd, int *abort_now); void sctp_handle_sack(struct mbuf *m, int offset_seg, int offset_dup, struct sctp_tcb *stcb, struct sctp_nets *net_from, uint16_t num_seg, uint16_t num_nr_seg, uint16_t num_dup, int *abort_now, uint8_t flags, uint32_t cum_ack, uint32_t rwnd); /* draft-ietf-tsvwg-usctp */ void sctp_handle_forward_tsn(struct sctp_tcb *, struct sctp_forward_tsn_chunk *, int *, struct mbuf *, int); struct sctp_tmit_chunk * sctp_try_advance_peer_ack_point(struct sctp_tcb *, struct sctp_association *); void sctp_service_queues(struct sctp_tcb *, struct sctp_association *); void sctp_update_acked(struct sctp_tcb *, struct sctp_shutdown_chunk *, struct sctp_nets *, int *); int sctp_process_data(struct mbuf **, int, int *, int, struct sctphdr *, struct sctp_inpcb *, struct sctp_tcb *, struct sctp_nets *, uint32_t *); void sctp_slide_mapping_arrays(struct sctp_tcb *stcb); void sctp_sack_check(struct sctp_tcb *, int, int *); #endif #endif diff --git a/sys/netinet/sctp_input.c b/sys/netinet/sctp_input.c index 664ff62d0de1..6fc8e5b7798b 100644 --- a/sys/netinet/sctp_input.c +++ b/sys/netinet/sctp_input.c @@ -1,5996 +1,5921 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_input.c,v 1.27 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static void sctp_stop_all_cookie_timers(struct sctp_tcb *stcb) { struct sctp_nets *net; /* * This now not only stops all cookie timers it also stops any INIT * timers as well. This will make sure that the timers are stopped * in all collision cases. */ SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->rxt_timer.type == SCTP_TIMER_TYPE_COOKIE) { sctp_timer_stop(SCTP_TIMER_TYPE_COOKIE, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_1); } else if (net->rxt_timer.type == SCTP_TIMER_TYPE_INIT) { sctp_timer_stop(SCTP_TIMER_TYPE_INIT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_2); } } } /* INIT handler */ static void sctp_handle_init(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_init_chunk *cp, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, int *abort_no_unlock, uint32_t vrf_id, uint16_t port) { struct sctp_init *init; struct mbuf *op_err; uint32_t init_limit; SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_init: handling INIT tcb:%p\n", stcb); if (stcb == NULL) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { goto outnow; } } op_err = NULL; init = &cp->init; /* First are we accepting? */ if ((inp->sctp_socket->so_qlimit == 0) && (stcb == NULL)) { SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_init: Abort, so_qlimit:%d\n", inp->sctp_socket->so_qlimit); /* * FIX ME ?? What about TCP model and we have a * match/restart case? Actually no fix is needed. the lookup * will always find the existing assoc so stcb would not be * NULL. It may be questionable to do this since we COULD * just send back the INIT-ACK and hope that the app did * accept()'s by the time the COOKIE was sent. But there is * a price to pay for COOKIE generation and I don't want to * pay it on the chance that the app will actually do some * accepts(). The App just looses and should NOT be in this * state :-) */ sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } if (ntohs(cp->ch.chunk_length) < sizeof(struct sctp_init_chunk)) { /* Invalid length */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } /* validate parameters */ if (init->initiate_tag == 0) { /* protocol error... send abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } if (ntohl(init->a_rwnd) < SCTP_MIN_RWND) { /* invalid parameter... send abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } if (init->num_inbound_streams == 0) { /* protocol error... send abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } if (init->num_outbound_streams == 0) { /* protocol error... send abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(inp, stcb, m, iphlen, sh, op_err, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } init_limit = offset + ntohs(cp->ch.chunk_length); if (sctp_validate_init_auth_params(m, offset + sizeof(*cp), init_limit)) { /* auth parameter(s) error... send abort */ sctp_abort_association(inp, stcb, m, iphlen, sh, NULL, vrf_id, port); if (stcb) *abort_no_unlock = 1; goto outnow; } /* send an INIT-ACK w/cookie */ SCTPDBG(SCTP_DEBUG_INPUT3, "sctp_handle_init: sending INIT-ACK\n"); sctp_send_initiate_ack(inp, stcb, m, iphlen, offset, sh, cp, vrf_id, port, ((stcb == NULL) ? SCTP_HOLDS_LOCK : SCTP_NOT_LOCKED)); outnow: if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); } } /* * process peer "INIT/INIT-ACK" chunk returns value < 0 on error */ int sctp_is_there_unsent_data(struct sctp_tcb *stcb) { int unsent_data = 0; unsigned int i; struct sctp_stream_queue_pending *sp; struct sctp_association *asoc; /* * This function returns the number of streams that have true unsent * data on them. Note that as it looks through it will clean up any * places that have old data that has been sent but left at top of * stream queue. */ asoc = &stcb->asoc; SCTP_TCB_SEND_LOCK(stcb); if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { /* Check to see if some data queued */ for (i = 0; i < stcb->asoc.streamoutcnt; i++) { /* sa_ignore FREED_MEMORY */ sp = TAILQ_FIRST(&stcb->asoc.strmout[i].outqueue); if (sp == NULL) { continue; } if ((sp->msg_is_complete) && (sp->length == 0) && (sp->sender_all_done)) { /* * We are doing differed cleanup. Last time * through when we took all the data the * sender_all_done was not set. */ if (sp->put_last_out == 0) { SCTP_PRINTF("Gak, put out entire msg with NO end!-1\n"); SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d\n", sp->sender_all_done, sp->length, sp->msg_is_complete, sp->put_last_out); } atomic_subtract_int(&stcb->asoc.stream_queue_cnt, 1); TAILQ_REMOVE(&stcb->asoc.strmout[i].outqueue, sp, next); if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } sctp_free_a_strmoq(stcb, sp); } else { unsent_data++; break; } } } SCTP_TCB_SEND_UNLOCK(stcb); return (unsent_data); } static int sctp_process_init(struct sctp_init_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_init *init; struct sctp_association *asoc; struct sctp_nets *lnet; unsigned int i; init = &cp->init; asoc = &stcb->asoc; /* save off parameters */ asoc->peer_vtag = ntohl(init->initiate_tag); asoc->peers_rwnd = ntohl(init->a_rwnd); + /* init tsn's */ + asoc->highest_tsn_inside_map = asoc->asconf_seq_in = ntohl(init->initial_tsn) - 1; + if (!TAILQ_EMPTY(&asoc->nets)) { /* update any ssthresh's that may have a default */ TAILQ_FOREACH(lnet, &asoc->nets, sctp_next) { lnet->ssthresh = asoc->peers_rwnd; - if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) { sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_INITIALIZATION); } } } SCTP_TCB_SEND_LOCK(stcb); if (asoc->pre_open_streams > ntohs(init->num_inbound_streams)) { unsigned int newcnt; struct sctp_stream_out *outs; struct sctp_stream_queue_pending *sp, *nsp; struct sctp_tmit_chunk *chk, *nchk; /* abandon the upper streams */ newcnt = ntohs(init->num_inbound_streams); TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { if (chk->rec.data.stream_number >= newcnt) { TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); asoc->send_queue_cnt--; if (chk->data != NULL) { sctp_free_bufspace(stcb, asoc, chk, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, chk, SCTP_SO_NOT_LOCKED); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ } } if (asoc->strmout) { for (i = newcnt; i < asoc->pre_open_streams; i++) { outs = &asoc->strmout[i]; TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { TAILQ_REMOVE(&outs->outqueue, sp, next); asoc->stream_queue_cnt--; sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, sp, SCTP_SO_NOT_LOCKED); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } /* Free the chunk */ sctp_free_a_strmoq(stcb, sp); /* sa_ignore FREED_MEMORY */ } } } /* cut back the count */ asoc->pre_open_streams = newcnt; } SCTP_TCB_SEND_UNLOCK(stcb); asoc->strm_realoutsize = asoc->streamoutcnt = asoc->pre_open_streams; - /* init tsn's */ - asoc->highest_tsn_inside_map = asoc->asconf_seq_in = ntohl(init->initial_tsn) - 1; + /* EY - nr_sack: initialize highest tsn in nr_mapping_array */ asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 5, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } /* This is the next one we expect */ asoc->str_reset_seq_in = asoc->asconf_seq_in + 1; asoc->mapping_array_base_tsn = ntohl(init->initial_tsn); asoc->tsn_last_delivered = asoc->cumulative_tsn = asoc->asconf_seq_in; - asoc->last_echo_tsn = asoc->asconf_seq_in; + asoc->advanced_peer_ack_point = asoc->last_acked_seq; /* open the requested streams */ if (asoc->strmin != NULL) { /* Free the old ones */ struct sctp_queued_to_read *ctl, *nctl; for (i = 0; i < asoc->streamincnt; i++) { TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) { TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next); sctp_free_remote_addr(ctl->whoFrom); ctl->whoFrom = NULL; sctp_m_freem(ctl->data); ctl->data = NULL; sctp_free_a_readq(stcb, ctl); } } SCTP_FREE(asoc->strmin, SCTP_M_STRMI); } asoc->streamincnt = ntohs(init->num_outbound_streams); if (asoc->streamincnt > MAX_SCTP_STREAMS) { asoc->streamincnt = MAX_SCTP_STREAMS; } SCTP_MALLOC(asoc->strmin, struct sctp_stream_in *, asoc->streamincnt * sizeof(struct sctp_stream_in), SCTP_M_STRMI); if (asoc->strmin == NULL) { /* we didn't get memory for the streams! */ SCTPDBG(SCTP_DEBUG_INPUT2, "process_init: couldn't get memory for the streams!\n"); return (-1); } for (i = 0; i < asoc->streamincnt; i++) { asoc->strmin[i].stream_no = i; asoc->strmin[i].last_sequence_delivered = 0xffff; /* * U-stream ranges will be set when the cookie is unpacked. * Or for the INIT sender they are un set (if pr-sctp not * supported) when the INIT-ACK arrives. */ TAILQ_INIT(&asoc->strmin[i].inqueue); asoc->strmin[i].delivery_started = 0; } /* * load_address_from_init will put the addresses into the * association when the COOKIE is processed or the INIT-ACK is * processed. Both types of COOKIE's existing and new call this * routine. It will remove addresses that are no longer in the * association (for the restarting case where addresses are * removed). Up front when the INIT arrives we will discard it if it * is a restart and new addresses have been added. */ /* sa_ignore MEMLEAK */ return (0); } /* * INIT-ACK message processing/consumption returns value < 0 on error */ static int sctp_process_init_ack(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_init_ack_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net, int *abort_no_unlock, uint32_t vrf_id) { struct sctp_association *asoc; struct mbuf *op_err; int retval, abort_flag; uint32_t initack_limit; int nat_friendly = 0; /* First verify that we have no illegal param's */ abort_flag = 0; op_err = NULL; op_err = sctp_arethere_unrecognized_parameters(m, (offset + sizeof(struct sctp_init_chunk)), &abort_flag, (struct sctp_chunkhdr *)cp, &nat_friendly); if (abort_flag) { /* Send an abort and notify peer */ sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_CAUSE_PROTOCOL_VIOLATION, op_err, SCTP_SO_NOT_LOCKED); *abort_no_unlock = 1; return (-1); } asoc = &stcb->asoc; asoc->peer_supports_nat = (uint8_t) nat_friendly; /* process the peer's parameters in the INIT-ACK */ retval = sctp_process_init((struct sctp_init_chunk *)cp, stcb, net); if (retval < 0) { return (retval); } initack_limit = offset + ntohs(cp->ch.chunk_length); /* load all addresses */ if ((retval = sctp_load_addresses_from_init(stcb, m, iphlen, (offset + sizeof(struct sctp_init_chunk)), initack_limit, sh, NULL))) { /* Huh, we should abort */ SCTPDBG(SCTP_DEBUG_INPUT1, "Load addresses from INIT causes an abort %d\n", retval); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, NULL, 0, net->port); *abort_no_unlock = 1; return (-1); } /* if the peer doesn't support asconf, flush the asconf queue */ if (asoc->peer_supports_asconf == 0) { struct sctp_asconf_addr *param, *nparam; TAILQ_FOREACH_SAFE(param, &asoc->asconf_queue, next, nparam) { TAILQ_REMOVE(&asoc->asconf_queue, param, next); SCTP_FREE(param, SCTP_M_ASC_ADDR); } } stcb->asoc.peer_hmac_id = sctp_negotiate_hmacid(stcb->asoc.peer_hmacs, stcb->asoc.local_hmacs); if (op_err) { sctp_queue_op_err(stcb, op_err); /* queuing will steal away the mbuf chain to the out queue */ op_err = NULL; } /* extract the cookie and queue it to "echo" it back... */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; net->error_count = 0; /* * Cancel the INIT timer, We do this first before queueing the * cookie. We always cancel at the primary to assue that we are * canceling the timer started by the INIT which always goes to the * primary. */ sctp_timer_stop(SCTP_TIMER_TYPE_INIT, stcb->sctp_ep, stcb, asoc->primary_destination, SCTP_FROM_SCTP_INPUT + SCTP_LOC_4); /* calculate the RTO */ net->RTO = sctp_calculate_rto(stcb, asoc, net, &asoc->time_entered, sctp_align_safe_nocopy); retval = sctp_send_cookie_echo(m, offset, stcb, net); if (retval < 0) { /* * No cookie, we probably should send a op error. But in any * case if there is no cookie in the INIT-ACK, we can * abandon the peer, its broke. */ if (retval == -3) { /* We abort with an error of missing mandatory param */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_MISSING_PARAM); if (op_err) { /* * Expand beyond to include the mandatory * param cookie */ struct sctp_inv_mandatory_param *mp; SCTP_BUF_LEN(op_err) = sizeof(struct sctp_inv_mandatory_param); mp = mtod(op_err, struct sctp_inv_mandatory_param *); /* Subtract the reserved param */ mp->length = htons(sizeof(struct sctp_inv_mandatory_param) - 2); mp->num_param = htonl(1); mp->param = htons(SCTP_STATE_COOKIE); mp->resv = 0; } sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; } return (retval); } return (0); } static void sctp_handle_heartbeat_ack(struct sctp_heartbeat_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sockaddr_storage store; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; struct sctp_nets *r_net, *f_net; struct timeval tv; int req_prim = 0; if (ntohs(cp->ch.chunk_length) != sizeof(struct sctp_heartbeat_chunk)) { /* Invalid length */ return; } sin = (struct sockaddr_in *)&store; sin6 = (struct sockaddr_in6 *)&store; memset(&store, 0, sizeof(store)); if (cp->heartbeat.hb_info.addr_family == AF_INET && cp->heartbeat.hb_info.addr_len == sizeof(struct sockaddr_in)) { sin->sin_family = cp->heartbeat.hb_info.addr_family; sin->sin_len = cp->heartbeat.hb_info.addr_len; sin->sin_port = stcb->rport; memcpy(&sin->sin_addr, cp->heartbeat.hb_info.address, sizeof(sin->sin_addr)); } else if (cp->heartbeat.hb_info.addr_family == AF_INET6 && cp->heartbeat.hb_info.addr_len == sizeof(struct sockaddr_in6)) { sin6->sin6_family = cp->heartbeat.hb_info.addr_family; sin6->sin6_len = cp->heartbeat.hb_info.addr_len; sin6->sin6_port = stcb->rport; memcpy(&sin6->sin6_addr, cp->heartbeat.hb_info.address, sizeof(sin6->sin6_addr)); } else { return; } r_net = sctp_findnet(stcb, (struct sockaddr *)sin); if (r_net == NULL) { SCTPDBG(SCTP_DEBUG_INPUT1, "Huh? I can't find the address I sent it to, discard\n"); return; } if ((r_net && (r_net->dest_state & SCTP_ADDR_UNCONFIRMED)) && (r_net->heartbeat_random1 == cp->heartbeat.hb_info.random_value1) && (r_net->heartbeat_random2 == cp->heartbeat.hb_info.random_value2)) { /* * If the its a HB and it's random value is correct when can * confirm the destination. */ r_net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; if (r_net->dest_state & SCTP_ADDR_REQ_PRIMARY) { stcb->asoc.primary_destination = r_net; r_net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY; r_net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY; f_net = TAILQ_FIRST(&stcb->asoc.nets); if (f_net != r_net) { /* * first one on the list is NOT the primary * sctp_cmpaddr() is much more efficent if * the primary is the first on the list, * make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, r_net, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, r_net, sctp_next); } req_prim = 1; } sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, stcb, 0, (void *)r_net, SCTP_SO_NOT_LOCKED); } r_net->error_count = 0; r_net->hb_responded = 1; tv.tv_sec = cp->heartbeat.hb_info.time_value_1; tv.tv_usec = cp->heartbeat.hb_info.time_value_2; if (r_net->dest_state & SCTP_ADDR_NOT_REACHABLE) { r_net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE; r_net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, SCTP_HEARTBEAT_SUCCESS, (void *)r_net, SCTP_SO_NOT_LOCKED); /* now was it the primary? if so restore */ if (r_net->dest_state & SCTP_ADDR_WAS_PRIMARY) { (void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, r_net); } } /* * JRS 5/14/07 - If CMT PF is on and the destination is in PF state, * set the destination to active state and set the cwnd to one or * two MTU's based on whether PF1 or PF2 is being used. If a T3 * timer is running, for the destination, stop the timer because a * PF-heartbeat was received. */ if ((stcb->asoc.sctp_cmt_on_off > 0) && (stcb->asoc.sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_5); } net->dest_state &= ~SCTP_ADDR_PF; net->cwnd = net->mtu * stcb->asoc.sctp_cmt_pf; SCTPDBG(SCTP_DEBUG_INPUT1, "Destination %p moved from PF to reachable with cwnd %d.\n", net, net->cwnd); } /* Now lets do a RTO with this */ r_net->RTO = sctp_calculate_rto(stcb, &stcb->asoc, r_net, &tv, sctp_align_safe_nocopy); /* Mobility adaptation */ if (req_prim) { if ((sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE) || sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_FASTHANDOFF)) && sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_PRIM_DELETED)) { sctp_timer_stop(SCTP_TIMER_TYPE_PRIM_DELETED, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_TIMER + SCTP_LOC_7); if (sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_FASTHANDOFF)) { sctp_assoc_immediate_retrans(stcb, stcb->asoc.primary_destination); } if (sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { sctp_move_chunks_from_net(stcb, stcb->asoc.deleted_primary); } sctp_delete_prim_timer(stcb->sctp_ep, stcb, stcb->asoc.deleted_primary); } } } static int sctp_handle_nat_colliding_state(struct sctp_tcb *stcb) { /* * return 0 means we want you to proceed with the abort non-zero * means no abort processing */ struct sctpasochead *head; if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_WAIT) { /* generate a new vtag and send init */ LIST_REMOVE(stcb, sctp_asocs); stcb->asoc.my_vtag = sctp_select_a_tag(stcb->sctp_ep, stcb->sctp_ep->sctp_lport, stcb->rport, 1); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* * put it in the bucket in the vtag hash of assoc's for the * system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); return (1); } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) { /* * treat like a case where the cookie expired i.e.: - dump * current cookie. - generate a new vtag. - resend init. */ /* generate a new vtag and send init */ LIST_REMOVE(stcb, sctp_asocs); stcb->asoc.state &= ~SCTP_STATE_COOKIE_ECHOED; stcb->asoc.state |= SCTP_STATE_COOKIE_WAIT; sctp_stop_all_cookie_timers(stcb); sctp_toss_old_cookies(stcb, &stcb->asoc); stcb->asoc.my_vtag = sctp_select_a_tag(stcb->sctp_ep, stcb->sctp_ep->sctp_lport, stcb->rport, 1); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* * put it in the bucket in the vtag hash of assoc's for the * system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); return (1); } return (0); } static int sctp_handle_nat_missing_state(struct sctp_tcb *stcb, struct sctp_nets *net) { /* * return 0 means we want you to proceed with the abort non-zero * means no abort processing */ if (stcb->asoc.peer_supports_auth == 0) { SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_nat_missing_state: Peer does not support AUTH, cannot send an asconf\n"); return (0); } sctp_asconf_send_nat_state_update(stcb, net); return (1); } static void sctp_handle_abort(struct sctp_abort_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif uint16_t len; SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_abort: handling ABORT\n"); if (stcb == NULL) return; len = ntohs(cp->ch.chunk_length); if (len > sizeof(struct sctp_chunkhdr)) { /* * Need to check the cause codes for our two magic nat * aborts which don't kill the assoc necessarily. */ struct sctp_abort_chunk *cpnext; struct sctp_missing_nat_state *natc; uint16_t cause; cpnext = cp; cpnext++; natc = (struct sctp_missing_nat_state *)cpnext; cause = ntohs(natc->cause); if (cause == SCTP_CAUSE_NAT_COLLIDING_STATE) { SCTPDBG(SCTP_DEBUG_INPUT2, "Received Colliding state abort flags:%x\n", cp->ch.chunk_flags); if (sctp_handle_nat_colliding_state(stcb)) { return; } } else if (cause == SCTP_CAUSE_NAT_MISSING_STATE) { SCTPDBG(SCTP_DEBUG_INPUT2, "Received missing state abort flags:%x\n", cp->ch.chunk_flags); if (sctp_handle_nat_missing_state(stcb, net)) { return; } } } /* stop any receive timers */ sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_6); /* notify user of the abort and clean up... */ sctp_abort_notification(stcb, 0, SCTP_SO_NOT_LOCKED); /* free the tcb */ #if defined(SCTP_PANIC_ON_ABORT) printf("stcb:%p state:%d rport:%d net:%p\n", stcb, stcb->asoc.state, stcb->rport, net); if (!(stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { panic("Received an ABORT"); } else { printf("No panic its in state %x closed\n", stcb->asoc.state); } #endif SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } #ifdef SCTP_ASOCLOG_OF_TSNS sctp_print_out_track_log(stcb); #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif stcb->asoc.state |= SCTP_STATE_WAS_ABORTED; (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_6); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_abort: finished\n"); } static void sctp_handle_shutdown(struct sctp_shutdown_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net, int *abort_flag) { struct sctp_association *asoc; int some_on_streamwheel; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_shutdown: handling SHUTDOWN\n"); if (stcb == NULL) return; asoc = &stcb->asoc; if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { return; } if (ntohs(cp->ch.chunk_length) != sizeof(struct sctp_shutdown_chunk)) { /* Shutdown NOT the expected size */ return; } else { sctp_update_acked(stcb, cp, net, abort_flag); if (*abort_flag) { return; } } if (asoc->control_pdapi) { /* * With a normal shutdown we assume the end of last record. */ SCTP_INP_READ_LOCK(stcb->sctp_ep); asoc->control_pdapi->end_added = 1; asoc->control_pdapi->pdapi_aborted = 1; asoc->control_pdapi = NULL; SCTP_INP_READ_UNLOCK(stcb->sctp_ep); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return; } #endif sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } /* goto SHUTDOWN_RECEIVED state to block new requests */ if (stcb->sctp_socket) { if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT)) { SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_RECEIVED); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); /* * notify upper layer that peer has initiated a * shutdown */ sctp_ulp_notify(SCTP_NOTIFY_PEER_SHUTDOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); /* reset time */ (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); } } if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) { /* * stop the shutdown timer, since we WILL move to * SHUTDOWN-ACK-SENT. */ sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_8); } /* Now is there unsent data on a stream somewhere? */ some_on_streamwheel = sctp_is_there_unsent_data(stcb); if (!TAILQ_EMPTY(&asoc->send_queue) || !TAILQ_EMPTY(&asoc->sent_queue) || some_on_streamwheel) { /* By returning we will push more data out */ return; } else { /* no outstanding data to send, so move on... */ /* send SHUTDOWN-ACK */ sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); /* move to SHUTDOWN-ACK-SENT state */ if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_ACK_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(stcb); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, stcb->sctp_ep, stcb, net); } } static void sctp_handle_shutdown_ack(struct sctp_shutdown_ack_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); #endif SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_shutdown_ack: handling SHUTDOWN ACK\n"); if (stcb == NULL) return; asoc = &stcb->asoc; /* process according to association state */ if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { /* unexpected SHUTDOWN-ACK... do OOTB handling... */ sctp_send_shutdown_complete(stcb, net, 1); SCTP_TCB_UNLOCK(stcb); return; } if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* unexpected SHUTDOWN-ACK... so ignore... */ SCTP_TCB_UNLOCK(stcb); return; } if (asoc->control_pdapi) { /* * With a normal shutdown we assume the end of last record. */ SCTP_INP_READ_LOCK(stcb->sctp_ep); asoc->control_pdapi->end_added = 1; asoc->control_pdapi->pdapi_aborted = 1; asoc->control_pdapi = NULL; SCTP_INP_READ_UNLOCK(stcb->sctp_ep); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return; } #endif sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } /* are the queues empty? */ if (!TAILQ_EMPTY(&asoc->send_queue) || !TAILQ_EMPTY(&asoc->sent_queue) || !stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { sctp_report_all_outbound(stcb, 0, SCTP_SO_NOT_LOCKED); } /* stop the timer */ sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_9); /* send SHUTDOWN-COMPLETE */ sctp_send_shutdown_complete(stcb, net, 0); /* notify upper layer protocol */ if (stcb->sctp_socket) { sctp_ulp_notify(SCTP_NOTIFY_ASSOC_DOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* Set the connected flag to disconnected */ stcb->sctp_ep->sctp_socket->so_snd.sb_cc = 0; } } SCTP_STAT_INCR_COUNTER32(sctps_shutdown); /* free the TCB but first save off the ep */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_10); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } /* * Skip past the param header and then we will find the chunk that caused the * problem. There are two possiblities ASCONF or FWD-TSN other than that and * our peer must be broken. */ static void sctp_process_unrecog_chunk(struct sctp_tcb *stcb, struct sctp_paramhdr *phdr, struct sctp_nets *net) { struct sctp_chunkhdr *chk; chk = (struct sctp_chunkhdr *)((caddr_t)phdr + sizeof(*phdr)); switch (chk->chunk_type) { case SCTP_ASCONF_ACK: case SCTP_ASCONF: sctp_asconf_cleanup(stcb, net); break; case SCTP_FORWARD_CUM_TSN: stcb->asoc.peer_supports_prsctp = 0; break; default: SCTPDBG(SCTP_DEBUG_INPUT2, "Peer does not support chunk type %d(%x)??\n", chk->chunk_type, (uint32_t) chk->chunk_type); break; } } /* * Skip past the param header and then we will find the param that caused the * problem. There are a number of param's in a ASCONF OR the prsctp param * these will turn of specific features. */ static void sctp_process_unrecog_param(struct sctp_tcb *stcb, struct sctp_paramhdr *phdr) { struct sctp_paramhdr *pbad; pbad = phdr + 1; switch (ntohs(pbad->param_type)) { /* pr-sctp draft */ case SCTP_PRSCTP_SUPPORTED: stcb->asoc.peer_supports_prsctp = 0; break; case SCTP_SUPPORTED_CHUNK_EXT: break; /* draft-ietf-tsvwg-addip-sctp */ case SCTP_HAS_NAT_SUPPORT: stcb->asoc.peer_supports_nat = 0; break; - case SCTP_ECN_NONCE_SUPPORTED: - stcb->asoc.peer_supports_ecn_nonce = 0; - stcb->asoc.ecn_nonce_allowed = 0; - stcb->asoc.ecn_allowed = 0; - break; case SCTP_ADD_IP_ADDRESS: case SCTP_DEL_IP_ADDRESS: case SCTP_SET_PRIM_ADDR: stcb->asoc.peer_supports_asconf = 0; break; case SCTP_SUCCESS_REPORT: case SCTP_ERROR_CAUSE_IND: SCTPDBG(SCTP_DEBUG_INPUT2, "Huh, the peer does not support success? or error cause?\n"); SCTPDBG(SCTP_DEBUG_INPUT2, "Turning off ASCONF to this strange peer\n"); stcb->asoc.peer_supports_asconf = 0; break; default: SCTPDBG(SCTP_DEBUG_INPUT2, "Peer does not support param type %d(%x)??\n", pbad->param_type, (uint32_t) pbad->param_type); break; } } static int sctp_handle_error(struct sctp_chunkhdr *ch, struct sctp_tcb *stcb, struct sctp_nets *net) { int chklen; struct sctp_paramhdr *phdr; uint16_t error_type; uint16_t error_len; struct sctp_association *asoc; int adjust; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif /* parse through all of the errors and process */ asoc = &stcb->asoc; phdr = (struct sctp_paramhdr *)((caddr_t)ch + sizeof(struct sctp_chunkhdr)); chklen = ntohs(ch->chunk_length) - sizeof(struct sctp_chunkhdr); while ((size_t)chklen >= sizeof(struct sctp_paramhdr)) { /* Process an Error Cause */ error_type = ntohs(phdr->param_type); error_len = ntohs(phdr->param_length); if ((error_len > chklen) || (error_len == 0)) { /* invalid param length for this param */ SCTPDBG(SCTP_DEBUG_INPUT1, "Bogus length in error param- chunk left:%d errorlen:%d\n", chklen, error_len); return (0); } switch (error_type) { case SCTP_CAUSE_INVALID_STREAM: case SCTP_CAUSE_MISSING_PARAM: case SCTP_CAUSE_INVALID_PARAM: case SCTP_CAUSE_NO_USER_DATA: SCTPDBG(SCTP_DEBUG_INPUT1, "Software error we got a %d back? We have a bug :/ (or do they?)\n", error_type); break; case SCTP_CAUSE_NAT_COLLIDING_STATE: SCTPDBG(SCTP_DEBUG_INPUT2, "Received Colliding state abort flags:%x\n", ch->chunk_flags); if (sctp_handle_nat_colliding_state(stcb)) { return (0); } break; case SCTP_CAUSE_NAT_MISSING_STATE: SCTPDBG(SCTP_DEBUG_INPUT2, "Received missing state abort flags:%x\n", ch->chunk_flags); if (sctp_handle_nat_missing_state(stcb, net)) { return (0); } break; case SCTP_CAUSE_STALE_COOKIE: /* * We only act if we have echoed a cookie and are * waiting. */ if (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) { int *p; p = (int *)((caddr_t)phdr + sizeof(*phdr)); /* Save the time doubled */ asoc->cookie_preserve_req = ntohl(*p) << 1; asoc->stale_cookie_count++; if (asoc->stale_cookie_count > asoc->max_init_times) { sctp_abort_notification(stcb, 0, SCTP_SO_NOT_LOCKED); /* now free the asoc */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_11); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif return (-1); } /* blast back to INIT state */ sctp_toss_old_cookies(stcb, &stcb->asoc); asoc->state &= ~SCTP_STATE_COOKIE_ECHOED; asoc->state |= SCTP_STATE_COOKIE_WAIT; sctp_stop_all_cookie_timers(stcb); sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); } break; case SCTP_CAUSE_UNRESOLVABLE_ADDR: /* * Nothing we can do here, we don't do hostname * addresses so if the peer does not like my IPv6 * (or IPv4 for that matter) it does not matter. If * they don't support that type of address, they can * NOT possibly get that packet type... i.e. with no * IPv6 you can't recieve a IPv6 packet. so we can * safely ignore this one. If we ever added support * for HOSTNAME Addresses, then we would need to do * something here. */ break; case SCTP_CAUSE_UNRECOG_CHUNK: sctp_process_unrecog_chunk(stcb, phdr, net); break; case SCTP_CAUSE_UNRECOG_PARAM: sctp_process_unrecog_param(stcb, phdr); break; case SCTP_CAUSE_COOKIE_IN_SHUTDOWN: /* * We ignore this since the timer will drive out a * new cookie anyway and there timer will drive us * to send a SHUTDOWN_COMPLETE. We can't send one * here since we don't have their tag. */ break; case SCTP_CAUSE_DELETING_LAST_ADDR: case SCTP_CAUSE_RESOURCE_SHORTAGE: case SCTP_CAUSE_DELETING_SRC_ADDR: /* * We should NOT get these here, but in a * ASCONF-ACK. */ SCTPDBG(SCTP_DEBUG_INPUT2, "Peer sends ASCONF errors in a Operational Error?<%d>?\n", error_type); break; case SCTP_CAUSE_OUT_OF_RESC: /* * And what, pray tell do we do with the fact that * the peer is out of resources? Not really sure we * could do anything but abort. I suspect this * should have came WITH an abort instead of in a * OP-ERROR. */ break; default: SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_handle_error: unknown error type = 0x%xh\n", error_type); break; } adjust = SCTP_SIZE32(error_len); chklen -= adjust; phdr = (struct sctp_paramhdr *)((caddr_t)phdr + adjust); } return (0); } static int sctp_handle_init_ack(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_init_ack_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net, int *abort_no_unlock, uint32_t vrf_id) { struct sctp_init_ack *init_ack; struct mbuf *op_err; SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_init_ack: handling INIT-ACK\n"); if (stcb == NULL) { SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_init_ack: TCB is null\n"); return (-1); } if (ntohs(cp->ch.chunk_length) < sizeof(struct sctp_init_ack_chunk)) { /* Invalid length */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; return (-1); } init_ack = &cp->init; /* validate parameters */ if (init_ack->initiate_tag == 0) { /* protocol error... send an abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; return (-1); } if (ntohl(init_ack->a_rwnd) < SCTP_MIN_RWND) { /* protocol error... send an abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; return (-1); } if (init_ack->num_inbound_streams == 0) { /* protocol error... send an abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; return (-1); } if (init_ack->num_outbound_streams == 0) { /* protocol error... send an abort */ op_err = sctp_generate_invmanparam(SCTP_CAUSE_INVALID_PARAM); sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, sh, op_err, 0, net->port); *abort_no_unlock = 1; return (-1); } /* process according to association state... */ switch (stcb->asoc.state & SCTP_STATE_MASK) { case SCTP_STATE_COOKIE_WAIT: /* this is the expected state for this chunk */ /* process the INIT-ACK parameters */ if (stcb->asoc.primary_destination->dest_state & SCTP_ADDR_UNCONFIRMED) { /* * The primary is where we sent the INIT, we can * always consider it confirmed when the INIT-ACK is * returned. Do this before we load addresses * though. */ stcb->asoc.primary_destination->dest_state &= ~SCTP_ADDR_UNCONFIRMED; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, stcb, 0, (void *)stcb->asoc.primary_destination, SCTP_SO_NOT_LOCKED); } if (sctp_process_init_ack(m, iphlen, offset, sh, cp, stcb, net, abort_no_unlock, vrf_id) < 0) { /* error in parsing parameters */ return (-1); } /* update our state */ SCTPDBG(SCTP_DEBUG_INPUT2, "moving to COOKIE-ECHOED state\n"); SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_ECHOED); /* reset the RTO calc */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); /* * collapse the init timer back in case of a exponential * backoff */ sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, stcb->sctp_ep, stcb, net); /* * the send at the end of the inbound data processing will * cause the cookie to be sent */ break; case SCTP_STATE_SHUTDOWN_SENT: /* incorrect state... discard */ break; case SCTP_STATE_COOKIE_ECHOED: /* incorrect state... discard */ break; case SCTP_STATE_OPEN: /* incorrect state... discard */ break; case SCTP_STATE_EMPTY: case SCTP_STATE_INUSE: default: /* incorrect state... discard */ return (-1); break; } SCTPDBG(SCTP_DEBUG_INPUT1, "Leaving handle-init-ack end\n"); return (0); } static struct sctp_tcb * sctp_process_cookie_new(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, struct sctp_inpcb *inp, struct sctp_nets **netp, struct sockaddr *init_src, int *notification, int auth_skipped, uint32_t auth_offset, uint32_t auth_len, uint32_t vrf_id, uint16_t port); /* * handle a state cookie for an existing association m: input packet mbuf * chain-- assumes a pullup on IP/SCTP/COOKIE-ECHO chunk note: this is a * "split" mbuf and the cookie signature does not exist offset: offset into * mbuf to the cookie-echo chunk */ static struct sctp_tcb * sctp_process_cookie_existing(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets **netp, struct sockaddr *init_src, int *notification, sctp_assoc_t * sac_assoc_id, uint32_t vrf_id, int auth_skipped, uint32_t auth_offset, uint32_t auth_len, uint16_t port) { struct sctp_association *asoc; struct sctp_init_chunk *init_cp, init_buf; struct sctp_init_ack_chunk *initack_cp, initack_buf; struct sctp_nets *net; struct mbuf *op_err; struct sctp_paramhdr *ph; int chk_length; int init_offset, initack_offset, i; int retval; int spec_flag = 0; uint32_t how_indx; net = *netp; /* I know that the TCB is non-NULL from the caller */ asoc = &stcb->asoc; for (how_indx = 0; how_indx < sizeof(asoc->cookie_how); how_indx++) { if (asoc->cookie_how[how_indx] == 0) break; } if (how_indx < sizeof(asoc->cookie_how)) { asoc->cookie_how[how_indx] = 1; } if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) { /* SHUTDOWN came in after sending INIT-ACK */ sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (op_err == NULL) { /* FOOBAR */ return (NULL); } /* Set the len */ SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_COOKIE_IN_SHUTDOWN); ph->param_length = htons(sizeof(struct sctp_paramhdr)); sctp_send_operr_to(m, iphlen, op_err, cookie->peers_vtag, vrf_id, net->port); if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 2; return (NULL); } /* * find and validate the INIT chunk in the cookie (peer's info) the * INIT should start after the cookie-echo header struct (chunk * header, state cookie header struct) */ init_offset = offset += sizeof(struct sctp_cookie_echo_chunk); init_cp = (struct sctp_init_chunk *) sctp_m_getptr(m, init_offset, sizeof(struct sctp_init_chunk), (uint8_t *) & init_buf); if (init_cp == NULL) { /* could not pull a INIT chunk in cookie */ return (NULL); } chk_length = ntohs(init_cp->ch.chunk_length); if (init_cp->ch.chunk_type != SCTP_INITIATION) { return (NULL); } /* * find and validate the INIT-ACK chunk in the cookie (my info) the * INIT-ACK follows the INIT chunk */ initack_offset = init_offset + SCTP_SIZE32(chk_length); initack_cp = (struct sctp_init_ack_chunk *) sctp_m_getptr(m, initack_offset, sizeof(struct sctp_init_ack_chunk), (uint8_t *) & initack_buf); if (initack_cp == NULL) { /* could not pull INIT-ACK chunk in cookie */ return (NULL); } chk_length = ntohs(initack_cp->ch.chunk_length); if (initack_cp->ch.chunk_type != SCTP_INITIATION_ACK) { return (NULL); } if ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && (ntohl(init_cp->init.initiate_tag) == asoc->peer_vtag)) { /* * case D in Section 5.2.4 Table 2: MMAA process accordingly * to get into the OPEN state */ if (ntohl(initack_cp->init.initial_tsn) != asoc->init_seq_number) { /*- * Opps, this means that we somehow generated two vtag's * the same. I.e. we did: * Us Peer * <---INIT(tag=a)------ * ----INIT-ACK(tag=t)--> * ----INIT(tag=t)------> *1 * <---INIT-ACK(tag=a)--- * <----CE(tag=t)------------- *2 * * At point *1 we should be generating a different * tag t'. Which means we would throw away the CE and send * ours instead. Basically this is case C (throw away side). */ if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 17; return (NULL); } switch SCTP_GET_STATE (asoc) { case SCTP_STATE_COOKIE_WAIT: case SCTP_STATE_COOKIE_ECHOED: /* * INIT was sent but got a COOKIE_ECHO with the * correct tags... just accept it...but we must * process the init so that we can make sure we have * the right seq no's. */ /* First we must process the INIT !! */ retval = sctp_process_init(init_cp, stcb, net); if (retval < 0) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 3; return (NULL); } /* we have already processed the INIT so no problem */ sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_12); sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_13); /* update current state */ if (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) SCTP_STAT_INCR_COUNTER32(sctps_activeestab); else SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } SCTP_STAT_INCR_GAUGE32(sctps_currestab); sctp_stop_all_cookie_timers(stcb); if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && (inp->sctp_socket->so_qlimit == 0) ) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif /* * Here is where collision would go if we * did a connect() and instead got a * init/init-ack/cookie done before the * init-ack came back.. */ stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return (NULL); } #endif soisconnected(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } /* notify upper layer */ *notification = SCTP_NOTIFY_ASSOC_UP; /* * since we did not send a HB make sure we don't * double things */ net->hb_responded = 1; net->RTO = sctp_calculate_rto(stcb, asoc, net, &cookie->time_entered, sctp_align_unsafe_makecopy); if (stcb->asoc.sctp_autoclose_ticks && (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE))) { sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); } break; default: /* * we're in the OPEN state (or beyond), so peer must * have simply lost the COOKIE-ACK */ break; } /* end switch */ sctp_stop_all_cookie_timers(stcb); /* * We ignore the return code here.. not sure if we should * somehow abort.. but we do have an existing asoc. This * really should not fail. */ if (sctp_load_addresses_from_init(stcb, m, iphlen, init_offset + sizeof(struct sctp_init_chunk), initack_offset, sh, init_src)) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 4; return (NULL); } /* respond with a COOKIE-ACK */ sctp_toss_old_cookies(stcb, asoc); sctp_send_cookie_ack(stcb); if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 5; return (stcb); } if (ntohl(initack_cp->init.initiate_tag) != asoc->my_vtag && ntohl(init_cp->init.initiate_tag) == asoc->peer_vtag && cookie->tie_tag_my_vtag == 0 && cookie->tie_tag_peer_vtag == 0) { /* * case C in Section 5.2.4 Table 2: XMOO silently discard */ if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 6; return (NULL); } /* * If nat support, and the below and stcb is established, send back * a ABORT(colliding state) if we are established. */ if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) && (asoc->peer_supports_nat) && ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && ((ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) || (asoc->peer_vtag == 0)))) { /* * Special case - Peer's support nat. We may have two init's * that we gave out the same tag on since one was not * established.. i.e. we get INIT from host-1 behind the nat * and we respond tag-a, we get a INIT from host-2 behind * the nat and we get tag-a again. Then we bring up host-1 * (or 2's) assoc, Then comes the cookie from hsot-2 (or 1). * Now we have colliding state. We must send an abort here * with colliding state indication. */ op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (op_err == NULL) { /* FOOBAR */ return (NULL); } /* pre-reserve some space */ #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); /* Set the len */ SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_NAT_COLLIDING_STATE); ph->param_length = htons(sizeof(struct sctp_paramhdr)); sctp_send_abort(m, iphlen, sh, 0, op_err, vrf_id, port); return (NULL); } if ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && ((ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) || (asoc->peer_vtag == 0))) { /* * case B in Section 5.2.4 Table 2: MXAA or MOAA my info * should be ok, re-accept peer info */ if (ntohl(initack_cp->init.initial_tsn) != asoc->init_seq_number) { /* * Extension of case C. If we hit this, then the * random number generator returned the same vtag * when we first sent our INIT-ACK and when we later * sent our INIT. The side with the seq numbers that * are different will be the one that normnally * would have hit case C. This in effect "extends" * our vtags in this collision case to be 64 bits. * The same collision could occur aka you get both * vtag and seq number the same twice in a row.. but * is much less likely. If it did happen then we * would proceed through and bring up the assoc.. we * may end up with the wrong stream setup however.. * which would be bad.. but there is no way to * tell.. until we send on a stream that does not * exist :-) */ if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 7; return (NULL); } if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 8; sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_14); sctp_stop_all_cookie_timers(stcb); /* * since we did not send a HB make sure we don't double * things */ net->hb_responded = 1; if (stcb->asoc.sctp_autoclose_ticks && sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); } asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); asoc->pre_open_streams = ntohs(initack_cp->init.num_outbound_streams); if (ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) { /* * Ok the peer probably discarded our data (if we * echoed a cookie+data). So anything on the * sent_queue should be marked for retransmit, we * may not get something to kick us so it COULD * still take a timeout to move these.. but it can't * hurt to mark them. */ struct sctp_tmit_chunk *chk; TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if (chk->sent < SCTP_DATAGRAM_RESEND) { chk->sent = SCTP_DATAGRAM_RESEND; sctp_flight_size_decrease(chk); sctp_total_flight_decrease(stcb, chk); sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); spec_flag++; } } } /* process the INIT info (peer's info) */ retval = sctp_process_init(init_cp, stcb, net); if (retval < 0) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 9; return (NULL); } if (sctp_load_addresses_from_init(stcb, m, iphlen, init_offset + sizeof(struct sctp_init_chunk), initack_offset, sh, init_src)) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 10; return (NULL); } if ((asoc->state & SCTP_STATE_COOKIE_WAIT) || (asoc->state & SCTP_STATE_COOKIE_ECHOED)) { *notification = SCTP_NOTIFY_ASSOC_UP; if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && (inp->sctp_socket->so_qlimit == 0)) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return (NULL); } #endif soisconnected(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } if (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) SCTP_STAT_INCR_COUNTER32(sctps_activeestab); else SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); SCTP_STAT_INCR_GAUGE32(sctps_currestab); } else if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_INCR_COUNTER32(sctps_restartestab); } else { SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); } SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } sctp_stop_all_cookie_timers(stcb); sctp_toss_old_cookies(stcb, asoc); sctp_send_cookie_ack(stcb); if (spec_flag) { /* * only if we have retrans set do we do this. What * this call does is get only the COOKIE-ACK out and * then when we return the normal call to * sctp_chunk_output will get the retrans out behind * this. */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_COOKIE_ACK, SCTP_SO_NOT_LOCKED); } if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 11; return (stcb); } if ((ntohl(initack_cp->init.initiate_tag) != asoc->my_vtag && ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) && cookie->tie_tag_my_vtag == asoc->my_vtag_nonce && cookie->tie_tag_peer_vtag == asoc->peer_vtag_nonce && cookie->tie_tag_peer_vtag != 0) { struct sctpasochead *head; if (asoc->peer_supports_nat) { /* * This is a gross gross hack. just call the * cookie_new code since we are allowing a duplicate * association. I hope this works... */ return (sctp_process_cookie_new(m, iphlen, offset, sh, cookie, cookie_len, inp, netp, init_src, notification, auth_skipped, auth_offset, auth_len, vrf_id, port)); } /* * case A in Section 5.2.4 Table 2: XXMM (peer restarted) */ /* temp code */ if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 12; sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_15); sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_16); *sac_assoc_id = sctp_get_associd(stcb); /* notify upper layer */ *notification = SCTP_NOTIFY_ASSOC_RESTART; atomic_add_int(&stcb->asoc.refcnt, 1); if ((SCTP_GET_STATE(asoc) != SCTP_STATE_OPEN) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT)) { SCTP_STAT_INCR_GAUGE32(sctps_currestab); } if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_INCR_GAUGE32(sctps_restartestab); } else if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) { SCTP_STAT_INCR_GAUGE32(sctps_collisionestab); } if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } else if (!(asoc->state & SCTP_STATE_SHUTDOWN_SENT)) { /* move to OPEN state, if not in SHUTDOWN_SENT */ SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); } asoc->pre_open_streams = ntohs(initack_cp->init.num_outbound_streams); asoc->init_seq_number = ntohl(initack_cp->init.initial_tsn); asoc->sending_seq = asoc->asconf_seq_out = asoc->str_reset_seq_out = asoc->init_seq_number; asoc->asconf_seq_out_acked = asoc->asconf_seq_out - 1; asoc->asconf_seq_in = asoc->last_acked_seq = asoc->init_seq_number - 1; asoc->str_reset_seq_in = asoc->init_seq_number; asoc->advanced_peer_ack_point = asoc->last_acked_seq; if (asoc->mapping_array) { memset(asoc->mapping_array, 0, asoc->mapping_array_size); } if (asoc->nr_mapping_array) { memset(asoc->nr_mapping_array, 0, asoc->mapping_array_size); } SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(stcb->sctp_ep); SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); /* send up all the data */ SCTP_TCB_SEND_LOCK(stcb); sctp_report_all_outbound(stcb, 1, SCTP_SO_NOT_LOCKED); for (i = 0; i < stcb->asoc.streamoutcnt; i++) { stcb->asoc.strmout[i].stream_no = i; stcb->asoc.strmout[i].next_sequence_sent = 0; stcb->asoc.strmout[i].last_msg_incomplete = 0; } /* process the INIT-ACK info (my info) */ asoc->my_vtag = ntohl(initack_cp->init.initiate_tag); asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); /* pull from vtag hash */ LIST_REMOVE(stcb, sctp_asocs); /* re-insert to new vtag position */ head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* * put it in the bucket in the vtag hash of assoc's for the * system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); /* process the INIT info (peer's info) */ SCTP_TCB_SEND_UNLOCK(stcb); SCTP_INP_WUNLOCK(stcb->sctp_ep); SCTP_INP_INFO_WUNLOCK(); retval = sctp_process_init(init_cp, stcb, net); if (retval < 0) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 13; return (NULL); } /* * since we did not send a HB make sure we don't double * things */ net->hb_responded = 1; if (sctp_load_addresses_from_init(stcb, m, iphlen, init_offset + sizeof(struct sctp_init_chunk), initack_offset, sh, init_src)) { if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 14; return (NULL); } /* respond with a COOKIE-ACK */ sctp_stop_all_cookie_timers(stcb); sctp_toss_old_cookies(stcb, asoc); sctp_send_cookie_ack(stcb); if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 15; return (stcb); } if (how_indx < sizeof(asoc->cookie_how)) asoc->cookie_how[how_indx] = 16; /* all other cases... */ return (NULL); } /* * handle a state cookie for a new association m: input packet mbuf chain-- * assumes a pullup on IP/SCTP/COOKIE-ECHO chunk note: this is a "split" mbuf * and the cookie signature does not exist offset: offset into mbuf to the * cookie-echo chunk length: length of the cookie chunk to: where the init * was from returns a new TCB */ struct sctp_tcb * sctp_process_cookie_new(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, struct sctp_inpcb *inp, struct sctp_nets **netp, struct sockaddr *init_src, int *notification, int auth_skipped, uint32_t auth_offset, uint32_t auth_len, uint32_t vrf_id, uint16_t port) { struct sctp_tcb *stcb; struct sctp_init_chunk *init_cp, init_buf; struct sctp_init_ack_chunk *initack_cp, initack_buf; struct sockaddr_storage sa_store; struct sockaddr *initack_src = (struct sockaddr *)&sa_store; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; struct sctp_association *asoc; int chk_length; int init_offset, initack_offset, initack_limit; int retval; int error = 0; uint32_t old_tag; uint8_t auth_chunk_buf[SCTP_PARAM_BUFFER_SIZE]; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(inp); #endif /* * find and validate the INIT chunk in the cookie (peer's info) the * INIT should start after the cookie-echo header struct (chunk * header, state cookie header struct) */ init_offset = offset + sizeof(struct sctp_cookie_echo_chunk); init_cp = (struct sctp_init_chunk *) sctp_m_getptr(m, init_offset, sizeof(struct sctp_init_chunk), (uint8_t *) & init_buf); if (init_cp == NULL) { /* could not pull a INIT chunk in cookie */ SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: could not pull INIT chunk hdr\n"); return (NULL); } chk_length = ntohs(init_cp->ch.chunk_length); if (init_cp->ch.chunk_type != SCTP_INITIATION) { SCTPDBG(SCTP_DEBUG_INPUT1, "HUH? process_cookie_new: could not find INIT chunk!\n"); return (NULL); } initack_offset = init_offset + SCTP_SIZE32(chk_length); /* * find and validate the INIT-ACK chunk in the cookie (my info) the * INIT-ACK follows the INIT chunk */ initack_cp = (struct sctp_init_ack_chunk *) sctp_m_getptr(m, initack_offset, sizeof(struct sctp_init_ack_chunk), (uint8_t *) & initack_buf); if (initack_cp == NULL) { /* could not pull INIT-ACK chunk in cookie */ SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: could not pull INIT-ACK chunk hdr\n"); return (NULL); } chk_length = ntohs(initack_cp->ch.chunk_length); if (initack_cp->ch.chunk_type != SCTP_INITIATION_ACK) { return (NULL); } /* * NOTE: We can't use the INIT_ACK's chk_length to determine the * "initack_limit" value. This is because the chk_length field * includes the length of the cookie, but the cookie is omitted when * the INIT and INIT_ACK are tacked onto the cookie... */ initack_limit = offset + cookie_len; /* * now that we know the INIT/INIT-ACK are in place, create a new TCB * and popluate */ /* * Here we do a trick, we set in NULL for the proc/thread argument. * We do this since in effect we only use the p argument when the * socket is unbound and we must do an implicit bind. Since we are * getting a cookie, we cannot be unbound. */ stcb = sctp_aloc_assoc(inp, init_src, &error, ntohl(initack_cp->init.initiate_tag), vrf_id, (struct thread *)NULL ); if (stcb == NULL) { struct mbuf *op_err; /* memory problem? */ SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: no room for another TCB!\n"); op_err = sctp_generate_invmanparam(SCTP_CAUSE_OUT_OF_RESC); sctp_abort_association(inp, (struct sctp_tcb *)NULL, m, iphlen, sh, op_err, vrf_id, port); return (NULL); } /* get the correct sctp_nets */ if (netp) *netp = sctp_findnet(stcb, init_src); asoc = &stcb->asoc; /* get scope variables out of cookie */ asoc->ipv4_local_scope = cookie->ipv4_scope; asoc->site_scope = cookie->site_scope; asoc->local_scope = cookie->local_scope; asoc->loopback_scope = cookie->loopback_scope; if ((asoc->ipv4_addr_legal != cookie->ipv4_addr_legal) || (asoc->ipv6_addr_legal != cookie->ipv6_addr_legal)) { struct mbuf *op_err; /* * Houston we have a problem. The EP changed while the * cookie was in flight. Only recourse is to abort the * association. */ atomic_add_int(&stcb->asoc.refcnt, 1); op_err = sctp_generate_invmanparam(SCTP_CAUSE_OUT_OF_RESC); sctp_abort_association(inp, (struct sctp_tcb *)NULL, m, iphlen, sh, op_err, vrf_id, port); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_16); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif atomic_subtract_int(&stcb->asoc.refcnt, 1); return (NULL); } /* process the INIT-ACK info (my info) */ old_tag = asoc->my_vtag; asoc->my_vtag = ntohl(initack_cp->init.initiate_tag); asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); asoc->pre_open_streams = ntohs(initack_cp->init.num_outbound_streams); asoc->init_seq_number = ntohl(initack_cp->init.initial_tsn); asoc->sending_seq = asoc->asconf_seq_out = asoc->str_reset_seq_out = asoc->init_seq_number; asoc->asconf_seq_out_acked = asoc->asconf_seq_out - 1; asoc->asconf_seq_in = asoc->last_acked_seq = asoc->init_seq_number - 1; asoc->str_reset_seq_in = asoc->init_seq_number; asoc->advanced_peer_ack_point = asoc->last_acked_seq; /* process the INIT info (peer's info) */ if (netp) retval = sctp_process_init(init_cp, stcb, *netp); else retval = 0; if (retval < 0) { atomic_add_int(&stcb->asoc.refcnt, 1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_16); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif atomic_subtract_int(&stcb->asoc.refcnt, 1); return (NULL); } /* load all addresses */ if (sctp_load_addresses_from_init(stcb, m, iphlen, init_offset + sizeof(struct sctp_init_chunk), initack_offset, sh, init_src)) { atomic_add_int(&stcb->asoc.refcnt, 1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_17); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif atomic_subtract_int(&stcb->asoc.refcnt, 1); return (NULL); } /* * verify any preceding AUTH chunk that was skipped */ /* pull the local authentication parameters from the cookie/init-ack */ sctp_auth_get_cookie_params(stcb, m, initack_offset + sizeof(struct sctp_init_ack_chunk), initack_limit - (initack_offset + sizeof(struct sctp_init_ack_chunk))); if (auth_skipped) { struct sctp_auth_chunk *auth; auth = (struct sctp_auth_chunk *) sctp_m_getptr(m, auth_offset, auth_len, auth_chunk_buf); if ((auth == NULL) || sctp_handle_auth(stcb, auth, m, auth_offset)) { /* auth HMAC failed, dump the assoc and packet */ SCTPDBG(SCTP_DEBUG_AUTH1, "COOKIE-ECHO: AUTH failed\n"); atomic_add_int(&stcb->asoc.refcnt, 1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_18); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif atomic_subtract_int(&stcb->asoc.refcnt, 1); return (NULL); } else { /* remaining chunks checked... good to go */ stcb->asoc.authenticated = 1; } } /* update current state */ SCTPDBG(SCTP_DEBUG_INPUT2, "moving to OPEN state\n"); SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } sctp_stop_all_cookie_timers(stcb); SCTP_STAT_INCR_COUNTER32(sctps_passiveestab); SCTP_STAT_INCR_GAUGE32(sctps_currestab); /* * if we're doing ASCONFs, check to see if we have any new local * addresses that need to get added to the peer (eg. addresses * changed while cookie echo in flight). This needs to be done * after we go to the OPEN state to do the correct asconf * processing. else, make sure we have the correct addresses in our * lists */ /* warning, we re-use sin, sin6, sa_store here! */ /* pull in local_address (our "from" address) */ if (cookie->laddr_type == SCTP_IPV4_ADDRESS) { /* source addr is IPv4 */ sin = (struct sockaddr_in *)initack_src; memset(sin, 0, sizeof(*sin)); sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_addr.s_addr = cookie->laddress[0]; } else if (cookie->laddr_type == SCTP_IPV6_ADDRESS) { /* source addr is IPv6 */ sin6 = (struct sockaddr_in6 *)initack_src; memset(sin6, 0, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); sin6->sin6_scope_id = cookie->scope_id; memcpy(&sin6->sin6_addr, cookie->laddress, sizeof(sin6->sin6_addr)); } else { atomic_add_int(&stcb->asoc.refcnt, 1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_19); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif atomic_subtract_int(&stcb->asoc.refcnt, 1); return (NULL); } /* set up to notify upper layer */ *notification = SCTP_NOTIFY_ASSOC_UP; if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && (inp->sctp_socket->so_qlimit == 0)) { /* * This is an endpoint that called connect() how it got a * cookie that is NEW is a bit of a mystery. It must be that * the INIT was sent, but before it got there.. a complete * INIT/INIT-ACK/COOKIE arrived. But of course then it * should have went to the other code.. not here.. oh well.. * a bit of protection is worth having.. */ stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return (NULL); } #endif soisconnected(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } else if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_socket->so_qlimit)) { /* * We don't want to do anything with this one. Since it is * the listening guy. The timer will get started for * accepted connections in the caller. */ ; } /* since we did not send a HB make sure we don't double things */ if ((netp) && (*netp)) (*netp)->hb_responded = 1; if (stcb->asoc.sctp_autoclose_ticks && sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); } /* calculate the RTT */ (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); if ((netp) && (*netp)) { (*netp)->RTO = sctp_calculate_rto(stcb, asoc, *netp, &cookie->time_entered, sctp_align_unsafe_makecopy); } /* respond with a COOKIE-ACK */ sctp_send_cookie_ack(stcb); /* * check the address lists for any ASCONFs that need to be sent * AFTER the cookie-ack is sent */ sctp_check_address_list(stcb, m, initack_offset + sizeof(struct sctp_init_ack_chunk), initack_limit - (initack_offset + sizeof(struct sctp_init_ack_chunk)), initack_src, cookie->local_scope, cookie->site_scope, cookie->ipv4_scope, cookie->loopback_scope); return (stcb); } /* * CODE LIKE THIS NEEDS TO RUN IF the peer supports the NAT extension, i.e * we NEED to make sure we are not already using the vtag. If so we * need to send back an ABORT-TRY-AGAIN-WITH-NEW-TAG No middle box bit! head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))]; LIST_FOREACH(stcb, head, sctp_asocs) { if ((stcb->asoc.my_vtag == tag) && (stcb->rport == rport) && (inp == stcb->sctp_ep)) { -- SEND ABORT - TRY AGAIN -- } } */ /* * handles a COOKIE-ECHO message stcb: modified to either a new or left as * existing (non-NULL) TCB */ static struct mbuf * sctp_handle_cookie_echo(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_cookie_echo_chunk *cp, struct sctp_inpcb **inp_p, struct sctp_tcb **stcb, struct sctp_nets **netp, int auth_skipped, uint32_t auth_offset, uint32_t auth_len, struct sctp_tcb **locked_tcb, uint32_t vrf_id, uint16_t port) { struct sctp_state_cookie *cookie; struct sockaddr_in6 sin6; struct sockaddr_in sin; struct sctp_tcb *l_stcb = *stcb; struct sctp_inpcb *l_inp; struct sockaddr *to; sctp_assoc_t sac_restart_id; struct sctp_pcb *ep; struct mbuf *m_sig; uint8_t calc_sig[SCTP_SIGNATURE_SIZE], tmp_sig[SCTP_SIGNATURE_SIZE]; uint8_t *sig; uint8_t cookie_ok = 0; unsigned int size_of_pkt, sig_offset, cookie_offset; unsigned int cookie_len; struct timeval now; struct timeval time_expires; struct sockaddr_storage dest_store; struct sockaddr *localep_sa = (struct sockaddr *)&dest_store; struct ip *iph; int notification = 0; struct sctp_nets *netl; int had_a_existing_tcb = 0; int send_int_conf = 0; SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_cookie: handling COOKIE-ECHO\n"); if (inp_p == NULL) { return (NULL); } /* First get the destination address setup too. */ iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *lsin; lsin = (struct sockaddr_in *)(localep_sa); memset(lsin, 0, sizeof(*lsin)); lsin->sin_family = AF_INET; lsin->sin_len = sizeof(*lsin); lsin->sin_port = sh->dest_port; lsin->sin_addr.s_addr = iph->ip_dst.s_addr; size_of_pkt = SCTP_GET_IPV4_LENGTH(iph); break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *lsin6; lsin6 = (struct sockaddr_in6 *)(localep_sa); memset(lsin6, 0, sizeof(*lsin6)); lsin6->sin6_family = AF_INET6; lsin6->sin6_len = sizeof(struct sockaddr_in6); ip6 = mtod(m, struct ip6_hdr *); lsin6->sin6_port = sh->dest_port; lsin6->sin6_addr = ip6->ip6_dst; size_of_pkt = SCTP_GET_IPV6_LENGTH(ip6) + iphlen; break; } #endif default: return (NULL); } cookie = &cp->cookie; cookie_offset = offset + sizeof(struct sctp_chunkhdr); cookie_len = ntohs(cp->ch.chunk_length); if ((cookie->peerport != sh->src_port) && (cookie->myport != sh->dest_port) && (cookie->my_vtag != sh->v_tag)) { /* * invalid ports or bad tag. Note that we always leave the * v_tag in the header in network order and when we stored * it in the my_vtag slot we also left it in network order. * This maintains the match even though it may be in the * opposite byte order of the machine :-> */ return (NULL); } if (cookie_len > size_of_pkt || cookie_len < sizeof(struct sctp_cookie_echo_chunk) + sizeof(struct sctp_init_chunk) + sizeof(struct sctp_init_ack_chunk) + SCTP_SIGNATURE_SIZE) { /* cookie too long! or too small */ return (NULL); } /* * split off the signature into its own mbuf (since it should not be * calculated in the sctp_hmac_m() call). */ sig_offset = offset + cookie_len - SCTP_SIGNATURE_SIZE; if (sig_offset > size_of_pkt) { /* packet not correct size! */ /* XXX this may already be accounted for earlier... */ return (NULL); } m_sig = m_split(m, sig_offset, M_DONTWAIT); if (m_sig == NULL) { /* out of memory or ?? */ return (NULL); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = m_sig; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_SPLIT); } mat = SCTP_BUF_NEXT(mat); } } #endif /* * compute the signature/digest for the cookie */ ep = &(*inp_p)->sctp_ep; l_inp = *inp_p; if (l_stcb) { SCTP_TCB_UNLOCK(l_stcb); } SCTP_INP_RLOCK(l_inp); if (l_stcb) { SCTP_TCB_LOCK(l_stcb); } /* which cookie is it? */ if ((cookie->time_entered.tv_sec < (long)ep->time_of_secret_change) && (ep->current_secret_number != ep->last_secret_number)) { /* it's the old cookie */ (void)sctp_hmac_m(SCTP_HMAC, (uint8_t *) ep->secret_key[(int)ep->last_secret_number], SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); } else { /* it's the current cookie */ (void)sctp_hmac_m(SCTP_HMAC, (uint8_t *) ep->secret_key[(int)ep->current_secret_number], SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); } /* get the signature */ SCTP_INP_RUNLOCK(l_inp); sig = (uint8_t *) sctp_m_getptr(m_sig, 0, SCTP_SIGNATURE_SIZE, (uint8_t *) & tmp_sig); if (sig == NULL) { /* couldn't find signature */ sctp_m_freem(m_sig); return (NULL); } /* compare the received digest with the computed digest */ if (memcmp(calc_sig, sig, SCTP_SIGNATURE_SIZE) != 0) { /* try the old cookie? */ if ((cookie->time_entered.tv_sec == (long)ep->time_of_secret_change) && (ep->current_secret_number != ep->last_secret_number)) { /* compute digest with old */ (void)sctp_hmac_m(SCTP_HMAC, (uint8_t *) ep->secret_key[(int)ep->last_secret_number], SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); /* compare */ if (memcmp(calc_sig, sig, SCTP_SIGNATURE_SIZE) == 0) cookie_ok = 1; } } else { cookie_ok = 1; } /* * Now before we continue we must reconstruct our mbuf so that * normal processing of any other chunks will work. */ { struct mbuf *m_at; m_at = m; while (SCTP_BUF_NEXT(m_at) != NULL) { m_at = SCTP_BUF_NEXT(m_at); } SCTP_BUF_NEXT(m_at) = m_sig; } if (cookie_ok == 0) { SCTPDBG(SCTP_DEBUG_INPUT2, "handle_cookie_echo: cookie signature validation failed!\n"); SCTPDBG(SCTP_DEBUG_INPUT2, "offset = %u, cookie_offset = %u, sig_offset = %u\n", (uint32_t) offset, cookie_offset, sig_offset); return (NULL); } /* * check the cookie timestamps to be sure it's not stale */ (void)SCTP_GETTIME_TIMEVAL(&now); /* Expire time is in Ticks, so we convert to seconds */ time_expires.tv_sec = cookie->time_entered.tv_sec + TICKS_TO_SEC(cookie->cookie_life); time_expires.tv_usec = cookie->time_entered.tv_usec; /* * TODO sctp_constants.h needs alternative time macros when _KERNEL * is undefined. */ if (timevalcmp(&now, &time_expires, >)) { /* cookie is stale! */ struct mbuf *op_err; struct sctp_stale_cookie_msg *scm; uint32_t tim; op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_stale_cookie_msg), 0, M_DONTWAIT, 1, MT_DATA); if (op_err == NULL) { /* FOOBAR */ return (NULL); } /* Set the len */ SCTP_BUF_LEN(op_err) = sizeof(struct sctp_stale_cookie_msg); scm = mtod(op_err, struct sctp_stale_cookie_msg *); scm->ph.param_type = htons(SCTP_CAUSE_STALE_COOKIE); scm->ph.param_length = htons((sizeof(struct sctp_paramhdr) + (sizeof(uint32_t)))); /* seconds to usec */ tim = (now.tv_sec - time_expires.tv_sec) * 1000000; /* add in usec */ if (tim == 0) tim = now.tv_usec - cookie->time_entered.tv_usec; scm->time_usec = htonl(tim); sctp_send_operr_to(m, iphlen, op_err, cookie->peers_vtag, vrf_id, port); return (NULL); } /* * Now we must see with the lookup address if we have an existing * asoc. This will only happen if we were in the COOKIE-WAIT state * and a INIT collided with us and somewhere the peer sent the * cookie on another address besides the single address our assoc * had for him. In this case we will have one of the tie-tags set at * least AND the address field in the cookie can be used to look it * up. */ to = NULL; if (cookie->addr_type == SCTP_IPV6_ADDRESS) { memset(&sin6, 0, sizeof(sin6)); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(sin6); sin6.sin6_port = sh->src_port; sin6.sin6_scope_id = cookie->scope_id; memcpy(&sin6.sin6_addr.s6_addr, cookie->address, sizeof(sin6.sin6_addr.s6_addr)); to = (struct sockaddr *)&sin6; } else if (cookie->addr_type == SCTP_IPV4_ADDRESS) { memset(&sin, 0, sizeof(sin)); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = sh->src_port; sin.sin_addr.s_addr = cookie->address[0]; to = (struct sockaddr *)&sin; } else { /* This should not happen */ return (NULL); } if ((*stcb == NULL) && to) { /* Yep, lets check */ *stcb = sctp_findassociation_ep_addr(inp_p, to, netp, localep_sa, NULL); if (*stcb == NULL) { /* * We should have only got back the same inp. If we * got back a different ep we have a problem. The * original findep got back l_inp and now */ if (l_inp != *inp_p) { SCTP_PRINTF("Bad problem find_ep got a diff inp then special_locate?\n"); } } else { if (*locked_tcb == NULL) { /* * In this case we found the assoc only * after we locked the create lock. This * means we are in a colliding case and we * must make sure that we unlock the tcb if * its one of the cases where we throw away * the incoming packets. */ *locked_tcb = *stcb; /* * We must also increment the inp ref count * since the ref_count flags was set when we * did not find the TCB, now we found it * which reduces the refcount.. we must * raise it back out to balance it all :-) */ SCTP_INP_INCR_REF((*stcb)->sctp_ep); if ((*stcb)->sctp_ep != l_inp) { SCTP_PRINTF("Huh? ep:%p diff then l_inp:%p?\n", (*stcb)->sctp_ep, l_inp); } } } } if (to == NULL) { return (NULL); } cookie_len -= SCTP_SIGNATURE_SIZE; if (*stcb == NULL) { /* this is the "normal" case... get a new TCB */ *stcb = sctp_process_cookie_new(m, iphlen, offset, sh, cookie, cookie_len, *inp_p, netp, to, ¬ification, auth_skipped, auth_offset, auth_len, vrf_id, port); } else { /* this is abnormal... cookie-echo on existing TCB */ had_a_existing_tcb = 1; *stcb = sctp_process_cookie_existing(m, iphlen, offset, sh, cookie, cookie_len, *inp_p, *stcb, netp, to, ¬ification, &sac_restart_id, vrf_id, auth_skipped, auth_offset, auth_len, port); } if (*stcb == NULL) { /* still no TCB... must be bad cookie-echo */ return (NULL); } /* * Ok, we built an association so confirm the address we sent the * INIT-ACK to. */ netl = sctp_findnet(*stcb, to); /* * This code should in theory NOT run but */ if (netl == NULL) { /* TSNH! Huh, why do I need to add this address here? */ int ret; ret = sctp_add_remote_addr(*stcb, to, SCTP_DONOT_SETSCOPE, SCTP_IN_COOKIE_PROC); netl = sctp_findnet(*stcb, to); } if (netl) { if (netl->dest_state & SCTP_ADDR_UNCONFIRMED) { netl->dest_state &= ~SCTP_ADDR_UNCONFIRMED; (void)sctp_set_primary_addr((*stcb), (struct sockaddr *)NULL, netl); send_int_conf = 1; } } if (*stcb) { sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, *inp_p, *stcb, NULL); } if ((*inp_p)->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { if (!had_a_existing_tcb || (((*inp_p)->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { /* * If we have a NEW cookie or the connect never * reached the connected state during collision we * must do the TCP accept thing. */ struct socket *so, *oso; struct sctp_inpcb *inp; if (notification == SCTP_NOTIFY_ASSOC_RESTART) { /* * For a restart we will keep the same * socket, no need to do anything. I THINK!! */ sctp_ulp_notify(notification, *stcb, 0, (void *)&sac_restart_id, SCTP_SO_NOT_LOCKED); if (send_int_conf) { sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); } return (m); } oso = (*inp_p)->sctp_socket; atomic_add_int(&(*stcb)->asoc.refcnt, 1); SCTP_TCB_UNLOCK((*stcb)); so = sonewconn(oso, 0 ); SCTP_TCB_LOCK((*stcb)); atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); if (so == NULL) { struct mbuf *op_err; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *pcb_so; #endif /* Too many sockets */ SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: no room for another socket!\n"); op_err = sctp_generate_invmanparam(SCTP_CAUSE_OUT_OF_RESC); sctp_abort_association(*inp_p, NULL, m, iphlen, sh, op_err, vrf_id, port); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) pcb_so = SCTP_INP_SO(*inp_p); atomic_add_int(&(*stcb)->asoc.refcnt, 1); SCTP_TCB_UNLOCK((*stcb)); SCTP_SOCKET_LOCK(pcb_so, 1); SCTP_TCB_LOCK((*stcb)); atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); #endif (void)sctp_free_assoc(*inp_p, *stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_20); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(pcb_so, 1); #endif return (NULL); } inp = (struct sctp_inpcb *)so->so_pcb; SCTP_INP_INCR_REF(inp); /* * We add the unbound flag here so that if we get an * soabort() before we get the move_pcb done, we * will properly cleanup. */ inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_CONNECTED | SCTP_PCB_FLAGS_IN_TCPPOOL | SCTP_PCB_FLAGS_UNBOUND | (SCTP_PCB_COPY_FLAGS & (*inp_p)->sctp_flags) | SCTP_PCB_FLAGS_DONT_WAKE); inp->sctp_features = (*inp_p)->sctp_features; inp->sctp_mobility_features = (*inp_p)->sctp_mobility_features; inp->sctp_socket = so; inp->sctp_frag_point = (*inp_p)->sctp_frag_point; inp->sctp_cmt_on_off = (*inp_p)->sctp_cmt_on_off; inp->partial_delivery_point = (*inp_p)->partial_delivery_point; inp->sctp_context = (*inp_p)->sctp_context; inp->inp_starting_point_for_iterator = NULL; /* * copy in the authentication parameters from the * original endpoint */ if (inp->sctp_ep.local_hmacs) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); inp->sctp_ep.local_hmacs = sctp_copy_hmaclist((*inp_p)->sctp_ep.local_hmacs); if (inp->sctp_ep.local_auth_chunks) sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); inp->sctp_ep.local_auth_chunks = sctp_copy_chunklist((*inp_p)->sctp_ep.local_auth_chunks); /* * Now we must move it from one hash table to * another and get the tcb in the right place. */ /* * This is where the one-2-one socket is put into * the accept state waiting for the accept! */ if (*stcb) { (*stcb)->asoc.state |= SCTP_STATE_IN_ACCEPT_QUEUE; } sctp_move_pcb_and_assoc(*inp_p, inp, *stcb); atomic_add_int(&(*stcb)->asoc.refcnt, 1); SCTP_TCB_UNLOCK((*stcb)); sctp_pull_off_control_to_new_inp((*inp_p), inp, *stcb, 0); SCTP_TCB_LOCK((*stcb)); atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); /* * now we must check to see if we were aborted while * the move was going on and the lock/unlock * happened. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* * yep it was, we leave the assoc attached * to the socket since the sctp_inpcb_free() * call will send an abort for us. */ SCTP_INP_DECR_REF(inp); return (NULL); } SCTP_INP_DECR_REF(inp); /* Switch over to the new guy */ *inp_p = inp; sctp_ulp_notify(notification, *stcb, 0, NULL, SCTP_SO_NOT_LOCKED); if (send_int_conf) { sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); } /* * Pull it from the incomplete queue and wake the * guy */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) atomic_add_int(&(*stcb)->asoc.refcnt, 1); SCTP_TCB_UNLOCK((*stcb)); SCTP_SOCKET_LOCK(so, 1); #endif soisconnected(so); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_TCB_LOCK((*stcb)); atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); SCTP_SOCKET_UNLOCK(so, 1); #endif return (m); } } if ((*inp_p)->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { if (notification) { sctp_ulp_notify(notification, *stcb, 0, NULL, SCTP_SO_NOT_LOCKED); } if (send_int_conf) { sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); } } return (m); } static void sctp_handle_cookie_ack(struct sctp_cookie_ack_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { /* cp must not be used, others call this without a c-ack :-) */ struct sctp_association *asoc; SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_cookie_ack: handling COOKIE-ACK\n"); if (stcb == NULL) return; asoc = &stcb->asoc; sctp_stop_all_cookie_timers(stcb); /* process according to association state */ if (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) { /* state change only needed when I am in right state */ SCTPDBG(SCTP_DEBUG_INPUT2, "moving to OPEN state\n"); SCTP_SET_STATE(asoc, SCTP_STATE_OPEN); if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } /* update RTO */ SCTP_STAT_INCR_COUNTER32(sctps_activeestab); SCTP_STAT_INCR_GAUGE32(sctps_currestab); if (asoc->overall_error_count == 0) { net->RTO = sctp_calculate_rto(stcb, asoc, net, &asoc->time_entered, sctp_align_safe_nocopy); } (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); sctp_ulp_notify(SCTP_NOTIFY_ASSOC_UP, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } #endif soisconnected(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); /* * since we did not send a HB make sure we don't double * things */ net->hb_responded = 1; if (stcb->asoc.sctp_autoclose_ticks && sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_AUTOCLOSE)) { sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, stcb->sctp_ep, stcb, NULL); } /* * send ASCONF if parameters are pending and ASCONFs are * allowed (eg. addresses changed when init/cookie echo were * in flight) */ if ((sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_DO_ASCONF)) && (stcb->asoc.peer_supports_asconf) && (!TAILQ_EMPTY(&stcb->asoc.asconf_queue))) { #ifdef SCTP_TIMER_BASED_ASCONF sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, stcb->sctp_ep, stcb, stcb->asoc.primary_destination); #else sctp_send_asconf(stcb, stcb->asoc.primary_destination, SCTP_ADDR_NOT_LOCKED); #endif } } /* Toss the cookie if I can */ sctp_toss_old_cookies(stcb, asoc); if (!TAILQ_EMPTY(&asoc->sent_queue)) { /* Restart the timer if we have pending data */ struct sctp_tmit_chunk *chk; chk = TAILQ_FIRST(&asoc->sent_queue); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); } } static void sctp_handle_ecn_echo(struct sctp_ecne_chunk *cp, struct sctp_tcb *stcb) { struct sctp_nets *net; struct sctp_tmit_chunk *lchk; struct sctp_ecne_chunk bkup; uint8_t override_bit = 0; uint32_t tsn, window_data_tsn; int len; int pkt_cnt; len = ntohs(cp->ch.chunk_length); if ((len != sizeof(struct sctp_ecne_chunk)) && (len != sizeof(struct old_sctp_ecne_chunk))) { return; } if (len == sizeof(struct old_sctp_ecne_chunk)) { /* Its the old format */ memcpy(&bkup, cp, sizeof(struct old_sctp_ecne_chunk)); bkup.num_pkts_since_cwr = htonl(1); cp = &bkup; } SCTP_STAT_INCR(sctps_recvecne); tsn = ntohl(cp->tsn); pkt_cnt = ntohl(cp->num_pkts_since_cwr); - /* ECN Nonce stuff: need a resync and disable the nonce sum check */ - /* Also we make sure we disable the nonce_wait */ lchk = TAILQ_LAST(&stcb->asoc.send_queue, sctpchunk_listhead); if (lchk == NULL) { - window_data_tsn = stcb->asoc.nonce_resync_tsn = stcb->asoc.sending_seq - 1; + window_data_tsn = stcb->asoc.sending_seq - 1; } else { - window_data_tsn = stcb->asoc.nonce_resync_tsn = lchk->rec.data.TSN_seq; + window_data_tsn = lchk->rec.data.TSN_seq; } - stcb->asoc.nonce_wait_for_ecne = 0; - stcb->asoc.nonce_sum_check = 0; /* Find where it was sent to if possible. */ net = NULL; TAILQ_FOREACH(lchk, &stcb->asoc.sent_queue, sctp_next) { if (lchk->rec.data.TSN_seq == tsn) { net = lchk->whoTo; break; } if (SCTP_TSN_GT(lchk->rec.data.TSN_seq, tsn)) { break; } } if (net == NULL) { /* * What to do. A previous send of a CWR was possibly lost. * See how old it is, we may have it marked on the actual * net. */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (tsn == net->last_cwr_tsn) { /* Found him, send it off */ goto out; } } /* * If we reach here, we need to send a special CWR that says * hey, we did this a long time ago and you lost the * response. */ net = TAILQ_FIRST(&stcb->asoc.nets); override_bit = SCTP_CWR_REDUCE_OVERRIDE; } out: - if (SCTP_TSN_GT(tsn, net->cwr_window_tsn)) { + if (SCTP_TSN_GT(tsn, net->cwr_window_tsn) && + ((override_bit & SCTP_CWR_REDUCE_OVERRIDE) == 0)) { /* * JRS - Use the congestion control given in the pluggable * CC module */ int ocwnd; ocwnd = net->cwnd; - stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo(stcb, net); + stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo(stcb, net, 0, pkt_cnt); /* * We reduce once every RTT. So we will only lower cwnd at * the next sending seq i.e. the window_data_tsn */ net->cwr_window_tsn = window_data_tsn; net->ecn_ce_pkt_cnt += pkt_cnt; net->lost_cnt = pkt_cnt; net->last_cwr_tsn = tsn; } else { override_bit |= SCTP_CWR_IN_SAME_WINDOW; - if (SCTP_TSN_GT(tsn, net->last_cwr_tsn)) { + if (SCTP_TSN_GT(tsn, net->last_cwr_tsn) && + ((override_bit & SCTP_CWR_REDUCE_OVERRIDE) == 0)) { /* - * Another loss in the same window update how man - * marks we have had + * Another loss in the same window update how many + * marks/packets lost we have had. */ + int cnt = 1; if (pkt_cnt > net->lost_cnt) { /* Should be the case */ - net->ecn_ce_pkt_cnt += (pkt_cnt - net->lost_cnt); - net->lost_cnt = pkt_cnt; + cnt = (pkt_cnt - net->lost_cnt); + net->ecn_ce_pkt_cnt += cnt; } + net->lost_cnt = pkt_cnt; net->last_cwr_tsn = tsn; + /* + * Most CC functions will ignore this call, since we + * are in-window yet of the initial CE the peer saw. + */ + stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo(stcb, net, 1, cnt); } } /* * We always send a CWR this way if our previous one was lost our * peer will get an update, or if it is not time again to reduce we * still get the cwr to the peer. Note we set the override when we * could not find the TSN on the chunk or the destination network. */ sctp_send_cwr(stcb, net, net->last_cwr_tsn, override_bit); } static void sctp_handle_ecn_cwr(struct sctp_cwr_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { /* * Here we get a CWR from the peer. We must look in the outqueue and * make sure that we have a covered ECNE in teh control chunk part. * If so remove it. */ struct sctp_tmit_chunk *chk; struct sctp_ecne_chunk *ecne; int override; uint32_t cwr_tsn; cwr_tsn = ntohl(cp->tsn); override = cp->ch.chunk_flags & SCTP_CWR_REDUCE_OVERRIDE; TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { if (chk->rec.chunk_id.id != SCTP_ECN_ECHO) { continue; } if ((override == 0) && (chk->whoTo != net)) { /* Must be from the right src unless override is set */ continue; } ecne = mtod(chk->data, struct sctp_ecne_chunk *); if (SCTP_TSN_GE(cwr_tsn, ntohl(ecne->tsn))) { /* this covers this ECNE, we can remove it */ stcb->asoc.ecn_echo_cnt_onq--; TAILQ_REMOVE(&stcb->asoc.control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } stcb->asoc.ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); if (override == 0) { break; } } } } static void sctp_handle_shutdown_complete(struct sctp_shutdown_complete_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_shutdown_complete: handling SHUTDOWN-COMPLETE\n"); if (stcb == NULL) return; asoc = &stcb->asoc; /* process according to association state */ if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT) { /* unexpected SHUTDOWN-COMPLETE... so ignore... */ SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_shutdown_complete: not in SCTP_STATE_SHUTDOWN_ACK_SENT --- ignore\n"); SCTP_TCB_UNLOCK(stcb); return; } /* notify upper layer protocol */ if (stcb->sctp_socket) { sctp_ulp_notify(SCTP_NOTIFY_ASSOC_DOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); /* are the queues empty? they should be */ if (!TAILQ_EMPTY(&asoc->send_queue) || !TAILQ_EMPTY(&asoc->sent_queue) || !stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { sctp_report_all_outbound(stcb, 0, SCTP_SO_NOT_LOCKED); } } /* stop the timer */ sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWNACK, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_22); SCTP_STAT_INCR_COUNTER32(sctps_shutdown); /* free the TCB */ SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_shutdown_complete: calls free-asoc\n"); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_23); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif return; } static int process_chunk_drop(struct sctp_tcb *stcb, struct sctp_chunk_desc *desc, struct sctp_nets *net, uint8_t flg) { switch (desc->chunk_type) { case SCTP_DATA: /* find the tsn to resend (possibly */ { uint32_t tsn; struct sctp_tmit_chunk *tp1; tsn = ntohl(desc->tsn_ifany); TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { if (tp1->rec.data.TSN_seq == tsn) { /* found it */ break; } if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, tsn)) { /* not found */ tp1 = NULL; break; } } if (tp1 == NULL) { /* * Do it the other way , aka without paying * attention to queue seq order. */ SCTP_STAT_INCR(sctps_pdrpdnfnd); TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { if (tp1->rec.data.TSN_seq == tsn) { /* found it */ break; } } } if (tp1 == NULL) { SCTP_STAT_INCR(sctps_pdrptsnnf); } if ((tp1) && (tp1->sent < SCTP_DATAGRAM_ACKED)) { uint8_t *ddp; if (((flg & SCTP_BADCRC) == 0) && ((flg & SCTP_FROM_MIDDLE_BOX) == 0)) { return (0); } if ((stcb->asoc.peers_rwnd == 0) && ((flg & SCTP_FROM_MIDDLE_BOX) == 0)) { SCTP_STAT_INCR(sctps_pdrpdiwnp); return (0); } if (stcb->asoc.peers_rwnd == 0 && (flg & SCTP_FROM_MIDDLE_BOX)) { SCTP_STAT_INCR(sctps_pdrpdizrw); return (0); } ddp = (uint8_t *) (mtod(tp1->data, caddr_t)+ sizeof(struct sctp_data_chunk)); { unsigned int iii; for (iii = 0; iii < sizeof(desc->data_bytes); iii++) { if (ddp[iii] != desc->data_bytes[iii]) { SCTP_STAT_INCR(sctps_pdrpbadd); return (-1); } } } - /* - * We zero out the nonce so resync not - * needed - */ - tp1->rec.data.ect_nonce = 0; if (tp1->do_rtt) { /* * this guy had a RTO calculation * pending on it, cancel it */ tp1->do_rtt = 0; } SCTP_STAT_INCR(sctps_pdrpmark); if (tp1->sent != SCTP_DATAGRAM_RESEND) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); /* * mark it as if we were doing a FR, since * we will be getting gap ack reports behind * the info from the router. */ tp1->rec.data.doing_fast_retransmit = 1; /* * mark the tsn with what sequences can * cause a new FR. */ if (TAILQ_EMPTY(&stcb->asoc.send_queue)) { tp1->rec.data.fast_retran_tsn = stcb->asoc.sending_seq; } else { tp1->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq; } /* restart the timer */ sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, tp1->whoTo, SCTP_FROM_SCTP_INPUT + SCTP_LOC_24); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, tp1->whoTo); /* fix counts and things */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PDRP, tp1->whoTo->flight_size, tp1->book_size, (uintptr_t) stcb, tp1->rec.data.TSN_seq); } if (tp1->sent < SCTP_DATAGRAM_RESEND) { sctp_flight_size_decrease(tp1); sctp_total_flight_decrease(stcb, tp1); } tp1->sent = SCTP_DATAGRAM_RESEND; } { /* audit code */ unsigned int audit; audit = 0; TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { if (tp1->sent == SCTP_DATAGRAM_RESEND) audit++; } TAILQ_FOREACH(tp1, &stcb->asoc.control_send_queue, sctp_next) { if (tp1->sent == SCTP_DATAGRAM_RESEND) audit++; } if (audit != stcb->asoc.sent_queue_retran_cnt) { SCTP_PRINTF("**Local Audit finds cnt:%d asoc cnt:%d\n", audit, stcb->asoc.sent_queue_retran_cnt); #ifndef SCTP_AUDITING_ENABLED stcb->asoc.sent_queue_retran_cnt = audit; #endif } } } break; case SCTP_ASCONF: { struct sctp_tmit_chunk *asconf; TAILQ_FOREACH(asconf, &stcb->asoc.control_send_queue, sctp_next) { if (asconf->rec.chunk_id.id == SCTP_ASCONF) { break; } } if (asconf) { if (asconf->sent != SCTP_DATAGRAM_RESEND) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); asconf->sent = SCTP_DATAGRAM_RESEND; asconf->snd_count--; } } break; case SCTP_INITIATION: /* resend the INIT */ stcb->asoc.dropped_special_cnt++; if (stcb->asoc.dropped_special_cnt < SCTP_RETRY_DROPPED_THRESH) { /* * If we can get it in, in a few attempts we do * this, otherwise we let the timer fire. */ sctp_timer_stop(SCTP_TIMER_TYPE_INIT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_25); sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); } break; case SCTP_SELECTIVE_ACK: case SCTP_NR_SELECTIVE_ACK: /* resend the sack */ sctp_send_sack(stcb); break; case SCTP_HEARTBEAT_REQUEST: /* resend a demand HB */ if ((stcb->asoc.overall_error_count + 3) < stcb->asoc.max_send_times) { /* * Only retransmit if we KNOW we wont destroy the * tcb */ (void)sctp_send_hb(stcb, 1, net); } break; case SCTP_SHUTDOWN: sctp_send_shutdown(stcb, net); break; case SCTP_SHUTDOWN_ACK: sctp_send_shutdown_ack(stcb, net); break; case SCTP_COOKIE_ECHO: { struct sctp_tmit_chunk *cookie; cookie = NULL; TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) { if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) { break; } } if (cookie) { if (cookie->sent != SCTP_DATAGRAM_RESEND) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); cookie->sent = SCTP_DATAGRAM_RESEND; sctp_stop_all_cookie_timers(stcb); } } break; case SCTP_COOKIE_ACK: sctp_send_cookie_ack(stcb); break; case SCTP_ASCONF_ACK: /* resend last asconf ack */ sctp_send_asconf_ack(stcb); break; case SCTP_FORWARD_CUM_TSN: send_forward_tsn(stcb, &stcb->asoc); break; /* can't do anything with these */ case SCTP_PACKET_DROPPED: case SCTP_INITIATION_ACK: /* this should not happen */ case SCTP_HEARTBEAT_ACK: case SCTP_ABORT_ASSOCIATION: case SCTP_OPERATION_ERROR: case SCTP_SHUTDOWN_COMPLETE: case SCTP_ECN_ECHO: case SCTP_ECN_CWR: default: break; } return (0); } void sctp_reset_in_stream(struct sctp_tcb *stcb, int number_entries, uint16_t * list) { int i; uint16_t temp; /* * We set things to 0xffff since this is the last delivered sequence * and we will be sending in 0 after the reset. */ if (number_entries) { for (i = 0; i < number_entries; i++) { temp = ntohs(list[i]); if (temp >= stcb->asoc.streamincnt) { continue; } stcb->asoc.strmin[temp].last_sequence_delivered = 0xffff; } } else { list = NULL; for (i = 0; i < stcb->asoc.streamincnt; i++) { stcb->asoc.strmin[i].last_sequence_delivered = 0xffff; } } sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_RECV, stcb, number_entries, (void *)list, SCTP_SO_NOT_LOCKED); } static void sctp_reset_out_streams(struct sctp_tcb *stcb, int number_entries, uint16_t * list) { int i; if (number_entries == 0) { for (i = 0; i < stcb->asoc.streamoutcnt; i++) { stcb->asoc.strmout[i].next_sequence_sent = 0; } } else if (number_entries) { for (i = 0; i < number_entries; i++) { uint16_t temp; temp = ntohs(list[i]); if (temp >= stcb->asoc.streamoutcnt) { /* no such stream */ continue; } stcb->asoc.strmout[temp].next_sequence_sent = 0; } } sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_SEND, stcb, number_entries, (void *)list, SCTP_SO_NOT_LOCKED); } struct sctp_stream_reset_out_request * sctp_find_stream_reset(struct sctp_tcb *stcb, uint32_t seq, struct sctp_tmit_chunk **bchk) { struct sctp_association *asoc; struct sctp_stream_reset_out_req *req; struct sctp_stream_reset_out_request *r; struct sctp_tmit_chunk *chk; int len, clen; asoc = &stcb->asoc; if (TAILQ_EMPTY(&stcb->asoc.control_send_queue)) { asoc->stream_reset_outstanding = 0; return (NULL); } if (stcb->asoc.str_reset == NULL) { asoc->stream_reset_outstanding = 0; return (NULL); } chk = stcb->asoc.str_reset; if (chk->data == NULL) { return (NULL); } if (bchk) { /* he wants a copy of the chk pointer */ *bchk = chk; } clen = chk->send_size; req = mtod(chk->data, struct sctp_stream_reset_out_req *); r = &req->sr_req; if (ntohl(r->request_seq) == seq) { /* found it */ return (r); } len = SCTP_SIZE32(ntohs(r->ph.param_length)); if (clen > (len + (int)sizeof(struct sctp_chunkhdr))) { /* move to the next one, there can only be a max of two */ r = (struct sctp_stream_reset_out_request *)((caddr_t)r + len); if (ntohl(r->request_seq) == seq) { return (r); } } /* that seq is not here */ return (NULL); } static void sctp_clean_up_stream_reset(struct sctp_tcb *stcb) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk = stcb->asoc.str_reset; if (stcb->asoc.str_reset == NULL) { return; } asoc = &stcb->asoc; sctp_timer_stop(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo, SCTP_FROM_SCTP_INPUT + SCTP_LOC_26); TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); /* sa_ignore NO_NULL_CHK */ stcb->asoc.str_reset = NULL; } static int sctp_handle_stream_reset_response(struct sctp_tcb *stcb, uint32_t seq, uint32_t action, struct sctp_stream_reset_response *respin) { uint16_t type; int lparm_len; struct sctp_association *asoc = &stcb->asoc; struct sctp_tmit_chunk *chk; struct sctp_stream_reset_out_request *srparam; int number_entries; if (asoc->stream_reset_outstanding == 0) { /* duplicate */ return (0); } if (seq == stcb->asoc.str_reset_seq_out) { srparam = sctp_find_stream_reset(stcb, seq, &chk); if (srparam) { stcb->asoc.str_reset_seq_out++; type = ntohs(srparam->ph.param_type); lparm_len = ntohs(srparam->ph.param_length); if (type == SCTP_STR_RESET_OUT_REQUEST) { number_entries = (lparm_len - sizeof(struct sctp_stream_reset_out_request)) / sizeof(uint16_t); asoc->stream_reset_out_is_outstanding = 0; if (asoc->stream_reset_outstanding) asoc->stream_reset_outstanding--; if (action == SCTP_STREAM_RESET_PERFORMED) { /* do it */ sctp_reset_out_streams(stcb, number_entries, srparam->list_of_streams); } else { sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_FAILED_OUT, stcb, number_entries, srparam->list_of_streams, SCTP_SO_NOT_LOCKED); } } else if (type == SCTP_STR_RESET_IN_REQUEST) { /* Answered my request */ number_entries = (lparm_len - sizeof(struct sctp_stream_reset_in_request)) / sizeof(uint16_t); if (asoc->stream_reset_outstanding) asoc->stream_reset_outstanding--; if (action != SCTP_STREAM_RESET_PERFORMED) { sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_FAILED_IN, stcb, number_entries, srparam->list_of_streams, SCTP_SO_NOT_LOCKED); } } else if (type == SCTP_STR_RESET_ADD_STREAMS) { /* Ok we now may have more streams */ if (asoc->stream_reset_outstanding) asoc->stream_reset_outstanding--; if (action == SCTP_STREAM_RESET_PERFORMED) { /* Put the new streams into effect */ stcb->asoc.streamoutcnt = stcb->asoc.strm_realoutsize; sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_ADD_OK, stcb, (uint32_t) stcb->asoc.streamoutcnt, NULL, SCTP_SO_NOT_LOCKED); } else { sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_ADD_FAIL, stcb, (uint32_t) stcb->asoc.streamoutcnt, NULL, SCTP_SO_NOT_LOCKED); } } else if (type == SCTP_STR_RESET_TSN_REQUEST) { /** * a) Adopt the new in tsn. * b) reset the map * c) Adopt the new out-tsn */ struct sctp_stream_reset_response_tsn *resp; struct sctp_forward_tsn_chunk fwdtsn; int abort_flag = 0; if (respin == NULL) { /* huh ? */ return (0); } if (action == SCTP_STREAM_RESET_PERFORMED) { resp = (struct sctp_stream_reset_response_tsn *)respin; asoc->stream_reset_outstanding--; fwdtsn.ch.chunk_length = htons(sizeof(struct sctp_forward_tsn_chunk)); fwdtsn.ch.chunk_type = SCTP_FORWARD_CUM_TSN; fwdtsn.new_cumulative_tsn = htonl(ntohl(resp->senders_next_tsn) - 1); sctp_handle_forward_tsn(stcb, &fwdtsn, &abort_flag, NULL, 0); if (abort_flag) { return (1); } stcb->asoc.highest_tsn_inside_map = (ntohl(resp->senders_next_tsn) - 1); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 7, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } stcb->asoc.tsn_last_delivered = stcb->asoc.cumulative_tsn = stcb->asoc.highest_tsn_inside_map; stcb->asoc.mapping_array_base_tsn = ntohl(resp->senders_next_tsn); memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size); stcb->asoc.highest_tsn_inside_nr_map = stcb->asoc.highest_tsn_inside_map; memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size); stcb->asoc.sending_seq = ntohl(resp->receivers_next_tsn); stcb->asoc.last_acked_seq = stcb->asoc.cumulative_tsn; sctp_reset_out_streams(stcb, 0, (uint16_t *) NULL); sctp_reset_in_stream(stcb, 0, (uint16_t *) NULL); } } /* get rid of the request and get the request flags */ if (asoc->stream_reset_outstanding == 0) { sctp_clean_up_stream_reset(stcb); } } } return (0); } static void sctp_handle_str_reset_request_in(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, struct sctp_stream_reset_in_request *req, int trunc) { uint32_t seq; int len, i; int number_entries; uint16_t temp; /* * peer wants me to send a str-reset to him for my outgoing seq's if * seq_in is right. */ struct sctp_association *asoc = &stcb->asoc; seq = ntohl(req->request_seq); if (asoc->str_reset_seq_in == seq) { if (trunc) { /* Can't do it, since they exceeded our buffer size */ asoc->last_reset_action[1] = asoc->last_reset_action[0]; asoc->last_reset_action[0] = SCTP_STREAM_RESET_DENIED; sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); } else if (stcb->asoc.stream_reset_out_is_outstanding == 0) { len = ntohs(req->ph.param_length); number_entries = ((len - sizeof(struct sctp_stream_reset_in_request)) / sizeof(uint16_t)); for (i = 0; i < number_entries; i++) { temp = ntohs(req->list_of_streams[i]); req->list_of_streams[i] = temp; } /* move the reset action back one */ asoc->last_reset_action[1] = asoc->last_reset_action[0]; asoc->last_reset_action[0] = SCTP_STREAM_RESET_PERFORMED; sctp_add_stream_reset_out(chk, number_entries, req->list_of_streams, asoc->str_reset_seq_out, seq, (asoc->sending_seq - 1)); asoc->stream_reset_out_is_outstanding = 1; asoc->str_reset = chk; sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo); stcb->asoc.stream_reset_outstanding++; } else { /* Can't do it, since we have sent one out */ asoc->last_reset_action[1] = asoc->last_reset_action[0]; asoc->last_reset_action[0] = SCTP_STREAM_RESET_TRY_LATER; sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); } asoc->str_reset_seq_in++; } else if (asoc->str_reset_seq_in - 1 == seq) { sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); } else if (asoc->str_reset_seq_in - 2 == seq) { sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); } else { sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_BAD_SEQNO); } } static int sctp_handle_str_reset_request_tsn(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, struct sctp_stream_reset_tsn_request *req) { /* reset all in and out and update the tsn */ /* * A) reset my str-seq's on in and out. B) Select a receive next, * and set cum-ack to it. Also process this selected number as a * fwd-tsn as well. C) set in the response my next sending seq. */ struct sctp_forward_tsn_chunk fwdtsn; struct sctp_association *asoc = &stcb->asoc; int abort_flag = 0; uint32_t seq; seq = ntohl(req->request_seq); if (asoc->str_reset_seq_in == seq) { fwdtsn.ch.chunk_length = htons(sizeof(struct sctp_forward_tsn_chunk)); fwdtsn.ch.chunk_type = SCTP_FORWARD_CUM_TSN; fwdtsn.ch.chunk_flags = 0; fwdtsn.new_cumulative_tsn = htonl(stcb->asoc.highest_tsn_inside_map + 1); sctp_handle_forward_tsn(stcb, &fwdtsn, &abort_flag, NULL, 0); if (abort_flag) { return (1); } stcb->asoc.highest_tsn_inside_map += SCTP_STREAM_RESET_TSN_DELTA; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 10, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } stcb->asoc.tsn_last_delivered = stcb->asoc.cumulative_tsn = stcb->asoc.highest_tsn_inside_map; stcb->asoc.mapping_array_base_tsn = stcb->asoc.highest_tsn_inside_map + 1; memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size); stcb->asoc.highest_tsn_inside_nr_map = stcb->asoc.highest_tsn_inside_map; memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size); atomic_add_int(&stcb->asoc.sending_seq, 1); /* save off historical data for retrans */ stcb->asoc.last_sending_seq[1] = stcb->asoc.last_sending_seq[0]; stcb->asoc.last_sending_seq[0] = stcb->asoc.sending_seq; stcb->asoc.last_base_tsnsent[1] = stcb->asoc.last_base_tsnsent[0]; stcb->asoc.last_base_tsnsent[0] = stcb->asoc.mapping_array_base_tsn; sctp_add_stream_reset_result_tsn(chk, ntohl(req->request_seq), SCTP_STREAM_RESET_PERFORMED, stcb->asoc.sending_seq, stcb->asoc.mapping_array_base_tsn); sctp_reset_out_streams(stcb, 0, (uint16_t *) NULL); sctp_reset_in_stream(stcb, 0, (uint16_t *) NULL); stcb->asoc.last_reset_action[1] = stcb->asoc.last_reset_action[0]; stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_PERFORMED; asoc->str_reset_seq_in++; } else if (asoc->str_reset_seq_in - 1 == seq) { sctp_add_stream_reset_result_tsn(chk, seq, asoc->last_reset_action[0], stcb->asoc.last_sending_seq[0], stcb->asoc.last_base_tsnsent[0] ); } else if (asoc->str_reset_seq_in - 2 == seq) { sctp_add_stream_reset_result_tsn(chk, seq, asoc->last_reset_action[1], stcb->asoc.last_sending_seq[1], stcb->asoc.last_base_tsnsent[1] ); } else { sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_BAD_SEQNO); } return (0); } static void sctp_handle_str_reset_request_out(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, struct sctp_stream_reset_out_request *req, int trunc) { uint32_t seq, tsn; int number_entries, len; struct sctp_association *asoc = &stcb->asoc; seq = ntohl(req->request_seq); /* now if its not a duplicate we process it */ if (asoc->str_reset_seq_in == seq) { len = ntohs(req->ph.param_length); number_entries = ((len - sizeof(struct sctp_stream_reset_out_request)) / sizeof(uint16_t)); /* * the sender is resetting, handle the list issue.. we must * a) verify if we can do the reset, if so no problem b) If * we can't do the reset we must copy the request. c) queue * it, and setup the data in processor to trigger it off * when needed and dequeue all the queued data. */ tsn = ntohl(req->send_reset_at_tsn); /* move the reset action back one */ asoc->last_reset_action[1] = asoc->last_reset_action[0]; if (trunc) { sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_DENIED); asoc->last_reset_action[0] = SCTP_STREAM_RESET_DENIED; } else if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) { /* we can do it now */ sctp_reset_in_stream(stcb, number_entries, req->list_of_streams); sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_PERFORMED); asoc->last_reset_action[0] = SCTP_STREAM_RESET_PERFORMED; } else { /* * we must queue it up and thus wait for the TSN's * to arrive that are at or before tsn */ struct sctp_stream_reset_list *liste; int siz; siz = sizeof(struct sctp_stream_reset_list) + (number_entries * sizeof(uint16_t)); SCTP_MALLOC(liste, struct sctp_stream_reset_list *, siz, SCTP_M_STRESET); if (liste == NULL) { /* gak out of memory */ sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_DENIED); asoc->last_reset_action[0] = SCTP_STREAM_RESET_DENIED; return; } liste->tsn = tsn; liste->number_entries = number_entries; memcpy(&liste->req, req, (sizeof(struct sctp_stream_reset_out_request) + (number_entries * sizeof(uint16_t)))); TAILQ_INSERT_TAIL(&asoc->resetHead, liste, next_resp); sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_PERFORMED); asoc->last_reset_action[0] = SCTP_STREAM_RESET_PERFORMED; } asoc->str_reset_seq_in++; } else if ((asoc->str_reset_seq_in - 1) == seq) { /* * one seq back, just echo back last action since my * response was lost. */ sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); } else if ((asoc->str_reset_seq_in - 2) == seq) { /* * two seq back, just echo back last action since my * response was lost. */ sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); } else { sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_BAD_SEQNO); } } static void sctp_handle_str_reset_add_strm(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, struct sctp_stream_reset_add_strm *str_add) { /* * Peer is requesting to add more streams. If its within our * max-streams we will allow it. */ uint16_t num_stream, i; uint32_t seq; struct sctp_association *asoc = &stcb->asoc; struct sctp_queued_to_read *ctl, *nctl; /* Get the number. */ seq = ntohl(str_add->request_seq); num_stream = ntohs(str_add->number_of_streams); /* Now what would be the new total? */ if (asoc->str_reset_seq_in == seq) { num_stream += stcb->asoc.streamincnt; if (num_stream > stcb->asoc.max_inbound_streams) { /* We must reject it they ask for to many */ denied: sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_DENIED); stcb->asoc.last_reset_action[1] = stcb->asoc.last_reset_action[0]; stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_DENIED; } else { /* Ok, we can do that :-) */ struct sctp_stream_in *oldstrm; /* save off the old */ oldstrm = stcb->asoc.strmin; SCTP_MALLOC(stcb->asoc.strmin, struct sctp_stream_in *, (num_stream * sizeof(struct sctp_stream_in)), SCTP_M_STRMI); if (stcb->asoc.strmin == NULL) { stcb->asoc.strmin = oldstrm; goto denied; } /* copy off the old data */ for (i = 0; i < stcb->asoc.streamincnt; i++) { TAILQ_INIT(&stcb->asoc.strmin[i].inqueue); stcb->asoc.strmin[i].stream_no = i; stcb->asoc.strmin[i].last_sequence_delivered = oldstrm[i].last_sequence_delivered; stcb->asoc.strmin[i].delivery_started = oldstrm[i].delivery_started; /* now anything on those queues? */ TAILQ_FOREACH_SAFE(ctl, &oldstrm[i].inqueue, next, nctl) { TAILQ_REMOVE(&oldstrm[i].inqueue, ctl, next); TAILQ_INSERT_TAIL(&stcb->asoc.strmin[i].inqueue, ctl, next); } } /* Init the new streams */ for (i = stcb->asoc.streamincnt; i < num_stream; i++) { TAILQ_INIT(&stcb->asoc.strmin[i].inqueue); stcb->asoc.strmin[i].stream_no = i; stcb->asoc.strmin[i].last_sequence_delivered = 0xffff; stcb->asoc.strmin[i].delivery_started = 0; } SCTP_FREE(oldstrm, SCTP_M_STRMI); /* update the size */ stcb->asoc.streamincnt = num_stream; /* Send the ack */ sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_PERFORMED); stcb->asoc.last_reset_action[1] = stcb->asoc.last_reset_action[0]; stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_PERFORMED; sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_INSTREAM_ADD_OK, stcb, (uint32_t) stcb->asoc.streamincnt, NULL, SCTP_SO_NOT_LOCKED); } } else if ((asoc->str_reset_seq_in - 1) == seq) { /* * one seq back, just echo back last action since my * response was lost. */ sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); } else if ((asoc->str_reset_seq_in - 2) == seq) { /* * two seq back, just echo back last action since my * response was lost. */ sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); } else { sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_BAD_SEQNO); } } #ifdef __GNUC__ __attribute__((noinline)) #endif static int sctp_handle_stream_reset(struct sctp_tcb *stcb, struct mbuf *m, int offset, struct sctp_stream_reset_out_req *sr_req) { int chk_length, param_len, ptype; struct sctp_paramhdr pstore; uint8_t cstore[SCTP_CHUNK_BUFFER_SIZE]; uint32_t seq; int num_req = 0; int trunc = 0; struct sctp_tmit_chunk *chk; struct sctp_chunkhdr *ch; struct sctp_paramhdr *ph; int ret_code = 0; int num_param = 0; /* now it may be a reset or a reset-response */ chk_length = ntohs(sr_req->ch.chunk_length); /* setup for adding the response */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return (ret_code); } chk->rec.chunk_id.id = SCTP_STREAM_RESET; chk->rec.chunk_id.can_take_data = 0; chk->asoc = &stcb->asoc; chk->no_fr_allowed = 0; chk->book_size = chk->send_size = sizeof(struct sctp_chunkhdr); chk->book_size_scale = 0; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { strres_nochunk: if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); return (ret_code); } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); /* setup chunk parameters */ chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = stcb->asoc.primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); ch = mtod(chk->data, struct sctp_chunkhdr *); ch->chunk_type = SCTP_STREAM_RESET; ch->chunk_flags = 0; ch->chunk_length = htons(chk->send_size); SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); offset += sizeof(struct sctp_chunkhdr); while ((size_t)chk_length >= sizeof(struct sctp_stream_reset_tsn_request)) { ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, sizeof(pstore), (uint8_t *) & pstore); if (ph == NULL) break; param_len = ntohs(ph->param_length); if (param_len < (int)sizeof(struct sctp_stream_reset_tsn_request)) { /* bad param */ break; } ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, min(param_len, (int)sizeof(cstore)), (uint8_t *) & cstore); ptype = ntohs(ph->param_type); num_param++; if (param_len > (int)sizeof(cstore)) { trunc = 1; } else { trunc = 0; } if (num_param > SCTP_MAX_RESET_PARAMS) { /* hit the max of parameters already sorry.. */ break; } if (ptype == SCTP_STR_RESET_OUT_REQUEST) { struct sctp_stream_reset_out_request *req_out; req_out = (struct sctp_stream_reset_out_request *)ph; num_req++; if (stcb->asoc.stream_reset_outstanding) { seq = ntohl(req_out->response_seq); if (seq == stcb->asoc.str_reset_seq_out) { /* implicit ack */ (void)sctp_handle_stream_reset_response(stcb, seq, SCTP_STREAM_RESET_PERFORMED, NULL); } } sctp_handle_str_reset_request_out(stcb, chk, req_out, trunc); } else if (ptype == SCTP_STR_RESET_ADD_STREAMS) { struct sctp_stream_reset_add_strm *str_add; str_add = (struct sctp_stream_reset_add_strm *)ph; num_req++; sctp_handle_str_reset_add_strm(stcb, chk, str_add); } else if (ptype == SCTP_STR_RESET_IN_REQUEST) { struct sctp_stream_reset_in_request *req_in; num_req++; req_in = (struct sctp_stream_reset_in_request *)ph; sctp_handle_str_reset_request_in(stcb, chk, req_in, trunc); } else if (ptype == SCTP_STR_RESET_TSN_REQUEST) { struct sctp_stream_reset_tsn_request *req_tsn; num_req++; req_tsn = (struct sctp_stream_reset_tsn_request *)ph; if (sctp_handle_str_reset_request_tsn(stcb, chk, req_tsn)) { ret_code = 1; goto strres_nochunk; } /* no more */ break; } else if (ptype == SCTP_STR_RESET_RESPONSE) { struct sctp_stream_reset_response *resp; uint32_t result; resp = (struct sctp_stream_reset_response *)ph; seq = ntohl(resp->response_seq); result = ntohl(resp->result); if (sctp_handle_stream_reset_response(stcb, seq, result, resp)) { ret_code = 1; goto strres_nochunk; } } else { break; } offset += SCTP_SIZE32(param_len); chk_length -= SCTP_SIZE32(param_len); } if (num_req == 0) { /* we have no response free the stuff */ goto strres_nochunk; } /* ok we have a chunk to link in */ TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); stcb->asoc.ctrl_queue_cnt++; return (ret_code); } /* * Handle a router or endpoints report of a packet loss, there are two ways * to handle this, either we get the whole packet and must disect it * ourselves (possibly with truncation and or corruption) or it is a summary * from a middle box that did the disectting for us. */ static void sctp_handle_packet_dropped(struct sctp_pktdrop_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t limit) { uint32_t bottle_bw, on_queue; uint16_t trunc_len; unsigned int chlen; unsigned int at; struct sctp_chunk_desc desc; struct sctp_chunkhdr *ch; chlen = ntohs(cp->ch.chunk_length); chlen -= sizeof(struct sctp_pktdrop_chunk); /* XXX possible chlen underflow */ if (chlen == 0) { ch = NULL; if (cp->ch.chunk_flags & SCTP_FROM_MIDDLE_BOX) SCTP_STAT_INCR(sctps_pdrpbwrpt); } else { ch = (struct sctp_chunkhdr *)(cp->data + sizeof(struct sctphdr)); chlen -= sizeof(struct sctphdr); /* XXX possible chlen underflow */ memset(&desc, 0, sizeof(desc)); } trunc_len = (uint16_t) ntohs(cp->trunc_len); if (trunc_len > limit) { trunc_len = limit; } /* now the chunks themselves */ while ((ch != NULL) && (chlen >= sizeof(struct sctp_chunkhdr))) { desc.chunk_type = ch->chunk_type; /* get amount we need to move */ at = ntohs(ch->chunk_length); if (at < sizeof(struct sctp_chunkhdr)) { /* corrupt chunk, maybe at the end? */ SCTP_STAT_INCR(sctps_pdrpcrupt); break; } if (trunc_len == 0) { /* we are supposed to have all of it */ if (at > chlen) { /* corrupt skip it */ SCTP_STAT_INCR(sctps_pdrpcrupt); break; } } else { /* is there enough of it left ? */ if (desc.chunk_type == SCTP_DATA) { if (chlen < (sizeof(struct sctp_data_chunk) + sizeof(desc.data_bytes))) { break; } } else { if (chlen < sizeof(struct sctp_chunkhdr)) { break; } } } if (desc.chunk_type == SCTP_DATA) { /* can we get out the tsn? */ if ((cp->ch.chunk_flags & SCTP_FROM_MIDDLE_BOX)) SCTP_STAT_INCR(sctps_pdrpmbda); if (chlen >= (sizeof(struct sctp_data_chunk) + sizeof(uint32_t))) { /* yep */ struct sctp_data_chunk *dcp; uint8_t *ddp; unsigned int iii; dcp = (struct sctp_data_chunk *)ch; ddp = (uint8_t *) (dcp + 1); for (iii = 0; iii < sizeof(desc.data_bytes); iii++) { desc.data_bytes[iii] = ddp[iii]; } desc.tsn_ifany = dcp->dp.tsn; } else { /* nope we are done. */ SCTP_STAT_INCR(sctps_pdrpnedat); break; } } else { if ((cp->ch.chunk_flags & SCTP_FROM_MIDDLE_BOX)) SCTP_STAT_INCR(sctps_pdrpmbct); } if (process_chunk_drop(stcb, &desc, net, cp->ch.chunk_flags)) { SCTP_STAT_INCR(sctps_pdrppdbrk); break; } if (SCTP_SIZE32(at) > chlen) { break; } chlen -= SCTP_SIZE32(at); if (chlen < sizeof(struct sctp_chunkhdr)) { /* done, none left */ break; } ch = (struct sctp_chunkhdr *)((caddr_t)ch + SCTP_SIZE32(at)); } /* Now update any rwnd --- possibly */ if ((cp->ch.chunk_flags & SCTP_FROM_MIDDLE_BOX) == 0) { /* From a peer, we get a rwnd report */ uint32_t a_rwnd; SCTP_STAT_INCR(sctps_pdrpfehos); bottle_bw = ntohl(cp->bottle_bw); on_queue = ntohl(cp->current_onq); if (bottle_bw && on_queue) { /* a rwnd report is in here */ if (bottle_bw > on_queue) a_rwnd = bottle_bw - on_queue; else a_rwnd = 0; if (a_rwnd == 0) stcb->asoc.peers_rwnd = 0; else { if (a_rwnd > stcb->asoc.total_flight) { stcb->asoc.peers_rwnd = a_rwnd - stcb->asoc.total_flight; } else { stcb->asoc.peers_rwnd = 0; } if (stcb->asoc.peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ stcb->asoc.peers_rwnd = 0; } } } } else { SCTP_STAT_INCR(sctps_pdrpfmbox); } /* now middle boxes in sat networks get a cwnd bump */ if ((cp->ch.chunk_flags & SCTP_FROM_MIDDLE_BOX) && (stcb->asoc.sat_t3_loss_recovery == 0) && (stcb->asoc.sat_network)) { /* * This is debateable but for sat networks it makes sense * Note if a T3 timer has went off, we will prohibit any * changes to cwnd until we exit the t3 loss recovery. */ stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped(stcb, net, cp, &bottle_bw, &on_queue); } } /* * handles all control chunks in a packet inputs: - m: mbuf chain, assumed to * still contain IP/SCTP header - stcb: is the tcb found for this packet - * offset: offset into the mbuf chain to first chunkhdr - length: is the * length of the complete packet outputs: - length: modified to remaining * length after control processing - netp: modified to new sctp_nets after * cookie-echo processing - return NULL to discard the packet (ie. no asoc, * bad packet,...) otherwise return the tcb for this packet */ #ifdef __GNUC__ __attribute__((noinline)) #endif static struct sctp_tcb * sctp_process_control(struct mbuf *m, int iphlen, int *offset, int length, struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets **netp, int *fwd_tsn_seen, uint32_t vrf_id, uint16_t port) { struct sctp_association *asoc; uint32_t vtag_in; int num_chunks = 0; /* number of control chunks processed */ uint32_t chk_length; int ret; int abort_no_unlock = 0; /* * How big should this be, and should it be alloc'd? Lets try the * d-mtu-ceiling for now (2k) and that should hopefully work ... * until we get into jumbo grams and such.. */ uint8_t chunk_buf[SCTP_CHUNK_BUFFER_SIZE]; struct sctp_tcb *locked_tcb = stcb; int got_auth = 0; uint32_t auth_offset = 0, auth_len = 0; int auth_skipped = 0; int asconf_cnt = 0; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_control: iphlen=%u, offset=%u, length=%u stcb:%p\n", iphlen, *offset, length, stcb); /* validate chunk header length... */ if (ntohs(ch->chunk_length) < sizeof(*ch)) { SCTPDBG(SCTP_DEBUG_INPUT1, "Invalid header length %d\n", ntohs(ch->chunk_length)); if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } /* * validate the verification tag */ vtag_in = ntohl(sh->v_tag); if (locked_tcb) { SCTP_TCB_LOCK_ASSERT(locked_tcb); } if (ch->chunk_type == SCTP_INITIATION) { SCTPDBG(SCTP_DEBUG_INPUT1, "Its an INIT of len:%d vtag:%x\n", ntohs(ch->chunk_length), vtag_in); if (vtag_in != 0) { /* protocol error- silently discard... */ SCTP_STAT_INCR(sctps_badvtag); if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } } else if (ch->chunk_type != SCTP_COOKIE_ECHO) { /* * If there is no stcb, skip the AUTH chunk and process * later after a stcb is found (to validate the lookup was * valid. */ if ((ch->chunk_type == SCTP_AUTHENTICATION) && (stcb == NULL) && !SCTP_BASE_SYSCTL(sctp_auth_disable)) { /* save this chunk for later processing */ auth_skipped = 1; auth_offset = *offset; auth_len = ntohs(ch->chunk_length); /* (temporarily) move past this chunk */ *offset += SCTP_SIZE32(auth_len); if (*offset >= length) { /* no more data left in the mbuf chain */ *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, sizeof(struct sctp_chunkhdr), chunk_buf); } if (ch == NULL) { /* Help */ *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if (ch->chunk_type == SCTP_COOKIE_ECHO) { goto process_control_chunks; } /* * first check if it's an ASCONF with an unknown src addr we * need to look inside to find the association */ if (ch->chunk_type == SCTP_ASCONF && stcb == NULL) { struct sctp_chunkhdr *asconf_ch = ch; uint32_t asconf_offset = 0, asconf_len = 0; /* inp's refcount may be reduced */ SCTP_INP_INCR_REF(inp); asconf_offset = *offset; do { asconf_len = ntohs(asconf_ch->chunk_length); if (asconf_len < sizeof(struct sctp_asconf_paramhdr)) break; stcb = sctp_findassociation_ep_asconf(m, iphlen, *offset, sh, &inp, netp, vrf_id); if (stcb != NULL) break; asconf_offset += SCTP_SIZE32(asconf_len); asconf_ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, asconf_offset, sizeof(struct sctp_chunkhdr), chunk_buf); } while (asconf_ch != NULL && asconf_ch->chunk_type == SCTP_ASCONF); if (stcb == NULL) { /* * reduce inp's refcount if not reduced in * sctp_findassociation_ep_asconf(). */ SCTP_INP_DECR_REF(inp); } else { locked_tcb = stcb; } /* now go back and verify any auth chunk to be sure */ if (auth_skipped && (stcb != NULL)) { struct sctp_auth_chunk *auth; auth = (struct sctp_auth_chunk *) sctp_m_getptr(m, auth_offset, auth_len, chunk_buf); got_auth = 1; auth_skipped = 0; if ((auth == NULL) || sctp_handle_auth(stcb, auth, m, auth_offset)) { /* auth HMAC failed so dump it */ *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } else { /* remaining chunks are HMAC checked */ stcb->asoc.authenticated = 1; } } } if (stcb == NULL) { /* no association, so it's out of the blue... */ sctp_handle_ootb(m, iphlen, *offset, sh, inp, NULL, vrf_id, port); *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } asoc = &stcb->asoc; /* ABORT and SHUTDOWN can use either v_tag... */ if ((ch->chunk_type == SCTP_ABORT_ASSOCIATION) || (ch->chunk_type == SCTP_SHUTDOWN_COMPLETE) || (ch->chunk_type == SCTP_PACKET_DROPPED)) { if ((vtag_in == asoc->my_vtag) || ((ch->chunk_flags & SCTP_HAD_NO_TCB) && (vtag_in == asoc->peer_vtag))) { /* this is valid */ } else { /* drop this packet... */ SCTP_STAT_INCR(sctps_badvtag); if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } } else if (ch->chunk_type == SCTP_SHUTDOWN_ACK) { if (vtag_in != asoc->my_vtag) { /* * this could be a stale SHUTDOWN-ACK or the * peer never got the SHUTDOWN-COMPLETE and * is still hung; we have started a new asoc * but it won't complete until the shutdown * is completed */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } sctp_handle_ootb(m, iphlen, *offset, sh, inp, NULL, vrf_id, port); return (NULL); } } else { /* for all other chunks, vtag must match */ if (vtag_in != asoc->my_vtag) { /* invalid vtag... */ SCTPDBG(SCTP_DEBUG_INPUT3, "invalid vtag: %xh, expect %xh\n", vtag_in, asoc->my_vtag); SCTP_STAT_INCR(sctps_badvtag); if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } } } /* end if !SCTP_COOKIE_ECHO */ /* * process all control chunks... */ if (((ch->chunk_type == SCTP_SELECTIVE_ACK) || /* EY */ (ch->chunk_type == SCTP_NR_SELECTIVE_ACK) || (ch->chunk_type == SCTP_HEARTBEAT_REQUEST)) && (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED)) { /* implied cookie-ack.. we must have lost the ack */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, *netp); } process_control_chunks: while (IS_SCTP_CONTROL(ch)) { /* validate chunk length */ chk_length = ntohs(ch->chunk_length); SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_process_control: processing a chunk type=%u, len=%u\n", ch->chunk_type, chk_length); SCTP_LTRACE_CHK(inp, stcb, ch->chunk_type, chk_length); if (chk_length < sizeof(*ch) || (*offset + (int)chk_length) > length) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } SCTP_STAT_INCR_COUNTER64(sctps_incontrolchunks); /* * INIT-ACK only gets the init ack "header" portion only * because we don't have to process the peer's COOKIE. All * others get a complete chunk. */ if ((ch->chunk_type == SCTP_INITIATION_ACK) || (ch->chunk_type == SCTP_INITIATION)) { /* get an init-ack chunk */ ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, sizeof(struct sctp_init_ack_chunk), chunk_buf); if (ch == NULL) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } } else { /* For cookies and all other chunks. */ if (chk_length > sizeof(chunk_buf)) { /* * use just the size of the chunk buffer so * the front part of our chunks fit in * contiguous space up to the chunk buffer * size (508 bytes). For chunks that need to * get more than that they must use the * sctp_m_getptr() function or other means * (e.g. know how to parse mbuf chains). * Cookies do this already. */ ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, (sizeof(chunk_buf) - 4), chunk_buf); if (ch == NULL) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } } else { /* We can fit it all */ ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, chk_length, chunk_buf); if (ch == NULL) { SCTP_PRINTF("sctp_process_control: Can't get the all data....\n"); *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } } } num_chunks++; /* Save off the last place we got a control from */ if (stcb != NULL) { if (((netp != NULL) && (*netp != NULL)) || (ch->chunk_type == SCTP_ASCONF)) { /* * allow last_control to be NULL if * ASCONF... ASCONF processing will find the * right net later */ if ((netp != NULL) && (*netp != NULL)) stcb->asoc.last_control_chunk_from = *netp; } } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xB0, ch->chunk_type); #endif /* check to see if this chunk required auth, but isn't */ if ((stcb != NULL) && !SCTP_BASE_SYSCTL(sctp_auth_disable) && sctp_auth_is_required_chunk(ch->chunk_type, stcb->asoc.local_auth_chunks) && !stcb->asoc.authenticated) { /* "silently" ignore */ SCTP_STAT_INCR(sctps_recvauthmissing); goto next_chunk; } switch (ch->chunk_type) { case SCTP_INITIATION: /* must be first and only chunk */ SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_INIT\n"); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore? */ if ((stcb) && (stcb->asoc.total_output_queue_size)) { /* * collision case where we are * sending to them too */ ; } else { if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } } if ((chk_length > SCTP_LARGEST_INIT_ACCEPTED) || (num_chunks > 1) || (SCTP_BASE_SYSCTL(sctp_strict_init) && (length - *offset > (int)SCTP_SIZE32(chk_length)))) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if ((stcb != NULL) && (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT)) { sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); *offset = length; sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if (netp) { sctp_handle_init(m, iphlen, *offset, sh, (struct sctp_init_chunk *)ch, inp, stcb, *netp, &abort_no_unlock, vrf_id, port); } if (abort_no_unlock) return (NULL); *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); break; case SCTP_PAD_CHUNK: break; case SCTP_INITIATION_ACK: /* must be first and only chunk */ SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_INIT-ACK\n"); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore */ if ((stcb) && (stcb->asoc.total_output_queue_size)) { ; } else { if (locked_tcb != stcb) { /* Very unlikely */ SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; if (stcb) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_27); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } return (NULL); } } if ((num_chunks > 1) || (SCTP_BASE_SYSCTL(sctp_strict_init) && (length - *offset > (int)SCTP_SIZE32(chk_length)))) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if ((netp) && (*netp)) { ret = sctp_handle_init_ack(m, iphlen, *offset, sh, (struct sctp_init_ack_chunk *)ch, stcb, *netp, &abort_no_unlock, vrf_id); } else { ret = -1; } /* * Special case, I must call the output routine to * get the cookie echoed */ if (abort_no_unlock) return (NULL); if ((stcb) && ret == 0) sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); break; case SCTP_SELECTIVE_ACK: { struct sctp_sack_chunk *sack; int abort_now = 0; uint32_t a_rwnd, cum_ack; uint16_t num_seg, num_dup; uint8_t flags; int offset_seg, offset_dup; - int nonce_sum_flag; SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SACK\n"); SCTP_STAT_INCR(sctps_recvsacks); if (stcb == NULL) { SCTPDBG(SCTP_DEBUG_INDATA1, "No stcb when processing SACK chunk\n"); break; } if (chk_length < sizeof(struct sctp_sack_chunk)) { SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size on SACK chunk, too small\n"); break; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) { /*- * If we have sent a shutdown-ack, we will pay no * attention to a sack sent in to us since * we don't care anymore. */ break; } sack = (struct sctp_sack_chunk *)ch; flags = ch->chunk_flags; - nonce_sum_flag = flags & SCTP_SACK_NONCE_SUM; cum_ack = ntohl(sack->sack.cum_tsn_ack); num_seg = ntohs(sack->sack.num_gap_ack_blks); num_dup = ntohs(sack->sack.num_dup_tsns); a_rwnd = (uint32_t) ntohl(sack->sack.a_rwnd); if (sizeof(struct sctp_sack_chunk) + num_seg * sizeof(struct sctp_gap_ack_block) + num_dup * sizeof(uint32_t) != chk_length) { SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size of SACK chunk\n"); break; } offset_seg = *offset + sizeof(struct sctp_sack_chunk); offset_dup = offset_seg + num_seg * sizeof(struct sctp_gap_ack_block); SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SACK process cum_ack:%x num_seg:%d a_rwnd:%d\n", cum_ack, num_seg, a_rwnd); stcb->asoc.seen_a_sack_this_pkt = 1; if ((stcb->asoc.pr_sctp_cnt == 0) && (num_seg == 0) && SCTP_TSN_GE(cum_ack, stcb->asoc.last_acked_seq) && (stcb->asoc.saw_sack_with_frags == 0) && (stcb->asoc.saw_sack_with_nr_frags == 0) && (!TAILQ_EMPTY(&stcb->asoc.sent_queue)) ) { /* * We have a SIMPLE sack having no * prior segments and data on sent * queue to be acked.. Use the * faster path sack processing. We * also allow window update sacks * with no missing segments to go * this way too. */ - sctp_express_handle_sack(stcb, cum_ack, a_rwnd, nonce_sum_flag, - &abort_now); + sctp_express_handle_sack(stcb, cum_ack, a_rwnd, &abort_now); } else { if (netp && *netp) sctp_handle_sack(m, offset_seg, offset_dup, stcb, *netp, num_seg, 0, num_dup, &abort_now, flags, cum_ack, a_rwnd); } if (abort_now) { /* ABORT signal from sack processing */ *offset = length; return (NULL); } if (TAILQ_EMPTY(&stcb->asoc.send_queue) && TAILQ_EMPTY(&stcb->asoc.sent_queue) && (stcb->asoc.stream_queue_cnt == 0)) { sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); } } break; /* * EY - nr_sack: If the received chunk is an * nr_sack chunk */ case SCTP_NR_SELECTIVE_ACK: { struct sctp_nr_sack_chunk *nr_sack; int abort_now = 0; uint32_t a_rwnd, cum_ack; uint16_t num_seg, num_nr_seg, num_dup; uint8_t flags; int offset_seg, offset_dup; - int nonce_sum_flag; SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_NR_SACK\n"); SCTP_STAT_INCR(sctps_recvsacks); if (stcb == NULL) { SCTPDBG(SCTP_DEBUG_INDATA1, "No stcb when processing NR-SACK chunk\n"); break; } if ((stcb->asoc.sctp_nr_sack_on_off == 0) || (stcb->asoc.peer_supports_nr_sack == 0)) { goto unknown_chunk; } if (chk_length < sizeof(struct sctp_nr_sack_chunk)) { SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size on NR-SACK chunk, too small\n"); break; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) { /*- * If we have sent a shutdown-ack, we will pay no * attention to a sack sent in to us since * we don't care anymore. */ break; } nr_sack = (struct sctp_nr_sack_chunk *)ch; flags = ch->chunk_flags; - nonce_sum_flag = flags & SCTP_SACK_NONCE_SUM; - cum_ack = ntohl(nr_sack->nr_sack.cum_tsn_ack); num_seg = ntohs(nr_sack->nr_sack.num_gap_ack_blks); num_nr_seg = ntohs(nr_sack->nr_sack.num_nr_gap_ack_blks); num_dup = ntohs(nr_sack->nr_sack.num_dup_tsns); a_rwnd = (uint32_t) ntohl(nr_sack->nr_sack.a_rwnd); if (sizeof(struct sctp_nr_sack_chunk) + (num_seg + num_nr_seg) * sizeof(struct sctp_gap_ack_block) + num_dup * sizeof(uint32_t) != chk_length) { SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size of NR_SACK chunk\n"); break; } offset_seg = *offset + sizeof(struct sctp_nr_sack_chunk); offset_dup = offset_seg + num_seg * sizeof(struct sctp_gap_ack_block); SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_NR_SACK process cum_ack:%x num_seg:%d a_rwnd:%d\n", cum_ack, num_seg, a_rwnd); stcb->asoc.seen_a_sack_this_pkt = 1; if ((stcb->asoc.pr_sctp_cnt == 0) && (num_seg == 0) && (num_nr_seg == 0) && SCTP_TSN_GE(cum_ack, stcb->asoc.last_acked_seq) && (stcb->asoc.saw_sack_with_frags == 0) && (stcb->asoc.saw_sack_with_nr_frags == 0) && (!TAILQ_EMPTY(&stcb->asoc.sent_queue))) { /* * We have a SIMPLE sack having no * prior segments and data on sent * queue to be acked. Use the faster * path sack processing. We also * allow window update sacks with no * missing segments to go this way * too. */ - sctp_express_handle_sack(stcb, cum_ack, a_rwnd, nonce_sum_flag, + sctp_express_handle_sack(stcb, cum_ack, a_rwnd, &abort_now); } else { if (netp && *netp) sctp_handle_sack(m, offset_seg, offset_dup, stcb, *netp, num_seg, num_nr_seg, num_dup, &abort_now, flags, cum_ack, a_rwnd); } if (abort_now) { /* ABORT signal from sack processing */ *offset = length; return (NULL); } if (TAILQ_EMPTY(&stcb->asoc.send_queue) && TAILQ_EMPTY(&stcb->asoc.sent_queue) && (stcb->asoc.stream_queue_cnt == 0)) { sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); } } break; case SCTP_HEARTBEAT_REQUEST: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_HEARTBEAT\n"); if ((stcb) && netp && *netp) { SCTP_STAT_INCR(sctps_recvheartbeat); sctp_send_heartbeat_ack(stcb, m, *offset, chk_length, *netp); /* He's alive so give him credit */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; } break; case SCTP_HEARTBEAT_ACK: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_HEARTBEAT-ACK\n"); if ((stcb == NULL) || (chk_length != sizeof(struct sctp_heartbeat_chunk))) { /* Its not ours */ *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } /* He's alive so give him credit */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; SCTP_STAT_INCR(sctps_recvheartbeatack); if (netp && *netp) sctp_handle_heartbeat_ack((struct sctp_heartbeat_chunk *)ch, stcb, *netp); break; case SCTP_ABORT_ASSOCIATION: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ABORT, stcb %p\n", stcb); if ((stcb) && netp && *netp) sctp_handle_abort((struct sctp_abort_chunk *)ch, stcb, *netp); *offset = length; return (NULL); break; case SCTP_SHUTDOWN: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN, stcb %p\n", stcb); if ((stcb == NULL) || (chk_length != sizeof(struct sctp_shutdown_chunk))) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if (netp && *netp) { int abort_flag = 0; sctp_handle_shutdown((struct sctp_shutdown_chunk *)ch, stcb, *netp, &abort_flag); if (abort_flag) { *offset = length; return (NULL); } } break; case SCTP_SHUTDOWN_ACK: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN-ACK, stcb %p\n", stcb); if ((stcb) && (netp) && (*netp)) sctp_handle_shutdown_ack((struct sctp_shutdown_ack_chunk *)ch, stcb, *netp); *offset = length; return (NULL); break; case SCTP_OPERATION_ERROR: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_OP-ERR\n"); if ((stcb) && netp && *netp && sctp_handle_error(ch, stcb, *netp) < 0) { *offset = length; return (NULL); } break; case SCTP_COOKIE_ECHO: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_COOKIE-ECHO, stcb %p\n", stcb); if ((stcb) && (stcb->asoc.total_output_queue_size)) { ; } else { if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore */ abend: if (stcb) { SCTP_TCB_UNLOCK(stcb); } *offset = length; return (NULL); } } /* * First are we accepting? We do this again here * since it is possible that a previous endpoint WAS * listening responded to a INIT-ACK and then * closed. We opened and bound.. and are now no * longer listening. */ if ((stcb == NULL) && (inp->sctp_socket->so_qlen >= inp->sctp_socket->so_qlimit)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit))) { struct mbuf *oper; struct sctp_paramhdr *phdr; oper = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr); phdr = mtod(oper, struct sctp_paramhdr *); phdr->param_type = htons(SCTP_CAUSE_OUT_OF_RESC); phdr->param_length = htons(sizeof(struct sctp_paramhdr)); } sctp_abort_association(inp, stcb, m, iphlen, sh, oper, vrf_id, port); } *offset = length; return (NULL); } else { struct mbuf *ret_buf; struct sctp_inpcb *linp; if (stcb) { linp = NULL; } else { linp = inp; } if (linp) { SCTP_ASOC_CREATE_LOCK(linp); if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { SCTP_ASOC_CREATE_UNLOCK(linp); goto abend; } } if (netp) { ret_buf = sctp_handle_cookie_echo(m, iphlen, *offset, sh, (struct sctp_cookie_echo_chunk *)ch, &inp, &stcb, netp, auth_skipped, auth_offset, auth_len, &locked_tcb, vrf_id, port); } else { ret_buf = NULL; } if (linp) { SCTP_ASOC_CREATE_UNLOCK(linp); } if (ret_buf == NULL) { if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } SCTPDBG(SCTP_DEBUG_INPUT3, "GAK, null buffer\n"); auth_skipped = 0; *offset = length; return (NULL); } /* if AUTH skipped, see if it verified... */ if (auth_skipped) { got_auth = 1; auth_skipped = 0; } if (!TAILQ_EMPTY(&stcb->asoc.sent_queue)) { /* * Restart the timer if we have * pending data */ struct sctp_tmit_chunk *chk; chk = TAILQ_FIRST(&stcb->asoc.sent_queue); sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); } } break; case SCTP_COOKIE_ACK: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_COOKIE-ACK, stcb %p\n", stcb); if ((stcb == NULL) || chk_length != sizeof(struct sctp_cookie_ack_chunk)) { if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore */ if ((stcb) && (stcb->asoc.total_output_queue_size)) { ; } else if (stcb) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_27); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif *offset = length; return (NULL); } } /* He's alive so give him credit */ if ((stcb) && netp && *netp) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, *netp); } break; case SCTP_ECN_ECHO: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ECN-ECHO\n"); /* He's alive so give him credit */ if ((stcb == NULL) || (chk_length != sizeof(struct sctp_ecne_chunk))) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if (stcb) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_ecn_echo((struct sctp_ecne_chunk *)ch, stcb); } break; case SCTP_ECN_CWR: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ECN-CWR\n"); /* He's alive so give him credit */ if ((stcb == NULL) || (chk_length != sizeof(struct sctp_cwr_chunk))) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if (stcb) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_ecn_cwr((struct sctp_cwr_chunk *)ch, stcb, *netp); } break; case SCTP_SHUTDOWN_COMPLETE: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN-COMPLETE, stcb %p\n", stcb); /* must be first and only chunk */ if ((num_chunks > 1) || (length - *offset > (int)SCTP_SIZE32(chk_length))) { *offset = length; if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } return (NULL); } if ((stcb) && netp && *netp) { sctp_handle_shutdown_complete((struct sctp_shutdown_complete_chunk *)ch, stcb, *netp); } *offset = length; return (NULL); break; case SCTP_ASCONF: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ASCONF\n"); /* He's alive so give him credit */ if (stcb) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_asconf(m, *offset, (struct sctp_asconf_chunk *)ch, stcb, asconf_cnt == 0); asconf_cnt++; } break; case SCTP_ASCONF_ACK: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ASCONF-ACK\n"); if (chk_length < sizeof(struct sctp_asconf_ack_chunk)) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if ((stcb) && netp && *netp) { /* He's alive so give him credit */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_asconf_ack(m, *offset, (struct sctp_asconf_ack_chunk *)ch, stcb, *netp, &abort_no_unlock); if (abort_no_unlock) return (NULL); } break; case SCTP_FORWARD_CUM_TSN: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_FWD-TSN\n"); if (chk_length < sizeof(struct sctp_forward_tsn_chunk)) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } /* He's alive so give him credit */ if (stcb) { int abort_flag = 0; stcb->asoc.overall_error_count = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } *fwd_tsn_seen = 1; if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_29); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif *offset = length; return (NULL); } sctp_handle_forward_tsn(stcb, (struct sctp_forward_tsn_chunk *)ch, &abort_flag, m, *offset); if (abort_flag) { *offset = length; return (NULL); } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; } } break; case SCTP_STREAM_RESET: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_STREAM_RESET\n"); if (((stcb == NULL) || (ch == NULL) || (chk_length < sizeof(struct sctp_stream_reset_tsn_req)))) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* We are not interested anymore */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_INPUT + SCTP_LOC_30); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif *offset = length; return (NULL); } if (stcb->asoc.peer_supports_strreset == 0) { /* * hmm, peer should have announced this, but * we will turn it on since he is sending us * a stream reset. */ stcb->asoc.peer_supports_strreset = 1; } if (sctp_handle_stream_reset(stcb, m, *offset, (struct sctp_stream_reset_out_req *)ch)) { /* stop processing */ *offset = length; return (NULL); } break; case SCTP_PACKET_DROPPED: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_PACKET_DROPPED\n"); /* re-get it all please */ if (chk_length < sizeof(struct sctp_pktdrop_chunk)) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if (ch && (stcb) && netp && (*netp)) { sctp_handle_packet_dropped((struct sctp_pktdrop_chunk *)ch, stcb, *netp, min(chk_length, (sizeof(chunk_buf) - 4))); } break; case SCTP_AUTHENTICATION: SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_AUTHENTICATION\n"); if (SCTP_BASE_SYSCTL(sctp_auth_disable)) goto unknown_chunk; if (stcb == NULL) { /* save the first AUTH for later processing */ if (auth_skipped == 0) { auth_offset = *offset; auth_len = chk_length; auth_skipped = 1; } /* skip this chunk (temporarily) */ goto next_chunk; } if ((chk_length < (sizeof(struct sctp_auth_chunk))) || (chk_length > (sizeof(struct sctp_auth_chunk) + SCTP_AUTH_DIGEST_LEN_MAX))) { /* Its not ours */ if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } if (got_auth == 1) { /* skip this chunk... it's already auth'd */ goto next_chunk; } got_auth = 1; if ((ch == NULL) || sctp_handle_auth(stcb, (struct sctp_auth_chunk *)ch, m, *offset)) { /* auth HMAC failed so dump the packet */ *offset = length; return (stcb); } else { /* remaining chunks are HMAC checked */ stcb->asoc.authenticated = 1; } break; default: unknown_chunk: /* it's an unknown chunk! */ if ((ch->chunk_type & 0x40) && (stcb != NULL)) { struct mbuf *mm; struct sctp_paramhdr *phd; mm = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (mm) { phd = mtod(mm, struct sctp_paramhdr *); /* * We cheat and use param type since * we did not bother to define a * error cause struct. They are the * same basic format with different * names. */ phd->param_type = htons(SCTP_CAUSE_UNRECOG_CHUNK); phd->param_length = htons(chk_length + sizeof(*phd)); SCTP_BUF_LEN(mm) = sizeof(*phd); SCTP_BUF_NEXT(mm) = SCTP_M_COPYM(m, *offset, SCTP_SIZE32(chk_length), M_DONTWAIT); if (SCTP_BUF_NEXT(mm)) { #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = SCTP_BUF_NEXT(mm); while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif sctp_queue_op_err(stcb, mm); } else { sctp_m_freem(mm); } } } if ((ch->chunk_type & 0x80) == 0) { /* discard this packet */ *offset = length; return (stcb); } /* else skip this bad chunk and continue... */ break; } /* switch (ch->chunk_type) */ next_chunk: /* get the next chunk */ *offset += SCTP_SIZE32(chk_length); if (*offset >= length) { /* no more data left in the mbuf chain */ break; } ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, sizeof(struct sctp_chunkhdr), chunk_buf); if (ch == NULL) { if (locked_tcb) { SCTP_TCB_UNLOCK(locked_tcb); } *offset = length; return (NULL); } } /* while */ if (asconf_cnt > 0 && stcb != NULL) { sctp_send_asconf_ack(stcb); } return (stcb); } -/* - * Process the ECN bits we have something set so we must look to see if it is - * ECN(0) or ECN(1) or CE - */ -static void -sctp_process_ecn_marked_a(struct sctp_tcb *stcb, struct sctp_nets *net, - uint8_t ecn_bits) -{ - if ((ecn_bits & SCTP_CE_BITS) == SCTP_CE_BITS) { - ; - } else if ((ecn_bits & SCTP_ECT1_BIT) == SCTP_ECT1_BIT) { - /* - * we only add to the nonce sum for ECT1, ECT0 does not - * change the NS bit (that we have yet to find a way to send - * it yet). - */ - - /* ECN Nonce stuff */ - stcb->asoc.receiver_nonce_sum++; - stcb->asoc.receiver_nonce_sum &= SCTP_SACK_NONCE_SUM; - - /* - * Drag up the last_echo point if cumack is larger since we - * don't want the point falling way behind by more than - * 2^^31 and then having it be incorrect. - */ - if (SCTP_TSN_GT(stcb->asoc.cumulative_tsn, stcb->asoc.last_echo_tsn)) { - stcb->asoc.last_echo_tsn = stcb->asoc.cumulative_tsn; - } - } else if ((ecn_bits & SCTP_ECT0_BIT) == SCTP_ECT0_BIT) { - /* - * Drag up the last_echo point if cumack is larger since we - * don't want the point falling way behind by more than - * 2^^31 and then having it be incorrect. - */ - if (SCTP_TSN_GT(stcb->asoc.cumulative_tsn, stcb->asoc.last_echo_tsn)) { - stcb->asoc.last_echo_tsn = stcb->asoc.cumulative_tsn; - } - } -} - -static void -sctp_process_ecn_marked_b(struct sctp_tcb *stcb, struct sctp_nets *net, - uint32_t high_tsn, uint8_t ecn_bits) -{ - if ((ecn_bits & SCTP_CE_BITS) == SCTP_CE_BITS) { - /* - * we possibly must notify the sender that a congestion - * window reduction is in order. We do this by adding a ECNE - * chunk to the output chunk queue. The incoming CWR will - * remove this chunk. - */ - if (SCTP_TSN_GT(high_tsn, stcb->asoc.last_echo_tsn)) { - /* Yep, we need to add a ECNE */ - sctp_send_ecn_echo(stcb, net, high_tsn); - stcb->asoc.last_echo_tsn = high_tsn; - } - } -} - #ifdef INVARIANTS #ifdef __GNUC__ __attribute__((noinline)) #endif void sctp_validate_no_locks(struct sctp_inpcb *inp) { struct sctp_tcb *lstcb; LIST_FOREACH(lstcb, &inp->sctp_asoc_list, sctp_tcblist) { if (mtx_owned(&lstcb->tcb_mtx)) { panic("Own lock on stcb at return from input"); } } if (mtx_owned(&inp->inp_create_mtx)) { panic("Own create lock on inp"); } if (mtx_owned(&inp->inp_mtx)) { panic("Own inp lock on inp"); } } #endif /* * common input chunk processing (v4 and v6) */ void sctp_common_input_processing(struct mbuf **mm, int iphlen, int offset, int length, struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint8_t ecn_bits, uint32_t vrf_id, uint16_t port) { /* * Control chunk processing */ uint32_t high_tsn; int fwd_tsn_seen = 0, data_processed = 0; struct mbuf *m = *mm; int abort_flag = 0; int un_sent; + int cnt_ctrl_ready = 0; SCTP_STAT_INCR(sctps_recvdatagrams); #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xE0, 1); sctp_auditing(0, inp, stcb, net); #endif SCTPDBG(SCTP_DEBUG_INPUT1, "Ok, Common input processing called, m:%p iphlen:%d offset:%d length:%d stcb:%p\n", m, iphlen, offset, length, stcb); if (stcb) { /* always clear this before beginning a packet */ stcb->asoc.authenticated = 0; stcb->asoc.seen_a_sack_this_pkt = 0; SCTPDBG(SCTP_DEBUG_INPUT1, "stcb:%p state:%x\n", stcb, stcb->asoc.state); if ((stcb->asoc.state & SCTP_STATE_WAS_ABORTED) || (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED)) { /*- * If we hit here, we had a ref count * up when the assoc was aborted and the * timer is clearing out the assoc, we should * NOT respond to any packet.. its OOTB. */ SCTP_TCB_UNLOCK(stcb); sctp_handle_ootb(m, iphlen, offset, sh, inp, NULL, vrf_id, port); goto out_now; } } if (IS_SCTP_CONTROL(ch)) { /* process the control portion of the SCTP packet */ /* sa_ignore NO_NULL_CHK */ stcb = sctp_process_control(m, iphlen, &offset, length, sh, ch, inp, stcb, &net, &fwd_tsn_seen, vrf_id, port); if (stcb) { /* * This covers us if the cookie-echo was there and * it changes our INP. */ inp = stcb->sctp_ep; if ((net) && (port)) { if (net->port == 0) { sctp_pathmtu_adjustment(inp, stcb, net, net->mtu - sizeof(struct udphdr)); } net->port = port; } } } else { /* * no control chunks, so pre-process DATA chunks (these * checks are taken care of by control processing) */ /* * if DATA only packet, and auth is required, then punt... * can't have authenticated without any AUTH (control) * chunks */ if ((stcb != NULL) && !SCTP_BASE_SYSCTL(sctp_auth_disable) && sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.local_auth_chunks)) { /* "silently" ignore */ SCTP_STAT_INCR(sctps_recvauthmissing); SCTP_TCB_UNLOCK(stcb); goto out_now; } if (stcb == NULL) { /* out of the blue DATA chunk */ sctp_handle_ootb(m, iphlen, offset, sh, inp, NULL, vrf_id, port); goto out_now; } if (stcb->asoc.my_vtag != ntohl(sh->v_tag)) { /* v_tag mismatch! */ SCTP_STAT_INCR(sctps_badvtag); SCTP_TCB_UNLOCK(stcb); goto out_now; } } if (stcb == NULL) { /* * no valid TCB for this packet, or we found it's a bad * packet while processing control, or we're done with this * packet (done or skip rest of data), so we drop it... */ goto out_now; } /* * DATA chunk processing */ /* plow through the data chunks while length > offset */ /* * Rest should be DATA only. Check authentication state if AUTH for * DATA is required. */ if ((length > offset) && (stcb != NULL) && !SCTP_BASE_SYSCTL(sctp_auth_disable) && sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.local_auth_chunks) && !stcb->asoc.authenticated) { /* "silently" ignore */ SCTP_STAT_INCR(sctps_recvauthmissing); SCTPDBG(SCTP_DEBUG_AUTH1, "Data chunk requires AUTH, skipped\n"); goto trigger_send; } if (length > offset) { int retval; /* * First check to make sure our state is correct. We would * not get here unless we really did have a tag, so we don't * abort if this happens, just dump the chunk silently. */ switch (SCTP_GET_STATE(&stcb->asoc)) { case SCTP_STATE_COOKIE_ECHOED: /* * we consider data with valid tags in this state * shows us the cookie-ack was lost. Imply it was * there. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_CLEAR, stcb->asoc.overall_error_count, 0, SCTP_FROM_SCTP_INPUT, __LINE__); } stcb->asoc.overall_error_count = 0; sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, net); break; case SCTP_STATE_COOKIE_WAIT: /* * We consider OOTB any data sent during asoc setup. */ sctp_handle_ootb(m, iphlen, offset, sh, inp, NULL, vrf_id, port); SCTP_TCB_UNLOCK(stcb); goto out_now; /* sa_ignore NOTREACHED */ break; case SCTP_STATE_EMPTY: /* should not happen */ case SCTP_STATE_INUSE: /* should not happen */ case SCTP_STATE_SHUTDOWN_RECEIVED: /* This is a peer error */ case SCTP_STATE_SHUTDOWN_ACK_SENT: default: SCTP_TCB_UNLOCK(stcb); goto out_now; /* sa_ignore NOTREACHED */ break; case SCTP_STATE_OPEN: case SCTP_STATE_SHUTDOWN_SENT: break; } - /* take care of ECN, part 1. */ - if (stcb->asoc.ecn_allowed && - (ecn_bits & (SCTP_ECT0_BIT | SCTP_ECT1_BIT))) { - sctp_process_ecn_marked_a(stcb, net, ecn_bits); - } /* plow through the data chunks while length > offset */ retval = sctp_process_data(mm, iphlen, &offset, length, sh, inp, stcb, net, &high_tsn); if (retval == 2) { /* * The association aborted, NO UNLOCK needed since * the association is destroyed. */ goto out_now; } data_processed = 1; - if (retval == 0) { - /* take care of ecn part 2. */ - if (stcb->asoc.ecn_allowed && - (ecn_bits & (SCTP_ECT0_BIT | SCTP_ECT1_BIT))) { - sctp_process_ecn_marked_b(stcb, net, high_tsn, - ecn_bits); - } - } /* * Anything important needs to have been m_copy'ed in * process_data */ } + /* take care of ecn */ + if (stcb->asoc.ecn_allowed && ((ecn_bits & SCTP_CE_BITS) == SCTP_CE_BITS)) { + /* Yep, we need to add a ECNE */ + sctp_send_ecn_echo(stcb, net, high_tsn); + } if ((data_processed == 0) && (fwd_tsn_seen)) { int was_a_gap; uint32_t highest_tsn; if (SCTP_TSN_GT(stcb->asoc.highest_tsn_inside_nr_map, stcb->asoc.highest_tsn_inside_map)) { highest_tsn = stcb->asoc.highest_tsn_inside_nr_map; } else { highest_tsn = stcb->asoc.highest_tsn_inside_map; } was_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); stcb->asoc.send_sack = 1; sctp_sack_check(stcb, was_a_gap, &abort_flag); if (abort_flag) { /* Again, we aborted so NO UNLOCK needed */ goto out_now; } } else if (fwd_tsn_seen) { stcb->asoc.send_sack = 1; } /* trigger send of any chunks in queue... */ trigger_send: #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xE0, 2); sctp_auditing(1, inp, stcb, net); #endif SCTPDBG(SCTP_DEBUG_INPUT1, "Check for chunk output prw:%d tqe:%d tf=%d\n", stcb->asoc.peers_rwnd, TAILQ_EMPTY(&stcb->asoc.control_send_queue), stcb->asoc.total_flight); un_sent = (stcb->asoc.total_output_queue_size - stcb->asoc.total_flight); - - if (!TAILQ_EMPTY(&stcb->asoc.control_send_queue) || + if (!TAILQ_EMPTY(&stcb->asoc.control_send_queue)) { + cnt_ctrl_ready = stcb->asoc.ctrl_queue_cnt - stcb->asoc.ecn_echo_cnt_onq; + } + if (cnt_ctrl_ready || ((un_sent) && (stcb->asoc.peers_rwnd > 0 || (stcb->asoc.peers_rwnd <= 0 && stcb->asoc.total_flight == 0)))) { SCTPDBG(SCTP_DEBUG_INPUT3, "Calling chunk OUTPUT\n"); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); SCTPDBG(SCTP_DEBUG_INPUT3, "chunk OUTPUT returns\n"); } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xE0, 3); sctp_auditing(2, inp, stcb, net); #endif SCTP_TCB_UNLOCK(stcb); out_now: #ifdef INVARIANTS sctp_validate_no_locks(inp); #endif return; } #if 0 static void sctp_print_mbuf_chain(struct mbuf *m) { for (; m; m = SCTP_BUF_NEXT(m)) { printf("%p: m_len = %ld\n", m, SCTP_BUF_LEN(m)); if (SCTP_BUF_IS_EXTENDED(m)) printf("%p: extend_size = %d\n", m, SCTP_BUF_EXTEND_SIZE(m)); } } #endif void sctp_input_with_port(struct mbuf *i_pak, int off, uint16_t port) { #ifdef SCTP_MBUF_LOGGING struct mbuf *mat; #endif struct mbuf *m; int iphlen; uint32_t vrf_id = 0; uint8_t ecn_bits; struct ip *ip; struct sctphdr *sh; struct sctp_inpcb *inp = NULL; struct sctp_nets *net; struct sctp_tcb *stcb = NULL; struct sctp_chunkhdr *ch; int refcount_up = 0; int length, mlen, offset; #if !defined(SCTP_WITH_NO_CSUM) uint32_t check, calc_check; #endif if (SCTP_GET_PKT_VRFID(i_pak, vrf_id)) { SCTP_RELEASE_PKT(i_pak); return; } mlen = SCTP_HEADER_LEN(i_pak); iphlen = off; m = SCTP_HEADER_TO_CHAIN(i_pak); net = NULL; SCTP_STAT_INCR(sctps_recvpackets); SCTP_STAT_INCR_COUNTER64(sctps_inpackets); #ifdef SCTP_MBUF_LOGGING /* Log in any input mbufs */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { mat = m; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_INPUT); } mat = SCTP_BUF_NEXT(mat); } } #endif #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(m, mlen); #endif /* * Must take out the iphlen, since mlen expects this (only effect lb * case) */ mlen -= iphlen; /* * Get IP, SCTP, and first chunk header together in first mbuf. */ ip = mtod(m, struct ip *); offset = iphlen + sizeof(*sh) + sizeof(*ch); if (SCTP_BUF_LEN(m) < offset) { if ((m = m_pullup(m, offset)) == 0) { SCTP_STAT_INCR(sctps_hdrops); return; } ip = mtod(m, struct ip *); } /* validate mbuf chain length with IP payload length */ if (mlen < (SCTP_GET_IPV4_LENGTH(ip) - iphlen)) { SCTP_STAT_INCR(sctps_hdrops); goto bad; } sh = (struct sctphdr *)((caddr_t)ip + iphlen); ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(*sh)); SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_input() length:%d iphlen:%d\n", mlen, iphlen); /* SCTP does not allow broadcasts or multicasts */ if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { goto bad; } if (SCTP_IS_IT_BROADCAST(ip->ip_dst, m)) { /* * We only look at broadcast if its a front state, All * others we will not have a tcb for anyway. */ goto bad; } /* validate SCTP checksum */ SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, "sctp_input(): Packet of length %d received on %s with csum_flags 0x%x.\n", m->m_pkthdr.len, if_name(m->m_pkthdr.rcvif), m->m_pkthdr.csum_flags); #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_recvnocrc); #else if (m->m_pkthdr.csum_flags & CSUM_SCTP_VALID) { SCTP_STAT_INCR(sctps_recvhwcrc); goto sctp_skip_csum_4; } check = sh->checksum; /* save incoming checksum */ sh->checksum = 0; /* prepare for calc */ calc_check = sctp_calculate_cksum(m, iphlen); sh->checksum = check; SCTP_STAT_INCR(sctps_recvswcrc); if (calc_check != check) { SCTPDBG(SCTP_DEBUG_INPUT1, "Bad CSUM on SCTP packet calc_check:%x check:%x m:%p mlen:%d iphlen:%d\n", calc_check, check, m, mlen, iphlen); stcb = sctp_findassociation_addr(m, iphlen, offset - sizeof(*ch), sh, ch, &inp, &net, vrf_id); if ((net) && (port)) { if (net->port == 0) { sctp_pathmtu_adjustment(inp, stcb, net, net->mtu - sizeof(struct udphdr)); } net->port = port; } if ((inp) && (stcb)) { sctp_send_packet_dropped(stcb, net, m, iphlen, 1); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_INPUT_ERROR, SCTP_SO_NOT_LOCKED); } else if ((inp != NULL) && (stcb == NULL)) { refcount_up = 1; } SCTP_STAT_INCR(sctps_badsum); SCTP_STAT_INCR_COUNTER32(sctps_checksumerrors); goto bad; } sctp_skip_csum_4: #endif /* destination port of 0 is illegal, based on RFC2960. */ if (sh->dest_port == 0) { SCTP_STAT_INCR(sctps_hdrops); goto bad; } /* * Locate pcb and tcb for datagram sctp_findassociation_addr() wants * IP/SCTP/first chunk header... */ stcb = sctp_findassociation_addr(m, iphlen, offset - sizeof(*ch), sh, ch, &inp, &net, vrf_id); if ((net) && (port)) { if (net->port == 0) { sctp_pathmtu_adjustment(inp, stcb, net, net->mtu - sizeof(struct udphdr)); } net->port = port; } /* inp's ref-count increased && stcb locked */ if (inp == NULL) { struct sctp_init_chunk *init_chk, chunk_buf; SCTP_STAT_INCR(sctps_noport); #ifdef ICMP_BANDLIM /* * we use the bandwidth limiting to protect against sending * too many ABORTS all at once. In this case these count the * same as an ICMP message. */ if (badport_bandlim(0) < 0) goto bad; #endif /* ICMP_BANDLIM */ SCTPDBG(SCTP_DEBUG_INPUT1, "Sending a ABORT from packet entry!\n"); if (ch->chunk_type == SCTP_INITIATION) { /* * we do a trick here to get the INIT tag, dig in * and get the tag from the INIT and put it in the * common header. */ init_chk = (struct sctp_init_chunk *)sctp_m_getptr(m, iphlen + sizeof(*sh), sizeof(*init_chk), (uint8_t *) & chunk_buf); if (init_chk != NULL) sh->v_tag = init_chk->init.initiate_tag; } if (ch->chunk_type == SCTP_SHUTDOWN_ACK) { sctp_send_shutdown_complete2(m, iphlen, sh, vrf_id, port); goto bad; } if (ch->chunk_type == SCTP_SHUTDOWN_COMPLETE) { goto bad; } if (ch->chunk_type != SCTP_ABORT_ASSOCIATION) sctp_send_abort(m, iphlen, sh, 0, NULL, vrf_id, port); goto bad; } else if (stcb == NULL) { refcount_up = 1; } #ifdef IPSEC /* * I very much doubt any of the IPSEC stuff will work but I have no * idea, so I will leave it in place. */ if (inp && ipsec4_in_reject(m, &inp->ip_inp.inp)) { MODULE_GLOBAL(ipsec4stat).in_polvio++; SCTP_STAT_INCR(sctps_hdrops); goto bad; } #endif /* IPSEC */ /* * common chunk processing */ length = ip->ip_len + iphlen; offset -= sizeof(struct sctp_chunkhdr); ecn_bits = ip->ip_tos; /* sa_ignore NO_NULL_CHK */ sctp_common_input_processing(&m, iphlen, offset, length, sh, ch, inp, stcb, net, ecn_bits, vrf_id, port); /* inp's ref-count reduced && stcb unlocked */ if (m) { sctp_m_freem(m); } if ((inp) && (refcount_up)) { /* reduce ref-count */ SCTP_INP_DECR_REF(inp); } return; bad: if (stcb) { SCTP_TCB_UNLOCK(stcb); } if ((inp) && (refcount_up)) { /* reduce ref-count */ SCTP_INP_DECR_REF(inp); } if (m) { sctp_m_freem(m); } return; } void sctp_input(i_pak, off) struct mbuf *i_pak; int off; { sctp_input_with_port(i_pak, off, 0); } diff --git a/sys/netinet/sctp_output.c b/sys/netinet/sctp_output.c index 06510546fd44..d1d914793b01 100644 --- a/sys/netinet/sctp_output.c +++ b/sys/netinet/sctp_output.c @@ -1,13464 +1,13393 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_output.c,v 1.46 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define SCTP_MAX_GAPS_INARRAY 4 struct sack_track { uint8_t right_edge; /* mergable on the right edge */ uint8_t left_edge; /* mergable on the left edge */ uint8_t num_entries; uint8_t spare; struct sctp_gap_ack_block gaps[SCTP_MAX_GAPS_INARRAY]; }; struct sack_track sack_array[256] = { {0, 0, 0, 0, /* 0x00 */ {{0, 0}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x01 */ {{0, 0}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x02 */ {{1, 1}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x03 */ {{0, 1}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x04 */ {{2, 2}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x05 */ {{0, 0}, {2, 2}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x06 */ {{1, 2}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x07 */ {{0, 2}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x08 */ {{3, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x09 */ {{0, 0}, {3, 3}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x0a */ {{1, 1}, {3, 3}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x0b */ {{0, 1}, {3, 3}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x0c */ {{2, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x0d */ {{0, 0}, {2, 3}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x0e */ {{1, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x0f */ {{0, 3}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x10 */ {{4, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x11 */ {{0, 0}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x12 */ {{1, 1}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x13 */ {{0, 1}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x14 */ {{2, 2}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x15 */ {{0, 0}, {2, 2}, {4, 4}, {0, 0} } }, {0, 0, 2, 0, /* 0x16 */ {{1, 2}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x17 */ {{0, 2}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x18 */ {{3, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x19 */ {{0, 0}, {3, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x1a */ {{1, 1}, {3, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x1b */ {{0, 1}, {3, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x1c */ {{2, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x1d */ {{0, 0}, {2, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x1e */ {{1, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x1f */ {{0, 4}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x20 */ {{5, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x21 */ {{0, 0}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x22 */ {{1, 1}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x23 */ {{0, 1}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x24 */ {{2, 2}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x25 */ {{0, 0}, {2, 2}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x26 */ {{1, 2}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x27 */ {{0, 2}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x28 */ {{3, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x29 */ {{0, 0}, {3, 3}, {5, 5}, {0, 0} } }, {0, 0, 3, 0, /* 0x2a */ {{1, 1}, {3, 3}, {5, 5}, {0, 0} } }, {1, 0, 3, 0, /* 0x2b */ {{0, 1}, {3, 3}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x2c */ {{2, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x2d */ {{0, 0}, {2, 3}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x2e */ {{1, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x2f */ {{0, 3}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x30 */ {{4, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x31 */ {{0, 0}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x32 */ {{1, 1}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x33 */ {{0, 1}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x34 */ {{2, 2}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x35 */ {{0, 0}, {2, 2}, {4, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x36 */ {{1, 2}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x37 */ {{0, 2}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x38 */ {{3, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x39 */ {{0, 0}, {3, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x3a */ {{1, 1}, {3, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x3b */ {{0, 1}, {3, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x3c */ {{2, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x3d */ {{0, 0}, {2, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x3e */ {{1, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x3f */ {{0, 5}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x40 */ {{6, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x41 */ {{0, 0}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x42 */ {{1, 1}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x43 */ {{0, 1}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x44 */ {{2, 2}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x45 */ {{0, 0}, {2, 2}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x46 */ {{1, 2}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x47 */ {{0, 2}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x48 */ {{3, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x49 */ {{0, 0}, {3, 3}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x4a */ {{1, 1}, {3, 3}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x4b */ {{0, 1}, {3, 3}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x4c */ {{2, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x4d */ {{0, 0}, {2, 3}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x4e */ {{1, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x4f */ {{0, 3}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x50 */ {{4, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x51 */ {{0, 0}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x52 */ {{1, 1}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x53 */ {{0, 1}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x54 */ {{2, 2}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 4, 0, /* 0x55 */ {{0, 0}, {2, 2}, {4, 4}, {6, 6} } }, {0, 0, 3, 0, /* 0x56 */ {{1, 2}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x57 */ {{0, 2}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x58 */ {{3, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x59 */ {{0, 0}, {3, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x5a */ {{1, 1}, {3, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x5b */ {{0, 1}, {3, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x5c */ {{2, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x5d */ {{0, 0}, {2, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x5e */ {{1, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x5f */ {{0, 4}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x60 */ {{5, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x61 */ {{0, 0}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x62 */ {{1, 1}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x63 */ {{0, 1}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x64 */ {{2, 2}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x65 */ {{0, 0}, {2, 2}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x66 */ {{1, 2}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x67 */ {{0, 2}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x68 */ {{3, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x69 */ {{0, 0}, {3, 3}, {5, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x6a */ {{1, 1}, {3, 3}, {5, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x6b */ {{0, 1}, {3, 3}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x6c */ {{2, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x6d */ {{0, 0}, {2, 3}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x6e */ {{1, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x6f */ {{0, 3}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x70 */ {{4, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x71 */ {{0, 0}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x72 */ {{1, 1}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x73 */ {{0, 1}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x74 */ {{2, 2}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x75 */ {{0, 0}, {2, 2}, {4, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x76 */ {{1, 2}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x77 */ {{0, 2}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x78 */ {{3, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x79 */ {{0, 0}, {3, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x7a */ {{1, 1}, {3, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x7b */ {{0, 1}, {3, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x7c */ {{2, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x7d */ {{0, 0}, {2, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x7e */ {{1, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x7f */ {{0, 6}, {0, 0}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0x80 */ {{7, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x81 */ {{0, 0}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x82 */ {{1, 1}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x83 */ {{0, 1}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x84 */ {{2, 2}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x85 */ {{0, 0}, {2, 2}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x86 */ {{1, 2}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x87 */ {{0, 2}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x88 */ {{3, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x89 */ {{0, 0}, {3, 3}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x8a */ {{1, 1}, {3, 3}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x8b */ {{0, 1}, {3, 3}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x8c */ {{2, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x8d */ {{0, 0}, {2, 3}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x8e */ {{1, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x8f */ {{0, 3}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x90 */ {{4, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x91 */ {{0, 0}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x92 */ {{1, 1}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x93 */ {{0, 1}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x94 */ {{2, 2}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0x95 */ {{0, 0}, {2, 2}, {4, 4}, {7, 7} } }, {0, 1, 3, 0, /* 0x96 */ {{1, 2}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x97 */ {{0, 2}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x98 */ {{3, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x99 */ {{0, 0}, {3, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x9a */ {{1, 1}, {3, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x9b */ {{0, 1}, {3, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x9c */ {{2, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x9d */ {{0, 0}, {2, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x9e */ {{1, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x9f */ {{0, 4}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xa0 */ {{5, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xa1 */ {{0, 0}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa2 */ {{1, 1}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xa3 */ {{0, 1}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa4 */ {{2, 2}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xa5 */ {{0, 0}, {2, 2}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xa6 */ {{1, 2}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xa7 */ {{0, 2}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa8 */ {{3, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xa9 */ {{0, 0}, {3, 3}, {5, 5}, {7, 7} } }, {0, 1, 4, 0, /* 0xaa */ {{1, 1}, {3, 3}, {5, 5}, {7, 7} } }, {1, 1, 4, 0, /* 0xab */ {{0, 1}, {3, 3}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xac */ {{2, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xad */ {{0, 0}, {2, 3}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xae */ {{1, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xaf */ {{0, 3}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xb0 */ {{4, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xb1 */ {{0, 0}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xb2 */ {{1, 1}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xb3 */ {{0, 1}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xb4 */ {{2, 2}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xb5 */ {{0, 0}, {2, 2}, {4, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xb6 */ {{1, 2}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xb7 */ {{0, 2}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xb8 */ {{3, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xb9 */ {{0, 0}, {3, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xba */ {{1, 1}, {3, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xbb */ {{0, 1}, {3, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xbc */ {{2, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xbd */ {{0, 0}, {2, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xbe */ {{1, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xbf */ {{0, 5}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xc0 */ {{6, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc1 */ {{0, 0}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc2 */ {{1, 1}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc3 */ {{0, 1}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc4 */ {{2, 2}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xc5 */ {{0, 0}, {2, 2}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xc6 */ {{1, 2}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc7 */ {{0, 2}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc8 */ {{3, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xc9 */ {{0, 0}, {3, 3}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xca */ {{1, 1}, {3, 3}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xcb */ {{0, 1}, {3, 3}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xcc */ {{2, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xcd */ {{0, 0}, {2, 3}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xce */ {{1, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xcf */ {{0, 3}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xd0 */ {{4, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xd1 */ {{0, 0}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xd2 */ {{1, 1}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xd3 */ {{0, 1}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xd4 */ {{2, 2}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xd5 */ {{0, 0}, {2, 2}, {4, 4}, {6, 7} } }, {0, 1, 3, 0, /* 0xd6 */ {{1, 2}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xd7 */ {{0, 2}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xd8 */ {{3, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xd9 */ {{0, 0}, {3, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xda */ {{1, 1}, {3, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xdb */ {{0, 1}, {3, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xdc */ {{2, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xdd */ {{0, 0}, {2, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xde */ {{1, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xdf */ {{0, 4}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xe0 */ {{5, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe1 */ {{0, 0}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe2 */ {{1, 1}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe3 */ {{0, 1}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe4 */ {{2, 2}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xe5 */ {{0, 0}, {2, 2}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xe6 */ {{1, 2}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe7 */ {{0, 2}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe8 */ {{3, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xe9 */ {{0, 0}, {3, 3}, {5, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xea */ {{1, 1}, {3, 3}, {5, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xeb */ {{0, 1}, {3, 3}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xec */ {{2, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xed */ {{0, 0}, {2, 3}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xee */ {{1, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xef */ {{0, 3}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xf0 */ {{4, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf1 */ {{0, 0}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xf2 */ {{1, 1}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf3 */ {{0, 1}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xf4 */ {{2, 2}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xf5 */ {{0, 0}, {2, 2}, {4, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xf6 */ {{1, 2}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf7 */ {{0, 2}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xf8 */ {{3, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf9 */ {{0, 0}, {3, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xfa */ {{1, 1}, {3, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xfb */ {{0, 1}, {3, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xfc */ {{2, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xfd */ {{0, 0}, {2, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xfe */ {{1, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 1, 0, /* 0xff */ {{0, 7}, {0, 0}, {0, 0}, {0, 0} } } }; int sctp_is_address_in_scope(struct sctp_ifa *ifa, int ipv4_addr_legal, int ipv6_addr_legal, int loopback_scope, int ipv4_local_scope, int local_scope, int site_scope, int do_update) { if ((loopback_scope == 0) && (ifa->ifn_p) && SCTP_IFN_IS_IFT_LOOP(ifa->ifn_p)) { /* * skip loopback if not in scope * */ return (0); } switch (ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&ifa->address.sin; if (sin->sin_addr.s_addr == 0) { /* not in scope , unspecified */ return (0); } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { /* private address not in scope */ return (0); } } else { return (0); } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; /* * Must update the flags, bummer, which means any * IFA locks must now be applied HERE <-> */ if (do_update) { sctp_gather_internal_ifa_flags(ifa); } if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { return (0); } /* ok to use deprecated addresses? */ sin6 = (struct sockaddr_in6 *)&ifa->address.sin6; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* skip unspecifed addresses */ return (0); } if ( /* (local_scope == 0) && */ (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr))) { return (0); } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { return (0); } } else { return (0); } break; #endif default: return (0); } return (1); } static struct mbuf * sctp_add_addr_to_mbuf(struct mbuf *m, struct sctp_ifa *ifa) { struct sctp_paramhdr *parmh; struct mbuf *mret; int len; if (ifa->address.sa.sa_family == AF_INET) { len = sizeof(struct sctp_ipv4addr_param); } else if (ifa->address.sa.sa_family == AF_INET6) { len = sizeof(struct sctp_ipv6addr_param); } else { /* unknown type */ return (m); } if (M_TRAILINGSPACE(m) >= len) { /* easy side we just drop it on the end */ parmh = (struct sctp_paramhdr *)(SCTP_BUF_AT(m, SCTP_BUF_LEN(m))); mret = m; } else { /* Need more space */ mret = m; while (SCTP_BUF_NEXT(mret) != NULL) { mret = SCTP_BUF_NEXT(mret); } SCTP_BUF_NEXT(mret) = sctp_get_mbuf_for_msg(len, 0, M_DONTWAIT, 1, MT_DATA); if (SCTP_BUF_NEXT(mret) == NULL) { /* We are hosed, can't add more addresses */ return (m); } mret = SCTP_BUF_NEXT(mret); parmh = mtod(mret, struct sctp_paramhdr *); } /* now add the parameter */ switch (ifa->address.sa.sa_family) { case AF_INET: { struct sctp_ipv4addr_param *ipv4p; struct sockaddr_in *sin; sin = (struct sockaddr_in *)&ifa->address.sin; ipv4p = (struct sctp_ipv4addr_param *)parmh; parmh->param_type = htons(SCTP_IPV4_ADDRESS); parmh->param_length = htons(len); ipv4p->addr = sin->sin_addr.s_addr; SCTP_BUF_LEN(mret) += len; break; } #ifdef INET6 case AF_INET6: { struct sctp_ipv6addr_param *ipv6p; struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&ifa->address.sin6; ipv6p = (struct sctp_ipv6addr_param *)parmh; parmh->param_type = htons(SCTP_IPV6_ADDRESS); parmh->param_length = htons(len); memcpy(ipv6p->addr, &sin6->sin6_addr, sizeof(ipv6p->addr)); /* clear embedded scope in the address */ in6_clearscope((struct in6_addr *)ipv6p->addr); SCTP_BUF_LEN(mret) += len; break; } #endif default: return (m); } return (mret); } struct mbuf * sctp_add_addresses_to_i_ia(struct sctp_inpcb *inp, struct sctp_scoping *scope, struct mbuf *m_at, int cnt_inits_to) { struct sctp_vrf *vrf = NULL; int cnt, limit_out = 0, total_count; uint32_t vrf_id; vrf_id = inp->def_vrf_id; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTP_IPI_ADDR_RUNLOCK(); return (m_at); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { struct sctp_ifa *sctp_ifap; struct sctp_ifn *sctp_ifnp; cnt = cnt_inits_to; if (vrf->total_ifa_count > SCTP_COUNT_LIMIT) { limit_out = 1; cnt = SCTP_ADDRESS_LIMIT; goto skip_count; } LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { if ((scope->loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { /* * Skip loopback devices if loopback_scope * not set */ continue; } LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { if (sctp_is_address_in_scope(sctp_ifap, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 1) == 0) { continue; } cnt++; if (cnt > SCTP_ADDRESS_LIMIT) { break; } } if (cnt > SCTP_ADDRESS_LIMIT) { break; } } skip_count: if (cnt > 1) { total_count = 0; LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { cnt = 0; if ((scope->loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { /* * Skip loopback devices if * loopback_scope not set */ continue; } LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { if (sctp_is_address_in_scope(sctp_ifap, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 0) == 0) { continue; } m_at = sctp_add_addr_to_mbuf(m_at, sctp_ifap); if (limit_out) { cnt++; total_count++; if (cnt >= 2) { /* * two from each * address */ break; } if (total_count > SCTP_ADDRESS_LIMIT) { /* No more addresses */ break; } } } } } } else { struct sctp_laddr *laddr; cnt = cnt_inits_to; /* First, how many ? */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) /* * Address being deleted by the system, dont * list. */ continue; if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* * Address being deleted on this ep don't * list. */ continue; } if (sctp_is_address_in_scope(laddr->ifa, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 1) == 0) { continue; } cnt++; } if (cnt > SCTP_ADDRESS_LIMIT) { limit_out = 1; } /* * To get through a NAT we only list addresses if we have * more than one. That way if you just bind a single address * we let the source of the init dictate our address. */ if (cnt > 1) { LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { cnt = 0; if (laddr->ifa == NULL) { continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) continue; if (sctp_is_address_in_scope(laddr->ifa, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 0) == 0) { continue; } m_at = sctp_add_addr_to_mbuf(m_at, laddr->ifa); cnt++; if (cnt >= SCTP_ADDRESS_LIMIT) { break; } } } } SCTP_IPI_ADDR_RUNLOCK(); return (m_at); } static struct sctp_ifa * sctp_is_ifa_addr_preferred(struct sctp_ifa *ifa, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { uint8_t dest_is_global = 0; /* dest_is_priv is true if destination is a private address */ /* dest_is_loop is true if destination is a loopback addresses */ /** * Here we determine if its a preferred address. A preferred address * means it is the same scope or higher scope then the destination. * L = loopback, P = private, G = global * ----------------------------------------- * src | dest | result * ---------------------------------------- * L | L | yes * ----------------------------------------- * P | L | yes-v4 no-v6 * ----------------------------------------- * G | L | yes-v4 no-v6 * ----------------------------------------- * L | P | no * ----------------------------------------- * P | P | yes * ----------------------------------------- * G | P | no * ----------------------------------------- * L | G | no * ----------------------------------------- * P | G | no * ----------------------------------------- * G | G | yes * ----------------------------------------- */ if (ifa->address.sa.sa_family != fam) { /* forget mis-matched family */ return (NULL); } if ((dest_is_priv == 0) && (dest_is_loop == 0)) { dest_is_global = 1; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "Is destination preferred:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &ifa->address.sa); /* Ok the address may be ok */ if (fam == AF_INET6) { /* ok to use deprecated addresses? no lets not! */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:1\n"); return (NULL); } if (ifa->src_is_priv && !ifa->src_is_loop) { if (dest_is_loop) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:2\n"); return (NULL); } } if (ifa->src_is_glob) { if (dest_is_loop) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:3\n"); return (NULL); } } } /* * Now that we know what is what, implement or table this could in * theory be done slicker (it used to be), but this is * straightforward and easier to validate :-) */ SCTPDBG(SCTP_DEBUG_OUTPUT3, "src_loop:%d src_priv:%d src_glob:%d\n", ifa->src_is_loop, ifa->src_is_priv, ifa->src_is_glob); SCTPDBG(SCTP_DEBUG_OUTPUT3, "dest_loop:%d dest_priv:%d dest_glob:%d\n", dest_is_loop, dest_is_priv, dest_is_global); if ((ifa->src_is_loop) && (dest_is_priv)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:4\n"); return (NULL); } if ((ifa->src_is_glob) && (dest_is_priv)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:5\n"); return (NULL); } if ((ifa->src_is_loop) && (dest_is_global)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:6\n"); return (NULL); } if ((ifa->src_is_priv) && (dest_is_global)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:7\n"); return (NULL); } SCTPDBG(SCTP_DEBUG_OUTPUT3, "YES\n"); /* its a preferred address */ return (ifa); } static struct sctp_ifa * sctp_is_ifa_addr_acceptable(struct sctp_ifa *ifa, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { uint8_t dest_is_global = 0; /* * Here we determine if its a acceptable address. A acceptable * address means it is the same scope or higher scope but we can * allow for NAT which means its ok to have a global dest and a * private src. * * L = loopback, P = private, G = global * ----------------------------------------- src | dest | result * ----------------------------------------- L | L | yes * ----------------------------------------- P | L | * yes-v4 no-v6 ----------------------------------------- G | * L | yes ----------------------------------------- L | * P | no ----------------------------------------- P | P * | yes ----------------------------------------- G | P * | yes - May not work ----------------------------------------- * L | G | no ----------------------------------------- P * | G | yes - May not work * ----------------------------------------- G | G | yes * ----------------------------------------- */ if (ifa->address.sa.sa_family != fam) { /* forget non matching family */ return (NULL); } /* Ok the address may be ok */ if ((dest_is_loop == 0) && (dest_is_priv == 0)) { dest_is_global = 1; } if (fam == AF_INET6) { /* ok to use deprecated addresses? */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { return (NULL); } if (ifa->src_is_priv) { /* Special case, linklocal to loop */ if (dest_is_loop) return (NULL); } } /* * Now that we know what is what, implement our table. This could in * theory be done slicker (it used to be), but this is * straightforward and easier to validate :-) */ if ((ifa->src_is_loop == 1) && (dest_is_priv)) { return (NULL); } if ((ifa->src_is_loop == 1) && (dest_is_global)) { return (NULL); } /* its an acceptable address */ return (ifa); } int sctp_is_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; if (stcb == NULL) { /* There are no restrictions, no TCB :-) */ return (0); } LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa == ifa) { /* Yes it is on the list */ return (1); } } return (0); } int sctp_is_addr_in_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; if (ifa == NULL) return (0); LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __FUNCTION__); continue; } if ((laddr->ifa == ifa) && laddr->action == 0) /* same pointer */ return (1); } return (0); } static struct sctp_ifa * sctp_choose_boundspecific_inp(struct sctp_inpcb *inp, sctp_route_t * ro, uint32_t vrf_id, int non_asoc_addr_ok, uint8_t dest_is_priv, uint8_t dest_is_loop, sa_family_t fam) { struct sctp_laddr *laddr, *starting_point; void *ifn; int resettotop = 0; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa, *sifa; struct sctp_vrf *vrf; uint32_t ifn_index; vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); sctp_ifn = sctp_find_ifn(ifn, ifn_index); /* * first question, is the ifn we will emit on in our list, if so, we * want such an address. Note that we first looked for a preferred * address. */ if (sctp_ifn) { /* is a preferred one on the interface we route out? */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_preferred(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (sctp_is_addr_in_ep(inp, sifa)) { atomic_add_int(&sifa->refcount, 1); return (sifa); } } } /* * ok, now we now need to find one on the list of the addresses. We * can't get one on the emitting interface so let's find first a * preferred one. If not that an acceptable one otherwise... we * return NULL. */ starting_point = inp->next_addr_touse; once_again: if (inp->next_addr_touse == NULL) { inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); resettotop = 1; } for (laddr = inp->next_addr_touse; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (resettotop == 0) { inp->next_addr_touse = NULL; goto once_again; } inp->next_addr_touse = starting_point; resettotop = 0; once_again_too: if (inp->next_addr_touse == NULL) { inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); resettotop = 1; } /* ok, what about an acceptable address in the inp */ for (laddr = inp->next_addr_touse; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (resettotop == 0) { inp->next_addr_touse = NULL; goto once_again_too; } /* * no address bound can be a source for the destination we are in * trouble */ return (NULL); } static struct sctp_ifa * sctp_choose_boundspecific_stcb(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, sctp_route_t * ro, uint32_t vrf_id, uint8_t dest_is_priv, uint8_t dest_is_loop, int non_asoc_addr_ok, sa_family_t fam) { struct sctp_laddr *laddr, *starting_point; void *ifn; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa, *sifa; uint8_t start_at_beginning = 0; struct sctp_vrf *vrf; uint32_t ifn_index; /* * first question, is the ifn we will emit on in our list, if so, we * want that one. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); sctp_ifn = sctp_find_ifn(ifn, ifn_index); /* * first question, is the ifn we will emit on in our list? If so, * we want that one. First we look for a preferred. Second, we go * for an acceptable. */ if (sctp_ifn) { /* first try for a preferred address on the ep */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; if (sctp_is_addr_in_ep(inp, sctp_ifa)) { sifa = sctp_is_ifa_addr_preferred(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } atomic_add_int(&sifa->refcount, 1); return (sifa); } } /* next try for an acceptable address on the ep */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; if (sctp_is_addr_in_ep(inp, sctp_ifa)) { sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } atomic_add_int(&sifa->refcount, 1); return (sifa); } } } /* * if we can't find one like that then we must look at all addresses * bound to pick one at first preferable then secondly acceptable. */ starting_point = stcb->asoc.last_used_address; sctp_from_the_top: if (stcb->asoc.last_used_address == NULL) { start_at_beginning = 1; stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); } /* search beginning with the last used address */ for (laddr = stcb->asoc.last_used_address; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } stcb->asoc.last_used_address = laddr; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (start_at_beginning == 0) { stcb->asoc.last_used_address = NULL; goto sctp_from_the_top; } /* now try for any higher scope than the destination */ stcb->asoc.last_used_address = starting_point; start_at_beginning = 0; sctp_from_the_top2: if (stcb->asoc.last_used_address == NULL) { start_at_beginning = 1; stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); } /* search beginning with the last used address */ for (laddr = stcb->asoc.last_used_address; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } stcb->asoc.last_used_address = laddr; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (start_at_beginning == 0) { stcb->asoc.last_used_address = NULL; goto sctp_from_the_top2; } return (NULL); } static struct sctp_ifa * sctp_select_nth_preferred_addr_from_ifn_boundall(struct sctp_ifn *ifn, struct sctp_tcb *stcb, int non_asoc_addr_ok, uint8_t dest_is_loop, uint8_t dest_is_priv, int addr_wanted, sa_family_t fam, sctp_route_t * ro ) { struct sctp_ifa *ifa, *sifa; int num_eligible_addr = 0; #ifdef INET6 struct sockaddr_in6 sin6, lsa6; if (fam == AF_INET6) { memcpy(&sin6, &ro->ro_dst, sizeof(struct sockaddr_in6)); (void)sa6_recoverscope(&sin6); } #endif /* INET6 */ LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; #ifdef INET6 if (fam == AF_INET6 && dest_is_loop && sifa->src_is_loop && sifa->src_is_priv) { /* * don't allow fe80::1 to be a src on loop ::1, we * don't list it to the peer so we will get an * abort. */ continue; } if (fam == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&sifa->address.sin6.sin6_addr) && IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { /* * link-local <-> link-local must belong to the same * scope. */ memcpy(&lsa6, &sifa->address.sin6, sizeof(struct sockaddr_in6)); (void)sa6_recoverscope(&lsa6); if (sin6.sin6_scope_id != lsa6.sin6_scope_id) { continue; } } #endif /* INET6 */ /* * Check if the IPv6 address matches to next-hop. In the * mobile case, old IPv6 address may be not deleted from the * interface. Then, the interface has previous and new * addresses. We should use one corresponding to the * next-hop. (by micchie) */ #ifdef INET6 if (stcb && fam == AF_INET6 && sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { if (sctp_v6src_match_nexthop(&sifa->address.sin6, ro) == 0) { continue; } } #endif /* Avoid topologically incorrect IPv4 address */ if (stcb && fam == AF_INET && sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { if (sctp_v4src_match_nexthop(sifa, ro) == 0) { continue; } } if (stcb) { if (sctp_is_address_in_scope(ifa, stcb->asoc.ipv4_addr_legal, stcb->asoc.ipv6_addr_legal, stcb->asoc.loopback_scope, stcb->asoc.ipv4_local_scope, stcb->asoc.local_scope, stcb->asoc.site_scope, 0) == 0) { continue; } if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } if (num_eligible_addr >= addr_wanted) { return (sifa); } num_eligible_addr++; } return (NULL); } static int sctp_count_num_preferred_boundall(struct sctp_ifn *ifn, struct sctp_tcb *stcb, int non_asoc_addr_ok, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { struct sctp_ifa *ifa, *sifa; int num_eligible_addr = 0; LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) { continue; } sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) { continue; } if (stcb) { if (sctp_is_address_in_scope(ifa, stcb->asoc.ipv4_addr_legal, stcb->asoc.ipv6_addr_legal, stcb->asoc.loopback_scope, stcb->asoc.ipv4_local_scope, stcb->asoc.local_scope, stcb->asoc.site_scope, 0) == 0) { continue; } if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } num_eligible_addr++; } return (num_eligible_addr); } static struct sctp_ifa * sctp_choose_boundall(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, sctp_route_t * ro, uint32_t vrf_id, uint8_t dest_is_priv, uint8_t dest_is_loop, int non_asoc_addr_ok, sa_family_t fam) { int cur_addr_num = 0, num_preferred = 0; void *ifn; struct sctp_ifn *sctp_ifn, *looked_at = NULL, *emit_ifn; struct sctp_ifa *sctp_ifa, *sifa; uint32_t ifn_index; struct sctp_vrf *vrf; /*- * For boundall we can use any address in the association. * If non_asoc_addr_ok is set we can use any address (at least in * theory). So we look for preferred addresses first. If we find one, * we use it. Otherwise we next try to get an address on the * interface, which we should be able to do (unless non_asoc_addr_ok * is false and we are routed out that way). In these cases where we * can't use the address of the interface we go through all the * ifn's looking for an address we can use and fill that in. Punting * means we send back address 0, which will probably cause problems * actually since then IP will fill in the address of the route ifn, * which means we probably already rejected it.. i.e. here comes an * abort :-<. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); emit_ifn = looked_at = sctp_ifn = sctp_find_ifn(ifn, ifn_index); if (sctp_ifn == NULL) { /* ?? We don't have this guy ?? */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "No ifn emit interface?\n"); goto bound_all_plan_b; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "ifn_index:%d name:%s is emit interface\n", ifn_index, sctp_ifn->ifn_name); if (net) { cur_addr_num = net->indx_of_eligible_next_to_use; } num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, fam); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Found %d preferred source addresses for intf:%s\n", num_preferred, sctp_ifn->ifn_name); if (num_preferred == 0) { /* * no eligible addresses, we must use some other interface * address if we can find one. */ goto bound_all_plan_b; } /* * Ok we have num_eligible_addr set with how many we can use, this * may vary from call to call due to addresses being deprecated * etc.. */ if (cur_addr_num >= num_preferred) { cur_addr_num = 0; } /* * select the nth address from the list (where cur_addr_num is the * nth) and 0 is the first one, 1 is the second one etc... */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "cur_addr_num:%d\n", cur_addr_num); sctp_ifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, cur_addr_num, fam, ro); /* if sctp_ifa is NULL something changed??, fall to plan b. */ if (sctp_ifa) { atomic_add_int(&sctp_ifa->refcount, 1); if (net) { /* save off where the next one we will want */ net->indx_of_eligible_next_to_use = cur_addr_num + 1; } return (sctp_ifa); } /* * plan_b: Look at all interfaces and find a preferred address. If * no preferred fall through to plan_c. */ bound_all_plan_b: SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan B\n"); LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "Examine interface %s\n", sctp_ifn->ifn_name); if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* wrong base scope */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "skip\n"); continue; } if ((sctp_ifn == looked_at) && looked_at) { /* already looked at this guy */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "already seen\n"); continue; } num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, fam); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Found ifn:%p %d preferred source addresses\n", ifn, num_preferred); if (num_preferred == 0) { /* None on this interface. */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "No prefered -- skipping to next\n"); continue; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "num preferred:%d on interface:%p cur_addr_num:%d\n", num_preferred, sctp_ifn, cur_addr_num); /* * Ok we have num_eligible_addr set with how many we can * use, this may vary from call to call due to addresses * being deprecated etc.. */ if (cur_addr_num >= num_preferred) { cur_addr_num = 0; } sifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, cur_addr_num, fam, ro); if (sifa == NULL) continue; if (net) { net->indx_of_eligible_next_to_use = cur_addr_num + 1; SCTPDBG(SCTP_DEBUG_OUTPUT2, "we selected %d\n", cur_addr_num); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Source:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Dest:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &net->ro._l_addr.sa); } atomic_add_int(&sifa->refcount, 1); return (sifa); } /* plan_c: do we have an acceptable address on the emit interface */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan C: find acceptable on interface\n"); if (emit_ifn == NULL) { goto plan_d; } LIST_FOREACH(sctp_ifa, &emit_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (stcb) { if (sctp_is_address_in_scope(sifa, stcb->asoc.ipv4_addr_legal, stcb->asoc.ipv6_addr_legal, stcb->asoc.loopback_scope, stcb->asoc.ipv4_local_scope, stcb->asoc.local_scope, stcb->asoc.site_scope, 0) == 0) { continue; } if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } atomic_add_int(&sifa->refcount, 1); return (sifa); } plan_d: /* * plan_d: We are in trouble. No preferred address on the emit * interface. And not even a preferred address on all interfaces. Go * out and see if we can find an acceptable address somewhere * amongst all interfaces. */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan D\n"); LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* wrong base scope */ continue; } if ((sctp_ifn == looked_at) && looked_at) /* already looked at this guy */ continue; LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (stcb) { if (sctp_is_address_in_scope(sifa, stcb->asoc.ipv4_addr_legal, stcb->asoc.ipv6_addr_legal, stcb->asoc.loopback_scope, stcb->asoc.ipv4_local_scope, stcb->asoc.local_scope, stcb->asoc.site_scope, 0) == 0) { continue; } if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some * reason.. probably not yet added. */ continue; } } atomic_add_int(&sifa->refcount, 1); return (sifa); } } /* * Ok we can find NO address to source from that is not on our * restricted list and non_asoc_address is NOT ok, or it is on our * restricted list. We can't source to it :-( */ return (NULL); } /* tcb may be NULL */ struct sctp_ifa * sctp_source_address_selection(struct sctp_inpcb *inp, struct sctp_tcb *stcb, sctp_route_t * ro, struct sctp_nets *net, int non_asoc_addr_ok, uint32_t vrf_id) { struct sockaddr_in *to = (struct sockaddr_in *)&ro->ro_dst; #ifdef INET6 struct sockaddr_in6 *to6 = (struct sockaddr_in6 *)&ro->ro_dst; #endif struct sctp_ifa *answer; uint8_t dest_is_priv, dest_is_loop; sa_family_t fam; /*- * Rules: - Find the route if needed, cache if I can. - Look at * interface address in route, Is it in the bound list. If so we * have the best source. - If not we must rotate amongst the * addresses. * * Cavets and issues * * Do we need to pay attention to scope. We can have a private address * or a global address we are sourcing or sending to. So if we draw * it out * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * For V4 * ------------------------------------------ * source * dest * result * ----------------------------------------- * Private * Global * NAT * ----------------------------------------- * Private * Private * No problem * ----------------------------------------- * Global * Private * Huh, How will this work? * ----------------------------------------- * Global * Global * No Problem *------------------------------------------ * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * For V6 *------------------------------------------ * source * dest * result * ----------------------------------------- * Linklocal * Global * * ----------------------------------------- * Linklocal * Linklocal * No problem * ----------------------------------------- * Global * Linklocal * Huh, How will this work? * ----------------------------------------- * Global * Global * No Problem *------------------------------------------ * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * * And then we add to that what happens if there are multiple addresses * assigned to an interface. Remember the ifa on a ifn is a linked * list of addresses. So one interface can have more than one IP * address. What happens if we have both a private and a global * address? Do we then use context of destination to sort out which * one is best? And what about NAT's sending P->G may get you a NAT * translation, or should you select the G thats on the interface in * preference. * * Decisions: * * - count the number of addresses on the interface. * - if it is one, no problem except case . * For we will assume a NAT out there. * - if there are more than one, then we need to worry about scope P * or G. We should prefer G -> G and P -> P if possible. * Then as a secondary fall back to mixed types G->P being a last * ditch one. * - The above all works for bound all, but bound specific we need to * use the same concept but instead only consider the bound * addresses. If the bound set is NOT assigned to the interface then * we must use rotation amongst the bound addresses.. */ if (ro->ro_rt == NULL) { /* * Need a route to cache. */ SCTP_RTALLOC(ro, vrf_id); } if (ro->ro_rt == NULL) { return (NULL); } fam = to->sin_family; dest_is_priv = dest_is_loop = 0; /* Setup our scopes for the destination */ switch (fam) { case AF_INET: /* Scope based on outbound address */ if (IN4_ISLOOPBACK_ADDRESS(&to->sin_addr)) { dest_is_loop = 1; if (net != NULL) { /* mark it as local */ net->addr_is_local = 1; } } else if ((IN4_ISPRIVATE_ADDRESS(&to->sin_addr))) { dest_is_priv = 1; } break; #ifdef INET6 case AF_INET6: /* Scope based on outbound address */ if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr) || SCTP_ROUTE_IS_REAL_LOOP(ro)) { /* * If the address is a loopback address, which * consists of "::1" OR "fe80::1%lo0", we are * loopback scope. But we don't use dest_is_priv * (link local addresses). */ dest_is_loop = 1; if (net != NULL) { /* mark it as local */ net->addr_is_local = 1; } } else if (IN6_IS_ADDR_LINKLOCAL(&to6->sin6_addr)) { dest_is_priv = 1; } break; #endif } SCTPDBG(SCTP_DEBUG_OUTPUT2, "Select source addr for:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)to); SCTP_IPI_ADDR_RLOCK(); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* * Bound all case */ answer = sctp_choose_boundall(inp, stcb, net, ro, vrf_id, dest_is_priv, dest_is_loop, non_asoc_addr_ok, fam); SCTP_IPI_ADDR_RUNLOCK(); return (answer); } /* * Subset bound case */ if (stcb) { answer = sctp_choose_boundspecific_stcb(inp, stcb, net, ro, vrf_id, dest_is_priv, dest_is_loop, non_asoc_addr_ok, fam); } else { answer = sctp_choose_boundspecific_inp(inp, ro, vrf_id, non_asoc_addr_ok, dest_is_priv, dest_is_loop, fam); } SCTP_IPI_ADDR_RUNLOCK(); return (answer); } static int sctp_find_cmsg(int c_type, void *data, struct mbuf *control, int cpsize) { struct cmsghdr cmh; int tlen, at; tlen = SCTP_BUF_LEN(control); at = 0; /* * Independent of how many mbufs, find the c_type inside the control * structure and copy out the data. */ while (at < tlen) { if ((tlen - at) < (int)CMSG_ALIGN(sizeof(cmh))) { /* not enough room for one more we are done. */ return (0); } m_copydata(control, at, sizeof(cmh), (caddr_t)&cmh); if (((int)cmh.cmsg_len + at) > tlen) { /* * this is real messed up since there is not enough * data here to cover the cmsg header. We are done. */ return (0); } if ((cmh.cmsg_level == IPPROTO_SCTP) && (c_type == cmh.cmsg_type)) { /* found the one we want, copy it out */ at += CMSG_ALIGN(sizeof(struct cmsghdr)); if ((int)(cmh.cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr))) < cpsize) { /* * space of cmsg_len after header not big * enough */ return (0); } m_copydata(control, at, cpsize, data); return (1); } else { at += CMSG_ALIGN(cmh.cmsg_len); if (cmh.cmsg_len == 0) { break; } } } /* not found */ return (0); } static struct mbuf * sctp_add_cookie(struct sctp_inpcb *inp, struct mbuf *init, int init_offset, struct mbuf *initack, int initack_offset, struct sctp_state_cookie *stc_in, uint8_t ** signature) { struct mbuf *copy_init, *copy_initack, *m_at, *sig, *mret; struct sctp_state_cookie *stc; struct sctp_paramhdr *ph; uint8_t *foo; int sig_offset; uint16_t cookie_sz; mret = NULL; mret = sctp_get_mbuf_for_msg((sizeof(struct sctp_state_cookie) + sizeof(struct sctp_paramhdr)), 0, M_DONTWAIT, 1, MT_DATA); if (mret == NULL) { return (NULL); } copy_init = SCTP_M_COPYM(init, init_offset, M_COPYALL, M_DONTWAIT); if (copy_init == NULL) { sctp_m_freem(mret); return (NULL); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = copy_init; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif copy_initack = SCTP_M_COPYM(initack, initack_offset, M_COPYALL, M_DONTWAIT); if (copy_initack == NULL) { sctp_m_freem(mret); sctp_m_freem(copy_init); return (NULL); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = copy_initack; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif /* easy side we just drop it on the end */ ph = mtod(mret, struct sctp_paramhdr *); SCTP_BUF_LEN(mret) = sizeof(struct sctp_state_cookie) + sizeof(struct sctp_paramhdr); stc = (struct sctp_state_cookie *)((caddr_t)ph + sizeof(struct sctp_paramhdr)); ph->param_type = htons(SCTP_STATE_COOKIE); ph->param_length = 0; /* fill in at the end */ /* Fill in the stc cookie data */ memcpy(stc, stc_in, sizeof(struct sctp_state_cookie)); /* tack the INIT and then the INIT-ACK onto the chain */ cookie_sz = 0; m_at = mret; for (m_at = mret; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { SCTP_BUF_NEXT(m_at) = copy_init; break; } } for (m_at = copy_init; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { SCTP_BUF_NEXT(m_at) = copy_initack; break; } } for (m_at = copy_initack; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { break; } } sig = sctp_get_mbuf_for_msg(SCTP_SECRET_SIZE, 0, M_DONTWAIT, 1, MT_DATA); if (sig == NULL) { /* no space, so free the entire chain */ sctp_m_freem(mret); return (NULL); } SCTP_BUF_LEN(sig) = 0; SCTP_BUF_NEXT(m_at) = sig; sig_offset = 0; foo = (uint8_t *) (mtod(sig, caddr_t)+sig_offset); memset(foo, 0, SCTP_SIGNATURE_SIZE); *signature = foo; SCTP_BUF_LEN(sig) += SCTP_SIGNATURE_SIZE; cookie_sz += SCTP_SIGNATURE_SIZE; ph->param_length = htons(cookie_sz); return (mret); } static uint8_t sctp_get_ect(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk) { - uint8_t this_random; - - /* Huh? */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 0) return (0); - if (SCTP_BASE_SYSCTL(sctp_ecn_nonce) == 0) - /* no nonce, always return ECT0 */ - return (SCTP_ECT0_BIT); - - if (stcb->asoc.peer_supports_ecn_nonce == 0) { - /* Peer does NOT support it, so we send a ECT0 only */ - return (SCTP_ECT0_BIT); - } - if (chk == NULL) - return (SCTP_ECT0_BIT); - - if ((stcb->asoc.hb_random_idx > 3) || - ((stcb->asoc.hb_random_idx == 3) && - (stcb->asoc.hb_ect_randombit > 7))) { - uint32_t rndval; - -warp_drive_sa: - rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); - memcpy(stcb->asoc.hb_random_values, &rndval, - sizeof(stcb->asoc.hb_random_values)); - this_random = stcb->asoc.hb_random_values[0]; - stcb->asoc.hb_random_idx = 0; - stcb->asoc.hb_ect_randombit = 0; - } else { - if (stcb->asoc.hb_ect_randombit > 7) { - stcb->asoc.hb_ect_randombit = 0; - stcb->asoc.hb_random_idx++; - if (stcb->asoc.hb_random_idx > 3) { - goto warp_drive_sa; - } - } - this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; - } - if ((this_random >> stcb->asoc.hb_ect_randombit) & 0x01) { - if (chk != NULL) - /* ECN Nonce stuff */ - chk->rec.data.ect_nonce = SCTP_ECT1_BIT; - stcb->asoc.hb_ect_randombit++; - return (SCTP_ECT1_BIT); - } else { - stcb->asoc.hb_ect_randombit++; - return (SCTP_ECT0_BIT); - } + return (SCTP_ECT0_BIT); } static int sctp_lowlevel_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, /* may be NULL */ struct sctp_nets *net, struct sockaddr *to, struct mbuf *m, uint32_t auth_offset, struct sctp_auth_chunk *auth, uint16_t auth_keyid, int nofragment_flag, int ecn_ok, struct sctp_tmit_chunk *chk, int out_of_asoc_ok, uint16_t src_port, uint16_t dest_port, uint32_t v_tag, uint16_t port, int so_locked, #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif union sctp_sockstore *over_addr ) /* nofragment_flag to tell if IP_DF should be set (IPv4 only) */ { /* * Given a mbuf chain (via SCTP_BUF_NEXT()) that holds a packet * header WITH an SCTPHDR but no IP header, endpoint inp and sa * structure: - fill in the HMAC digest of any AUTH chunk in the * packet. - calculate and fill in the SCTP checksum. - prepend an * IP address header. - if boundall use INADDR_ANY. - if * boundspecific do source address selection. - set fragmentation * option for ipV4. - On return from IP output, check/adjust mtu * size of output interface and smallest_mtu size as well. */ /* Will need ifdefs around this */ struct mbuf *o_pak; struct mbuf *newm; struct sctphdr *sctphdr; int packet_length; int ret; uint32_t vrf_id; sctp_route_t *ro = NULL; struct udphdr *udp = NULL; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so = NULL; #endif if ((net) && (net->dest_state & SCTP_ADDR_OUT_OF_SCOPE)) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); sctp_m_freem(m); return (EFAULT); } if (stcb) { vrf_id = stcb->asoc.vrf_id; } else { vrf_id = inp->def_vrf_id; } /* fill in the HMAC digest for any AUTH chunk in the packet */ if ((auth != NULL) && (stcb != NULL)) { sctp_fill_hmac_digest_m(m, auth_offset, auth, stcb, auth_keyid); } if (to->sa_family == AF_INET) { struct ip *ip = NULL; sctp_route_t iproute; uint8_t tos_value; int len; len = sizeof(struct ip) + sizeof(struct sctphdr); if (port) { len += sizeof(struct udphdr); } newm = sctp_get_mbuf_for_msg(len, 1, M_DONTWAIT, 1, MT_DATA); if (newm == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_ALIGN_TO_END(newm, len); SCTP_BUF_LEN(newm) = len; SCTP_BUF_NEXT(newm) = m; m = newm; packet_length = sctp_calculate_len(m); ip = mtod(m, struct ip *); ip->ip_v = IPVERSION; ip->ip_hl = (sizeof(struct ip) >> 2); if (net) { tos_value = net->tos_flowlabel & 0x000000ff; } else { tos_value = inp->ip_inp.inp.inp_ip_tos; } if ((nofragment_flag) && (port == 0)) { ip->ip_off = IP_DF; } else ip->ip_off = 0; /* FreeBSD has a function for ip_id's */ ip->ip_id = ip_newid(); ip->ip_ttl = inp->ip_inp.inp.inp_ip_ttl; ip->ip_len = packet_length; if (stcb) { if ((stcb->asoc.ecn_allowed) && ecn_ok) { /* Enable ECN */ ip->ip_tos = ((u_char)(tos_value & 0xfc) | sctp_get_ect(stcb, chk)); } else { /* No ECN */ ip->ip_tos = (u_char)(tos_value & 0xfc); } } else { /* no association at all */ ip->ip_tos = (tos_value & 0xfc); } if (port) { ip->ip_p = IPPROTO_UDP; } else { ip->ip_p = IPPROTO_SCTP; } ip->ip_sum = 0; if (net == NULL) { ro = &iproute; memset(&iproute, 0, sizeof(iproute)); memcpy(&ro->ro_dst, to, to->sa_len); } else { ro = (sctp_route_t *) & net->ro; } /* Now the address selection part */ ip->ip_dst.s_addr = ((struct sockaddr_in *)to)->sin_addr.s_addr; /* call the routine to select the src address */ if (net && out_of_asoc_ok == 0) { if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED | SCTP_ADDR_IFA_UNUSEABLE))) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (net->src_addr_selected == 0) { /* Cache the source address */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, ro, net, 0, vrf_id); net->src_addr_selected = 1; } if (net->ro._s_addr == NULL) { /* No route to host */ net->src_addr_selected = 0; goto no_route; } ip->ip_src = net->ro._s_addr->address.sin.sin_addr; } else { if (over_addr == NULL) { struct sctp_ifa *_lsrc; _lsrc = sctp_source_address_selection(inp, stcb, ro, net, out_of_asoc_ok, vrf_id); if (_lsrc == NULL) { goto no_route; } ip->ip_src = _lsrc->address.sin.sin_addr; sctp_free_ifa(_lsrc); } else { ip->ip_src = over_addr->sin.sin_addr; SCTP_RTALLOC(ro, vrf_id); } } if (port) { udp = (struct udphdr *)((caddr_t)ip + sizeof(struct ip)); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(packet_length - sizeof(struct ip)); udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); sctphdr = (struct sctphdr *)((caddr_t)udp + sizeof(struct udphdr)); } else { sctphdr = (struct sctphdr *)((caddr_t)ip + sizeof(struct ip)); } sctphdr->src_port = src_port; sctphdr->dest_port = dest_port; sctphdr->v_tag = v_tag; sctphdr->checksum = 0; /* * If source address selection fails and we find no route * then the ip_output should fail as well with a * NO_ROUTE_TO_HOST type error. We probably should catch * that somewhere and abort the association right away * (assuming this is an INIT being sent). */ if ((ro->ro_rt == NULL)) { /* * src addr selection failed to find a route (or * valid source addr), so we can't get there from * here (yet)! */ no_route: SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: dropped packet - no valid source addr\n", __FUNCTION__); if (net) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Destination was "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT1, &net->ro._l_addr.sa); if (net->dest_state & SCTP_ADDR_CONFIRMED) { if ((net->dest_state & SCTP_ADDR_REACHABLE) && stcb) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "no route takes interface %p down\n", net); sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, SCTP_FAILED_THRESHOLD, (void *)net, so_locked); net->dest_state &= ~SCTP_ADDR_REACHABLE; net->dest_state |= SCTP_ADDR_NOT_REACHABLE; /* * JRS 5/14/07 - If a * destination is * unreachable, the PF bit * is turned off. This * allows an unambiguous use * of the PF bit for * destinations that are * reachable but potentially * failed. If the * destination is set to the * unreachable state, also * set the destination to * the PF state. */ /* * Add debug message here if * destination is not in PF * state. */ /* * Stop any running T3 * timers here? */ if ((stcb->asoc.sctp_cmt_on_off > 0) && (stcb->asoc.sctp_cmt_pf > 0)) { net->dest_state &= ~SCTP_ADDR_PF; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Destination %p moved from PF to unreachable.\n", net); } } } if (stcb) { if (net == stcb->asoc.primary_destination) { /* need a new primary */ struct sctp_nets *alt; alt = sctp_find_alternate_net(stcb, net, 0); if (alt != net) { if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, alt) == 0) { net->dest_state |= SCTP_ADDR_WAS_PRIMARY; if (net->ro._s_addr) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } } } } } SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); sctp_m_freem(m); return (EHOSTUNREACH); } if (ro != &iproute) { memcpy(&iproute, ro, sizeof(*ro)); } SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv4 output routine from low level src addr:%x\n", (uint32_t) (ntohl(ip->ip_src.s_addr))); SCTPDBG(SCTP_DEBUG_OUTPUT3, "Destination is %x\n", (uint32_t) (ntohl(ip->ip_dst.s_addr))); SCTPDBG(SCTP_DEBUG_OUTPUT3, "RTP route is %p through\n", ro->ro_rt); if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* failed to prepend data, give up */ SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); sctp_m_freem(m); return (ENOMEM); } #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(m, packet_length); #endif SCTP_ATTACH_CHAIN(o_pak, m, packet_length); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else if (!(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && (stcb) && (stcb->asoc.loopback_scope))) { sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip) + sizeof(struct udphdr)); SCTP_STAT_INCR(sctps_sendswcrc); } else { SCTP_STAT_INCR(sctps_sendnocrc); } #endif SCTP_ENABLE_UDP_CSUM(o_pak); } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else m->m_pkthdr.csum_flags = CSUM_SCTP; m->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } /* send it out. table id is taken from stcb */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { so = SCTP_INP_SO(inp); SCTP_SOCKET_UNLOCK(so, 0); } #endif SCTP_IP_OUTPUT(ret, o_pak, ro, stcb, vrf_id); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 0); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); if (ret) SCTP_STAT_INCR(sctps_senderrors); SCTPDBG(SCTP_DEBUG_OUTPUT3, "IP output returns %d\n", ret); if (net == NULL) { /* free tempy routes */ if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } else { /* PMTU check versus smallest asoc MTU goes here */ if ((ro->ro_rt != NULL) && (net->ro._s_addr)) { uint32_t mtu; mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); if (net->port) { mtu -= sizeof(struct udphdr); } if (mtu && (stcb->asoc.smallest_mtu > mtu)) { sctp_mtu_size_reset(inp, &stcb->asoc, mtu); net->mtu = mtu; } } else if (ro->ro_rt == NULL) { /* route was freed */ if (net->ro._s_addr && net->src_addr_selected) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } } return (ret); } #ifdef INET6 else if (to->sa_family == AF_INET6) { uint32_t flowlabel; struct ip6_hdr *ip6h; struct route_in6 ip6route; struct ifnet *ifp; u_char flowTop; uint16_t flowBottom; u_char tosBottom, tosTop; struct sockaddr_in6 *sin6, tmp, *lsa6, lsa6_tmp; int prev_scope = 0; struct sockaddr_in6 lsa6_storage; int error; u_short prev_port = 0; int len; if (net != NULL) { flowlabel = net->tos_flowlabel; } else { flowlabel = ((struct in6pcb *)inp)->in6p_flowinfo; } len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr); if (port) { len += sizeof(struct udphdr); } newm = sctp_get_mbuf_for_msg(len, 1, M_DONTWAIT, 1, MT_DATA); if (newm == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_ALIGN_TO_END(newm, len); SCTP_BUF_LEN(newm) = len; SCTP_BUF_NEXT(newm) = m; m = newm; packet_length = sctp_calculate_len(m); ip6h = mtod(m, struct ip6_hdr *); /* * We assume here that inp_flow is in host byte order within * the TCB! */ flowBottom = flowlabel & 0x0000ffff; flowTop = ((flowlabel & 0x000f0000) >> 16); tosTop = (((flowlabel & 0xf0) >> 4) | IPV6_VERSION); /* protect *sin6 from overwrite */ sin6 = (struct sockaddr_in6 *)to; tmp = *sin6; sin6 = &tmp; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (net == NULL) { memset(&ip6route, 0, sizeof(ip6route)); ro = (sctp_route_t *) & ip6route; memcpy(&ro->ro_dst, sin6, sin6->sin6_len); } else { ro = (sctp_route_t *) & net->ro; } if (stcb != NULL) { if ((stcb->asoc.ecn_allowed) && ecn_ok) { /* Enable ECN */ tosBottom = (((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) | sctp_get_ect(stcb, chk)) << 4); } else { /* No ECN */ tosBottom = ((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) << 4); } } else { /* we could get no asoc if it is a O-O-T-B packet */ tosBottom = ((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) << 4); } ip6h->ip6_flow = htonl(((tosTop << 24) | ((tosBottom | flowTop) << 16) | flowBottom)); if (port) { ip6h->ip6_nxt = IPPROTO_UDP; } else { ip6h->ip6_nxt = IPPROTO_SCTP; } ip6h->ip6_plen = (packet_length - sizeof(struct ip6_hdr)); ip6h->ip6_dst = sin6->sin6_addr; /* * Add SRC address selection here: we can only reuse to a * limited degree the kame src-addr-sel, since we can try * their selection but it may not be bound. */ bzero(&lsa6_tmp, sizeof(lsa6_tmp)); lsa6_tmp.sin6_family = AF_INET6; lsa6_tmp.sin6_len = sizeof(lsa6_tmp); lsa6 = &lsa6_tmp; if (net && out_of_asoc_ok == 0) { if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED | SCTP_ADDR_IFA_UNUSEABLE))) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (net->src_addr_selected == 0) { sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } /* Cache the source address */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, ro, net, 0, vrf_id); (void)sa6_recoverscope(sin6); net->src_addr_selected = 1; } if (net->ro._s_addr == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "V6:No route to host\n"); net->src_addr_selected = 0; goto no_route; } lsa6->sin6_addr = net->ro._s_addr->address.sin6.sin6_addr; } else { sin6 = (struct sockaddr_in6 *)&ro->ro_dst; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (over_addr == NULL) { struct sctp_ifa *_lsrc; _lsrc = sctp_source_address_selection(inp, stcb, ro, net, out_of_asoc_ok, vrf_id); if (_lsrc == NULL) { goto no_route; } lsa6->sin6_addr = _lsrc->address.sin6.sin6_addr; sctp_free_ifa(_lsrc); } else { lsa6->sin6_addr = over_addr->sin6.sin6_addr; SCTP_RTALLOC(ro, vrf_id); } (void)sa6_recoverscope(sin6); } lsa6->sin6_port = inp->sctp_lport; if (ro->ro_rt == NULL) { /* * src addr selection failed to find a route (or * valid source addr), so we can't get there from * here! */ goto no_route; } /* * XXX: sa6 may not have a valid sin6_scope_id in the * non-SCOPEDROUTING case. */ bzero(&lsa6_storage, sizeof(lsa6_storage)); lsa6_storage.sin6_family = AF_INET6; lsa6_storage.sin6_len = sizeof(lsa6_storage); lsa6_storage.sin6_addr = lsa6->sin6_addr; if ((error = sa6_recoverscope(&lsa6_storage)) != 0) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "recover scope fails error %d\n", error); sctp_m_freem(m); return (error); } /* XXX */ lsa6_storage.sin6_addr = lsa6->sin6_addr; lsa6_storage.sin6_port = inp->sctp_lport; lsa6 = &lsa6_storage; ip6h->ip6_src = lsa6->sin6_addr; if (port) { udp = (struct udphdr *)((caddr_t)ip6h + sizeof(struct ip6_hdr)); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(packet_length - sizeof(struct ip6_hdr)); udp->uh_sum = 0; sctphdr = (struct sctphdr *)((caddr_t)udp + sizeof(struct udphdr)); } else { sctphdr = (struct sctphdr *)((caddr_t)ip6h + sizeof(struct ip6_hdr)); } sctphdr->src_port = src_port; sctphdr->dest_port = dest_port; sctphdr->v_tag = v_tag; sctphdr->checksum = 0; /* * We set the hop limit now since there is a good chance * that our ro pointer is now filled */ ip6h->ip6_hlim = SCTP_GET_HLIM(inp, ro); ifp = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); #ifdef SCTP_DEBUG /* Copy to be sure something bad is not happening */ sin6->sin6_addr = ip6h->ip6_dst; lsa6->sin6_addr = ip6h->ip6_src; #endif SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv6 output routine from low level\n"); SCTPDBG(SCTP_DEBUG_OUTPUT3, "src: "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)lsa6); SCTPDBG(SCTP_DEBUG_OUTPUT3, "dst: "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)sin6); if (net) { sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; /* preserve the port and scope for link local send */ prev_scope = sin6->sin6_scope_id; prev_port = sin6->sin6_port; } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* failed to prepend data, give up */ sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(m, packet_length); #endif SCTP_ATTACH_CHAIN(o_pak, m, packet_length); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else if (!(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && (stcb) && (stcb->asoc.loopback_scope))) { sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); SCTP_STAT_INCR(sctps_sendswcrc); } else { SCTP_STAT_INCR(sctps_sendnocrc); } #endif if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), packet_length - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else m->m_pkthdr.csum_flags = CSUM_SCTP; m->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } /* send it out. table id is taken from stcb */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { so = SCTP_INP_SO(inp); SCTP_SOCKET_UNLOCK(so, 0); } #endif SCTP_IP6_OUTPUT(ret, o_pak, (struct route_in6 *)ro, &ifp, stcb, vrf_id); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 0); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif if (net) { /* for link local this must be done */ sin6->sin6_scope_id = prev_scope; sin6->sin6_port = prev_port; } SCTPDBG(SCTP_DEBUG_OUTPUT3, "return from send is %d\n", ret); SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); if (ret) { SCTP_STAT_INCR(sctps_senderrors); } if (net == NULL) { /* Now if we had a temp route free it */ if (ro->ro_rt) { RTFREE(ro->ro_rt); } } else { /* PMTU check versus smallest asoc MTU goes here */ if (ro->ro_rt == NULL) { /* Route was freed */ if (net->ro._s_addr && net->src_addr_selected) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } if ((ro->ro_rt != NULL) && (net->ro._s_addr)) { uint32_t mtu; mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); if (mtu && (stcb->asoc.smallest_mtu > mtu)) { sctp_mtu_size_reset(inp, &stcb->asoc, mtu); net->mtu = mtu; if (net->port) { net->mtu -= sizeof(struct udphdr); } } } else if (ifp) { if (ND_IFINFO(ifp)->linkmtu && (stcb->asoc.smallest_mtu > ND_IFINFO(ifp)->linkmtu)) { sctp_mtu_size_reset(inp, &stcb->asoc, ND_IFINFO(ifp)->linkmtu); } } } return (ret); } #endif else { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Unknown protocol (TSNH) type %d\n", ((struct sockaddr *)to)->sa_family); sctp_m_freem(m); SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); return (EFAULT); } } void sctp_send_initiate(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m, *m_at, *mp_last; struct sctp_nets *net; struct sctp_init_chunk *init; struct sctp_supported_addr_param *sup_addr; struct sctp_adaptation_layer_indication *ali; struct sctp_ecn_supported_param *ecn; struct sctp_prsctp_supported_param *prsctp; - struct sctp_ecn_nonce_supported_param *ecn_nonce; struct sctp_supported_chunk_types_param *pr_supported; int cnt_inits_to = 0; int padval, ret; int num_ext; int p_len; /* INIT's always go to the primary (and usually ONLY address) */ mp_last = NULL; net = stcb->asoc.primary_destination; if (net == NULL) { net = TAILQ_FIRST(&stcb->asoc.nets); if (net == NULL) { /* TSNH */ return; } /* we confirm any address we send an INIT to */ net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; (void)sctp_set_primary_addr(stcb, NULL, net); } else { /* we confirm any address we send an INIT to */ net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; } SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT\n"); #ifdef INET6 if (((struct sockaddr *)&(net->ro._l_addr))->sa_family == AF_INET6) { /* * special hook, if we are sending to link local it will not * show up in our private address count. */ struct sockaddr_in6 *sin6l; sin6l = &net->ro._l_addr.sin6; if (IN6_IS_ADDR_LINKLOCAL(&sin6l->sin6_addr)) cnt_inits_to = 1; } #endif if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* This case should not happen */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - failed timer?\n"); return; } /* start the INIT timer */ sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, net); m = sctp_get_mbuf_for_msg(MCLBYTES, 1, M_DONTWAIT, 1, MT_DATA); if (m == NULL) { /* No memory, INIT timer will re-attempt. */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - mbuf?\n"); return; } SCTP_BUF_LEN(m) = sizeof(struct sctp_init_chunk); /* * assume peer supports asconf in order to be able to queue local * address changes while an INIT is in flight and before the assoc * is established. */ stcb->asoc.peer_supports_asconf = 1; /* Now lets put the SCTP header in place */ init = mtod(m, struct sctp_init_chunk *); /* now the chunk header */ init->ch.chunk_type = SCTP_INITIATION; init->ch.chunk_flags = 0; /* fill in later from mbuf we build */ init->ch.chunk_length = 0; /* place in my tag */ init->init.initiate_tag = htonl(stcb->asoc.my_vtag); /* set up some of the credits. */ init->init.a_rwnd = htonl(max(inp->sctp_socket ? SCTP_SB_LIMIT_RCV(inp->sctp_socket) : 0, SCTP_MINIMAL_RWND)); init->init.num_outbound_streams = htons(stcb->asoc.pre_open_streams); init->init.num_inbound_streams = htons(stcb->asoc.max_inbound_streams); init->init.initial_tsn = htonl(stcb->asoc.init_seq_number); /* now the address restriction */ sup_addr = (struct sctp_supported_addr_param *)((caddr_t)init + sizeof(*init)); sup_addr->ph.param_type = htons(SCTP_SUPPORTED_ADDRTYPE); #ifdef INET6 /* we support 2 types: IPv6/IPv4 */ sup_addr->ph.param_length = htons(sizeof(*sup_addr) + sizeof(uint16_t)); sup_addr->addr_type[0] = htons(SCTP_IPV4_ADDRESS); sup_addr->addr_type[1] = htons(SCTP_IPV6_ADDRESS); #else /* we support 1 type: IPv4 */ sup_addr->ph.param_length = htons(sizeof(*sup_addr) + sizeof(uint8_t)); sup_addr->addr_type[0] = htons(SCTP_IPV4_ADDRESS); sup_addr->addr_type[1] = htons(0); /* this is the padding */ #endif SCTP_BUF_LEN(m) += sizeof(*sup_addr) + sizeof(uint16_t); /* adaptation layer indication parameter */ ali = (struct sctp_adaptation_layer_indication *)((caddr_t)sup_addr + sizeof(*sup_addr) + sizeof(uint16_t)); ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); ali->ph.param_length = htons(sizeof(*ali)); ali->indication = ntohl(inp->sctp_ep.adaptation_layer_indicator); SCTP_BUF_LEN(m) += sizeof(*ali); ecn = (struct sctp_ecn_supported_param *)((caddr_t)ali + sizeof(*ali)); if (SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly)) { /* Add NAT friendly parameter */ struct sctp_paramhdr *ph; ph = (struct sctp_paramhdr *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); ph->param_type = htons(SCTP_HAS_NAT_SUPPORT); ph->param_length = htons(sizeof(struct sctp_paramhdr)); SCTP_BUF_LEN(m) += sizeof(struct sctp_paramhdr); ecn = (struct sctp_ecn_supported_param *)((caddr_t)ph + sizeof(*ph)); } /* now any cookie time extensions */ if (stcb->asoc.cookie_preserve_req) { struct sctp_cookie_perserve_param *cookie_preserve; cookie_preserve = (struct sctp_cookie_perserve_param *)(ecn); cookie_preserve->ph.param_type = htons(SCTP_COOKIE_PRESERVE); cookie_preserve->ph.param_length = htons( sizeof(*cookie_preserve)); cookie_preserve->time = htonl(stcb->asoc.cookie_preserve_req); SCTP_BUF_LEN(m) += sizeof(*cookie_preserve); ecn = (struct sctp_ecn_supported_param *)( (caddr_t)cookie_preserve + sizeof(*cookie_preserve)); stcb->asoc.cookie_preserve_req = 0; } /* ECN parameter */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 1) { ecn->ph.param_type = htons(SCTP_ECN_CAPABLE); ecn->ph.param_length = htons(sizeof(*ecn)); SCTP_BUF_LEN(m) += sizeof(*ecn); prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn + sizeof(*ecn)); } else { prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn); } /* And now tell the peer we do pr-sctp */ prsctp->ph.param_type = htons(SCTP_PRSCTP_SUPPORTED); prsctp->ph.param_length = htons(sizeof(*prsctp)); SCTP_BUF_LEN(m) += sizeof(*prsctp); /* And now tell the peer we do all the extensions */ pr_supported = (struct sctp_supported_chunk_types_param *) ((caddr_t)prsctp + sizeof(*prsctp)); pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); num_ext = 0; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) { pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; } if (stcb->asoc.sctp_nr_sack_on_off == 1) { pr_supported->chunk_types[num_ext++] = SCTP_NR_SELECTIVE_ACK; } p_len = sizeof(*pr_supported) + num_ext; pr_supported->ph.param_length = htons(p_len); bzero((caddr_t)pr_supported + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); - /* ECN nonce: And now tell the peer we support ECN nonce */ - if (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) { - ecn_nonce = (struct sctp_ecn_nonce_supported_param *) - ((caddr_t)pr_supported + SCTP_SIZE32(p_len)); - ecn_nonce->ph.param_type = htons(SCTP_ECN_NONCE_SUPPORTED); - ecn_nonce->ph.param_length = htons(sizeof(*ecn_nonce)); - SCTP_BUF_LEN(m) += sizeof(*ecn_nonce); - } /* add authentication parameters */ if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) { struct sctp_auth_random *randp; struct sctp_auth_hmac_algo *hmacs; struct sctp_auth_chunk_list *chunks; /* attach RANDOM parameter, if available */ if (stcb->asoc.authinfo.random != NULL) { randp = (struct sctp_auth_random *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sizeof(*randp) + stcb->asoc.authinfo.random_len; /* random key already contains the header */ bcopy(stcb->asoc.authinfo.random->key, randp, p_len); /* zero out any padding required */ bzero((caddr_t)randp + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add HMAC_ALGO parameter */ hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_hmaclist(stcb->asoc.local_hmacs, (uint8_t *) hmacs->hmac_ids); if (p_len > 0) { p_len += sizeof(*hmacs); hmacs->ph.param_type = htons(SCTP_HMAC_LIST); hmacs->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)hmacs + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add CHUNKS parameter */ chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_auth_chunks(stcb->asoc.local_auth_chunks, chunks->chunk_types); if (p_len > 0) { p_len += sizeof(*chunks); chunks->ph.param_type = htons(SCTP_CHUNK_LIST); chunks->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)chunks + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } } m_at = m; /* now the addresses */ { struct sctp_scoping scp; /* * To optimize this we could put the scoping stuff into a * structure and remove the individual uint8's from the * assoc structure. Then we could just sifa in the address * within the stcb.. but for now this is a quick hack to get * the address stuff teased apart. */ scp.ipv4_addr_legal = stcb->asoc.ipv4_addr_legal; scp.ipv6_addr_legal = stcb->asoc.ipv6_addr_legal; scp.loopback_scope = stcb->asoc.loopback_scope; scp.ipv4_local_scope = stcb->asoc.ipv4_local_scope; scp.local_scope = stcb->asoc.local_scope; scp.site_scope = stcb->asoc.site_scope; m_at = sctp_add_addresses_to_i_ia(inp, &scp, m_at, cnt_inits_to); } /* calulate the size and update pkt header and chunk header */ p_len = 0; for (m_at = m; m_at; m_at = SCTP_BUF_NEXT(m_at)) { if (SCTP_BUF_NEXT(m_at) == NULL) mp_last = m_at; p_len += SCTP_BUF_LEN(m_at); } init->ch.chunk_length = htons(p_len); /* * We sifa 0 here to NOT set IP_DF if its IPv4, we ignore the return * here since the timer will drive a retranmission. */ /* I don't expect this to execute but we will be safe here */ padval = p_len % 4; if ((padval) && (mp_last)) { /* * The compiler worries that mp_last may not be set even * though I think it is impossible :-> however we add * mp_last here just in case. */ ret = sctp_add_pad_tombuf(mp_last, (4 - padval)); if (ret) { /* Houston we have a problem, no space */ sctp_m_freem(m); return; } p_len += padval; } SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - calls lowlevel_output\n"); ret = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, m, 0, NULL, 0, 0, 0, NULL, 0, inp->sctp_lport, stcb->rport, htonl(0), net->port, so_locked, NULL); SCTPDBG(SCTP_DEBUG_OUTPUT4, "lowlevel_output - %d\n", ret); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); } struct mbuf * sctp_arethere_unrecognized_parameters(struct mbuf *in_initpkt, int param_offset, int *abort_processing, struct sctp_chunkhdr *cp, int *nat_friendly) { /* * Given a mbuf containing an INIT or INIT-ACK with the param_offset * being equal to the beginning of the params i.e. (iphlen + * sizeof(struct sctp_init_msg) parse through the parameters to the * end of the mbuf verifying that all parameters are known. * * For unknown parameters build and return a mbuf with * UNRECOGNIZED_PARAMETER errors. If the flags indicate to stop * processing this chunk stop, and set *abort_processing to 1. * * By having param_offset be pre-set to where parameters begin it is * hoped that this routine may be reused in the future by new * features. */ struct sctp_paramhdr *phdr, params; struct mbuf *mat, *op_err; char tempbuf[SCTP_PARAM_BUFFER_SIZE]; int at, limit, pad_needed; uint16_t ptype, plen, padded_size; int err_at; *abort_processing = 0; mat = in_initpkt; err_at = 0; limit = ntohs(cp->chunk_length) - sizeof(struct sctp_init_chunk); at = param_offset; op_err = NULL; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Check for unrecognized param's\n"); phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); while ((phdr != NULL) && ((size_t)limit >= sizeof(struct sctp_paramhdr))) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if ((plen > limit) || (plen < sizeof(struct sctp_paramhdr))) { /* wacked parameter */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error %d\n", plen); goto invalid_size; } limit -= SCTP_SIZE32(plen); /*- * All parameters for all chunks that we know/understand are * listed here. We process them other places and make * appropriate stop actions per the upper bits. However this * is the generic routine processor's can call to get back * an operr.. to either incorporate (init-ack) or send. */ padded_size = SCTP_SIZE32(plen); switch (ptype) { /* Param's with variable size */ case SCTP_HEARTBEAT_INFO: case SCTP_STATE_COOKIE: case SCTP_UNRECOG_PARAM: case SCTP_ERROR_CAUSE_IND: /* ok skip fwd */ at += padded_size; break; /* Param's with variable size within a range */ case SCTP_CHUNK_LIST: case SCTP_SUPPORTED_CHUNK_EXT: if (padded_size > (sizeof(struct sctp_supported_chunk_types_param) + (sizeof(uint8_t) * SCTP_MAX_SUPPORTED_EXT))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error chklist %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SUPPORTED_ADDRTYPE: if (padded_size > SCTP_MAX_ADDR_PARAMS_SIZE) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error supaddrtype %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_RANDOM: if (padded_size > (sizeof(struct sctp_auth_random) + SCTP_RANDOM_MAX_SIZE)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error random %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SET_PRIM_ADDR: case SCTP_DEL_IP_ADDRESS: case SCTP_ADD_IP_ADDRESS: if ((padded_size != sizeof(struct sctp_asconf_addrv4_param)) && (padded_size != sizeof(struct sctp_asconf_addr_param))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error setprim %d\n", plen); goto invalid_size; } at += padded_size; break; /* Param's with a fixed size */ case SCTP_IPV4_ADDRESS: if (padded_size != sizeof(struct sctp_ipv4addr_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv4 addr %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_IPV6_ADDRESS: if (padded_size != sizeof(struct sctp_ipv6addr_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv6 addr %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_COOKIE_PRESERVE: if (padded_size != sizeof(struct sctp_cookie_perserve_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error cookie-preserve %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_HAS_NAT_SUPPORT: *nat_friendly = 1; /* fall through */ - case SCTP_ECN_NONCE_SUPPORTED: case SCTP_PRSCTP_SUPPORTED: if (padded_size != sizeof(struct sctp_paramhdr)) { - SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ecnnonce/prsctp/nat support %d\n", plen); + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error prsctp/nat support %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_ECN_CAPABLE: if (padded_size != sizeof(struct sctp_ecn_supported_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ecn %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_ULP_ADAPTATION: if (padded_size != sizeof(struct sctp_adaptation_layer_indication)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error adapatation %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SUCCESS_REPORT: if (padded_size != sizeof(struct sctp_asconf_paramhdr)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error success %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_HOSTNAME_ADDRESS: { /* We can NOT handle HOST NAME addresses!! */ int l_len; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Can't handle hostname addresses.. abort processing\n"); *abort_processing = 1; if (op_err == NULL) { /* Ok need to try to get a mbuf */ #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += plen; l_len += sizeof(struct sctp_paramhdr); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; /* * pre-reserve space for ip * and sctp header and * chunk hdr */ #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if (op_err) { /* If we have space */ struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_CAUSE_UNRESOLVABLE_ADDR); s.param_length = htons(sizeof(s) + plen); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); phdr = sctp_get_next_param(mat, at, (struct sctp_paramhdr *)tempbuf, min(sizeof(tempbuf), plen)); if (phdr == NULL) { sctp_m_freem(op_err); /* * we are out of memory but * we still need to have a * look at what to do (the * system is in trouble * though). */ return (NULL); } m_copyback(op_err, err_at, plen, (caddr_t)phdr); err_at += plen; } return (op_err); break; } default: /* * we do not recognize the parameter figure out what * we do. */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Hit default param %x\n", ptype); if ((ptype & 0x4000) == 0x4000) { /* Report bit is set?? */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "report op err\n"); if (op_err == NULL) { int l_len; /* Ok need to try to get an mbuf */ #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += plen; l_len += sizeof(struct sctp_paramhdr); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if (op_err) { /* If we have space */ struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_UNRECOG_PARAM); s.param_length = htons(sizeof(s) + plen); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); if (plen > sizeof(tempbuf)) { plen = sizeof(tempbuf); } phdr = sctp_get_next_param(mat, at, (struct sctp_paramhdr *)tempbuf, min(sizeof(tempbuf), plen)); if (phdr == NULL) { sctp_m_freem(op_err); /* * we are out of memory but * we still need to have a * look at what to do (the * system is in trouble * though). */ op_err = NULL; goto more_processing; } m_copyback(op_err, err_at, plen, (caddr_t)phdr); err_at += plen; } } more_processing: if ((ptype & 0x8000) == 0x0000) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "stop proc\n"); return (op_err); } else { /* skip this chunk and continue processing */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "move on\n"); at += SCTP_SIZE32(plen); } break; } phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); } return (op_err); invalid_size: SCTPDBG(SCTP_DEBUG_OUTPUT1, "abort flag set\n"); *abort_processing = 1; if ((op_err == NULL) && phdr) { int l_len; #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += (2 * sizeof(struct sctp_paramhdr)); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if ((op_err) && phdr) { struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); s.param_length = htons(sizeof(s) + sizeof(struct sctp_paramhdr)); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); /* Only copy back the p-hdr that caused the issue */ m_copyback(op_err, err_at, sizeof(struct sctp_paramhdr), (caddr_t)phdr); } return (op_err); } static int sctp_are_there_new_addresses(struct sctp_association *asoc, struct mbuf *in_initpkt, int iphlen, int offset) { /* * Given a INIT packet, look through the packet to verify that there * are NO new addresses. As we go through the parameters add reports * of any un-understood parameters that require an error. Also we * must return (1) to drop the packet if we see a un-understood * parameter that tells us to drop the chunk. */ struct sockaddr_in sin4, *sa4; #ifdef INET6 struct sockaddr_in6 sin6, *sa6; #endif struct sockaddr *sa_touse; struct sockaddr *sa; struct sctp_paramhdr *phdr, params; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6h; #endif struct mbuf *mat; uint16_t ptype, plen; int err_at; uint8_t fnd; struct sctp_nets *net; memset(&sin4, 0, sizeof(sin4)); #ifdef INET6 memset(&sin6, 0, sizeof(sin6)); #endif sin4.sin_family = AF_INET; sin4.sin_len = sizeof(sin4); #ifdef INET6 sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(sin6); #endif sa_touse = NULL; /* First what about the src address of the pkt ? */ iph = mtod(in_initpkt, struct ip *); switch (iph->ip_v) { case IPVERSION: /* source addr is IPv4 */ sin4.sin_addr = iph->ip_src; sa_touse = (struct sockaddr *)&sin4; break; #ifdef INET6 case IPV6_VERSION >> 4: /* source addr is IPv6 */ ip6h = mtod(in_initpkt, struct ip6_hdr *); sin6.sin6_addr = ip6h->ip6_src; sa_touse = (struct sockaddr *)&sin6; break; #endif default: return (1); } fnd = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sa = (struct sockaddr *)&net->ro._l_addr; if (sa->sa_family == sa_touse->sa_family) { if (sa->sa_family == AF_INET) { sa4 = (struct sockaddr_in *)sa; if (sa4->sin_addr.s_addr == sin4.sin_addr.s_addr) { fnd = 1; break; } } #ifdef INET6 if (sa->sa_family == AF_INET6) { sa6 = (struct sockaddr_in6 *)sa; if (SCTP6_ARE_ADDR_EQUAL(sa6, &sin6)) { fnd = 1; break; } } #endif } } if (fnd == 0) { /* New address added! no need to look futher. */ return (1); } /* Ok so far lets munge through the rest of the packet */ mat = in_initpkt; err_at = 0; sa_touse = NULL; offset += sizeof(struct sctp_init_chunk); phdr = sctp_get_next_param(mat, offset, ¶ms, sizeof(params)); while (phdr) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (ptype == SCTP_IPV4_ADDRESS) { struct sctp_ipv4addr_param *p4, p4_buf; phdr = sctp_get_next_param(mat, offset, (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); if (plen != sizeof(struct sctp_ipv4addr_param) || phdr == NULL) { return (1); } p4 = (struct sctp_ipv4addr_param *)phdr; sin4.sin_addr.s_addr = p4->addr; sa_touse = (struct sockaddr *)&sin4; } else if (ptype == SCTP_IPV6_ADDRESS) { struct sctp_ipv6addr_param *p6, p6_buf; phdr = sctp_get_next_param(mat, offset, (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); if (plen != sizeof(struct sctp_ipv6addr_param) || phdr == NULL) { return (1); } p6 = (struct sctp_ipv6addr_param *)phdr; #ifdef INET6 memcpy((caddr_t)&sin6.sin6_addr, p6->addr, sizeof(p6->addr)); #endif sa_touse = (struct sockaddr *)&sin4; } if (sa_touse) { /* ok, sa_touse points to one to check */ fnd = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sa = (struct sockaddr *)&net->ro._l_addr; if (sa->sa_family != sa_touse->sa_family) { continue; } if (sa->sa_family == AF_INET) { sa4 = (struct sockaddr_in *)sa; if (sa4->sin_addr.s_addr == sin4.sin_addr.s_addr) { fnd = 1; break; } } #ifdef INET6 if (sa->sa_family == AF_INET6) { sa6 = (struct sockaddr_in6 *)sa; if (SCTP6_ARE_ADDR_EQUAL( sa6, &sin6)) { fnd = 1; break; } } #endif } if (!fnd) { /* New addr added! no need to look further */ return (1); } } offset += SCTP_SIZE32(plen); phdr = sctp_get_next_param(mat, offset, ¶ms, sizeof(params)); } return (0); } /* * Given a MBUF chain that was sent into us containing an INIT. Build a * INIT-ACK with COOKIE and send back. We assume that the in_initpkt has done * a pullup to include IPv6/4header, SCTP header and initial part of INIT * message (i.e. the struct sctp_init_msg). */ void sctp_send_initiate_ack(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct mbuf *init_pkt, int iphlen, int offset, struct sctphdr *sh, struct sctp_init_chunk *init_chk, uint32_t vrf_id, uint16_t port, int hold_inp_lock) { struct sctp_association *asoc; struct mbuf *m, *m_at, *m_tmp, *m_cookie, *op_err, *mp_last; struct sctp_init_ack_chunk *initack; struct sctp_adaptation_layer_indication *ali; struct sctp_ecn_supported_param *ecn; struct sctp_prsctp_supported_param *prsctp; - struct sctp_ecn_nonce_supported_param *ecn_nonce; struct sctp_supported_chunk_types_param *pr_supported; union sctp_sockstore store, store1, *over_addr; struct sockaddr_in *sin, *to_sin; #ifdef INET6 struct sockaddr_in6 *sin6, *to_sin6; #endif struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6; #endif struct sockaddr *to; struct sctp_state_cookie stc; struct sctp_nets *net = NULL; uint8_t *signature = NULL; int cnt_inits_to = 0; uint16_t his_limit, i_want; int abort_flag, padval; int num_ext; int p_len; int nat_friendly = 0; struct socket *so; if (stcb) asoc = &stcb->asoc; else asoc = NULL; mp_last = NULL; if ((asoc != NULL) && (SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) && (sctp_are_there_new_addresses(asoc, init_pkt, iphlen, offset))) { /* new addresses, out of here in non-cookie-wait states */ /* * Send a ABORT, we don't add the new address error clause * though we even set the T bit and copy in the 0 tag.. this * looks no different than if no listener was present. */ sctp_send_abort(init_pkt, iphlen, sh, 0, NULL, vrf_id, port); return; } abort_flag = 0; op_err = sctp_arethere_unrecognized_parameters(init_pkt, (offset + sizeof(struct sctp_init_chunk)), &abort_flag, (struct sctp_chunkhdr *)init_chk, &nat_friendly); if (abort_flag) { do_a_abort: sctp_send_abort(init_pkt, iphlen, sh, init_chk->init.initiate_tag, op_err, vrf_id, port); return; } m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (m == NULL) { /* No memory, INIT timer will re-attempt. */ if (op_err) sctp_m_freem(op_err); return; } SCTP_BUF_LEN(m) = sizeof(struct sctp_init_chunk); /* the time I built cookie */ (void)SCTP_GETTIME_TIMEVAL(&stc.time_entered); /* populate any tie tags */ if (asoc != NULL) { /* unlock before tag selections */ stc.tie_tag_my_vtag = asoc->my_vtag_nonce; stc.tie_tag_peer_vtag = asoc->peer_vtag_nonce; stc.cookie_life = asoc->cookie_life; net = asoc->primary_destination; } else { stc.tie_tag_my_vtag = 0; stc.tie_tag_peer_vtag = 0; /* life I will award this cookie */ stc.cookie_life = inp->sctp_ep.def_cookie_life; } /* copy in the ports for later check */ stc.myport = sh->dest_port; stc.peerport = sh->src_port; /* * If we wanted to honor cookie life extentions, we would add to * stc.cookie_life. For now we should NOT honor any extension */ stc.site_scope = stc.local_scope = stc.loopback_scope = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { struct inpcb *in_inp; /* Its a V6 socket */ in_inp = (struct inpcb *)inp; stc.ipv6_addr_legal = 1; /* Now look at the binding flag to see if V4 will be legal */ if (SCTP_IPV6_V6ONLY(in_inp) == 0) { stc.ipv4_addr_legal = 1; } else { /* V4 addresses are NOT legal on the association */ stc.ipv4_addr_legal = 0; } } else { /* Its a V4 socket, no - V6 */ stc.ipv4_addr_legal = 1; stc.ipv6_addr_legal = 0; } #ifdef SCTP_DONT_DO_PRIVADDR_SCOPE stc.ipv4_scope = 1; #else stc.ipv4_scope = 0; #endif /* now for scope setup */ memset((caddr_t)&store, 0, sizeof(store)); memset((caddr_t)&store1, 0, sizeof(store1)); sin = &store.sin; to_sin = &store1.sin; #ifdef INET6 sin6 = &store.sin6; to_sin6 = &store1.sin6; #endif iph = mtod(init_pkt, struct ip *); /* establish the to_addr's */ switch (iph->ip_v) { case IPVERSION: to_sin->sin_port = sh->dest_port; to_sin->sin_family = AF_INET; to_sin->sin_len = sizeof(struct sockaddr_in); to_sin->sin_addr = iph->ip_dst; break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = mtod(init_pkt, struct ip6_hdr *); to_sin6->sin6_addr = ip6->ip6_dst; to_sin6->sin6_scope_id = 0; to_sin6->sin6_port = sh->dest_port; to_sin6->sin6_family = AF_INET6; to_sin6->sin6_len = sizeof(struct sockaddr_in6); break; #endif default: goto do_a_abort; break; }; if (net == NULL) { to = (struct sockaddr *)&store; switch (iph->ip_v) { case IPVERSION: { sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_port = sh->src_port; sin->sin_addr = iph->ip_src; /* lookup address */ stc.address[0] = sin->sin_addr.s_addr; stc.address[1] = 0; stc.address[2] = 0; stc.address[3] = 0; stc.addr_type = SCTP_IPV4_ADDRESS; /* local from address */ stc.laddress[0] = to_sin->sin_addr.s_addr; stc.laddress[1] = 0; stc.laddress[2] = 0; stc.laddress[3] = 0; stc.laddr_type = SCTP_IPV4_ADDRESS; /* scope_id is only for v6 */ stc.scope_id = 0; #ifndef SCTP_DONT_DO_PRIVADDR_SCOPE if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { stc.ipv4_scope = 1; } #else stc.ipv4_scope = 1; #endif /* SCTP_DONT_DO_PRIVADDR_SCOPE */ /* Must use the address in this case */ if (sctp_is_address_on_local_host((struct sockaddr *)sin, vrf_id)) { stc.loopback_scope = 1; stc.ipv4_scope = 1; stc.site_scope = 1; stc.local_scope = 0; } break; } #ifdef INET6 case IPV6_VERSION >> 4: { ip6 = mtod(init_pkt, struct ip6_hdr *); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); sin6->sin6_port = sh->src_port; sin6->sin6_addr = ip6->ip6_src; /* lookup address */ memcpy(&stc.address, &sin6->sin6_addr, sizeof(struct in6_addr)); sin6->sin6_scope_id = 0; stc.addr_type = SCTP_IPV6_ADDRESS; stc.scope_id = 0; if (sctp_is_address_on_local_host((struct sockaddr *)sin6, vrf_id)) { /* * FIX ME: does this have scope from * rcvif? */ (void)sa6_recoverscope(sin6); stc.scope_id = sin6->sin6_scope_id; sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); stc.loopback_scope = 1; stc.local_scope = 0; stc.site_scope = 1; stc.ipv4_scope = 1; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { /* * If the new destination is a * LINK_LOCAL we must have common * both site and local scope. Don't * set local scope though since we * must depend on the source to be * added implicitly. We cannot * assure just because we share one * link that all links are common. */ stc.local_scope = 0; stc.site_scope = 1; stc.ipv4_scope = 1; /* * we start counting for the private * address stuff at 1. since the * link local we source from won't * show up in our scoped count. */ cnt_inits_to = 1; /* * pull out the scope_id from * incoming pkt */ /* * FIX ME: does this have scope from * rcvif? */ (void)sa6_recoverscope(sin6); stc.scope_id = sin6->sin6_scope_id; sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { /* * If the new destination is * SITE_LOCAL then we must have site * scope in common. */ stc.site_scope = 1; } memcpy(&stc.laddress, &to_sin6->sin6_addr, sizeof(struct in6_addr)); stc.laddr_type = SCTP_IPV6_ADDRESS; break; } #endif default: /* TSNH */ goto do_a_abort; break; } } else { /* set the scope per the existing tcb */ #ifdef INET6 struct sctp_nets *lnet; #endif stc.loopback_scope = asoc->loopback_scope; stc.ipv4_scope = asoc->ipv4_local_scope; stc.site_scope = asoc->site_scope; stc.local_scope = asoc->local_scope; #ifdef INET6 /* Why do we not consider IPv4 LL addresses? */ TAILQ_FOREACH(lnet, &asoc->nets, sctp_next) { if (lnet->ro._l_addr.sin6.sin6_family == AF_INET6) { if (IN6_IS_ADDR_LINKLOCAL(&lnet->ro._l_addr.sin6.sin6_addr)) { /* * if we have a LL address, start * counting at 1. */ cnt_inits_to = 1; } } } #endif /* use the net pointer */ to = (struct sockaddr *)&net->ro._l_addr; switch (to->sa_family) { case AF_INET: sin = (struct sockaddr_in *)to; stc.address[0] = sin->sin_addr.s_addr; stc.address[1] = 0; stc.address[2] = 0; stc.address[3] = 0; stc.addr_type = SCTP_IPV4_ADDRESS; if (net->src_addr_selected == 0) { /* * strange case here, the INIT should have * did the selection. */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (net->ro._s_addr == NULL) return; net->src_addr_selected = 1; } stc.laddress[0] = net->ro._s_addr->address.sin.sin_addr.s_addr; stc.laddress[1] = 0; stc.laddress[2] = 0; stc.laddress[3] = 0; stc.laddr_type = SCTP_IPV4_ADDRESS; break; #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)to; memcpy(&stc.address, &sin6->sin6_addr, sizeof(struct in6_addr)); stc.addr_type = SCTP_IPV6_ADDRESS; if (net->src_addr_selected == 0) { /* * strange case here, the INIT should have * did the selection. */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (net->ro._s_addr == NULL) return; net->src_addr_selected = 1; } memcpy(&stc.laddress, &net->ro._s_addr->address.sin6.sin6_addr, sizeof(struct in6_addr)); stc.laddr_type = SCTP_IPV6_ADDRESS; break; #endif } } /* Now lets put the SCTP header in place */ initack = mtod(m, struct sctp_init_ack_chunk *); /* Save it off for quick ref */ stc.peers_vtag = init_chk->init.initiate_tag; /* who are we */ memcpy(stc.identification, SCTP_VERSION_STRING, min(strlen(SCTP_VERSION_STRING), sizeof(stc.identification))); /* now the chunk header */ initack->ch.chunk_type = SCTP_INITIATION_ACK; initack->ch.chunk_flags = 0; /* fill in later from mbuf we build */ initack->ch.chunk_length = 0; /* place in my tag */ if ((asoc != NULL) && ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_INUSE) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED))) { /* re-use the v-tags and init-seq here */ initack->init.initiate_tag = htonl(asoc->my_vtag); initack->init.initial_tsn = htonl(asoc->init_seq_number); } else { uint32_t vtag, itsn; if (hold_inp_lock) { SCTP_INP_INCR_REF(inp); SCTP_INP_RUNLOCK(inp); } if (asoc) { atomic_add_int(&asoc->refcnt, 1); SCTP_TCB_UNLOCK(stcb); new_tag: vtag = sctp_select_a_tag(inp, inp->sctp_lport, sh->src_port, 1); if ((asoc->peer_supports_nat) && (vtag == asoc->my_vtag)) { /* * Got a duplicate vtag on some guy behind a * nat make sure we don't use it. */ goto new_tag; } initack->init.initiate_tag = htonl(vtag); /* get a TSN to use too */ itsn = sctp_select_initial_TSN(&inp->sctp_ep); initack->init.initial_tsn = htonl(itsn); SCTP_TCB_LOCK(stcb); atomic_add_int(&asoc->refcnt, -1); } else { vtag = sctp_select_a_tag(inp, inp->sctp_lport, sh->src_port, 1); initack->init.initiate_tag = htonl(vtag); /* get a TSN to use too */ initack->init.initial_tsn = htonl(sctp_select_initial_TSN(&inp->sctp_ep)); } if (hold_inp_lock) { SCTP_INP_RLOCK(inp); SCTP_INP_DECR_REF(inp); } } /* save away my tag to */ stc.my_vtag = initack->init.initiate_tag; /* set up some of the credits. */ so = inp->sctp_socket; if (so == NULL) { /* memory problem */ sctp_m_freem(m); return; } else { initack->init.a_rwnd = htonl(max(SCTP_SB_LIMIT_RCV(so), SCTP_MINIMAL_RWND)); } /* set what I want */ his_limit = ntohs(init_chk->init.num_inbound_streams); /* choose what I want */ if (asoc != NULL) { if (asoc->streamoutcnt > inp->sctp_ep.pre_open_stream_count) { i_want = asoc->streamoutcnt; } else { i_want = inp->sctp_ep.pre_open_stream_count; } } else { i_want = inp->sctp_ep.pre_open_stream_count; } if (his_limit < i_want) { /* I Want more :< */ initack->init.num_outbound_streams = init_chk->init.num_inbound_streams; } else { /* I can have what I want :> */ initack->init.num_outbound_streams = htons(i_want); } /* tell him his limt. */ initack->init.num_inbound_streams = htons(inp->sctp_ep.max_open_streams_intome); /* adaptation layer indication parameter */ ali = (struct sctp_adaptation_layer_indication *)((caddr_t)initack + sizeof(*initack)); ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); ali->ph.param_length = htons(sizeof(*ali)); ali->indication = ntohl(inp->sctp_ep.adaptation_layer_indicator); SCTP_BUF_LEN(m) += sizeof(*ali); ecn = (struct sctp_ecn_supported_param *)((caddr_t)ali + sizeof(*ali)); /* ECN parameter */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 1) { ecn->ph.param_type = htons(SCTP_ECN_CAPABLE); ecn->ph.param_length = htons(sizeof(*ecn)); SCTP_BUF_LEN(m) += sizeof(*ecn); prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn + sizeof(*ecn)); } else { prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn); } /* And now tell the peer we do pr-sctp */ prsctp->ph.param_type = htons(SCTP_PRSCTP_SUPPORTED); prsctp->ph.param_length = htons(sizeof(*prsctp)); SCTP_BUF_LEN(m) += sizeof(*prsctp); if (nat_friendly) { /* Add NAT friendly parameter */ struct sctp_paramhdr *ph; ph = (struct sctp_paramhdr *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); ph->param_type = htons(SCTP_HAS_NAT_SUPPORT); ph->param_length = htons(sizeof(struct sctp_paramhdr)); SCTP_BUF_LEN(m) += sizeof(struct sctp_paramhdr); } /* And now tell the peer we do all the extensions */ pr_supported = (struct sctp_supported_chunk_types_param *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); num_ext = 0; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; if (SCTP_BASE_SYSCTL(sctp_nr_sack_on_off)) pr_supported->chunk_types[num_ext++] = SCTP_NR_SELECTIVE_ACK; p_len = sizeof(*pr_supported) + num_ext; pr_supported->ph.param_length = htons(p_len); bzero((caddr_t)pr_supported + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); - /* ECN nonce: And now tell the peer we support ECN nonce */ - if (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) { - ecn_nonce = (struct sctp_ecn_nonce_supported_param *) - ((caddr_t)pr_supported + SCTP_SIZE32(p_len)); - ecn_nonce->ph.param_type = htons(SCTP_ECN_NONCE_SUPPORTED); - ecn_nonce->ph.param_length = htons(sizeof(*ecn_nonce)); - SCTP_BUF_LEN(m) += sizeof(*ecn_nonce); - } /* add authentication parameters */ if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) { struct sctp_auth_random *randp; struct sctp_auth_hmac_algo *hmacs; struct sctp_auth_chunk_list *chunks; uint16_t random_len; /* generate and add RANDOM parameter */ random_len = SCTP_AUTH_RANDOM_SIZE_DEFAULT; randp = (struct sctp_auth_random *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); randp->ph.param_type = htons(SCTP_RANDOM); p_len = sizeof(*randp) + random_len; randp->ph.param_length = htons(p_len); SCTP_READ_RANDOM(randp->random_data, random_len); /* zero out any padding required */ bzero((caddr_t)randp + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); /* add HMAC_ALGO parameter */ hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_hmaclist(inp->sctp_ep.local_hmacs, (uint8_t *) hmacs->hmac_ids); if (p_len > 0) { p_len += sizeof(*hmacs); hmacs->ph.param_type = htons(SCTP_HMAC_LIST); hmacs->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)hmacs + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add CHUNKS parameter */ chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_auth_chunks(inp->sctp_ep.local_auth_chunks, chunks->chunk_types); if (p_len > 0) { p_len += sizeof(*chunks); chunks->ph.param_type = htons(SCTP_CHUNK_LIST); chunks->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)chunks + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } } m_at = m; /* now the addresses */ { struct sctp_scoping scp; /* * To optimize this we could put the scoping stuff into a * structure and remove the individual uint8's from the stc * structure. Then we could just sifa in the address within * the stc.. but for now this is a quick hack to get the * address stuff teased apart. */ scp.ipv4_addr_legal = stc.ipv4_addr_legal; scp.ipv6_addr_legal = stc.ipv6_addr_legal; scp.loopback_scope = stc.loopback_scope; scp.ipv4_local_scope = stc.ipv4_scope; scp.local_scope = stc.local_scope; scp.site_scope = stc.site_scope; m_at = sctp_add_addresses_to_i_ia(inp, &scp, m_at, cnt_inits_to); } /* tack on the operational error if present */ if (op_err) { struct mbuf *ol; int llen; llen = 0; ol = op_err; while (ol) { llen += SCTP_BUF_LEN(ol); ol = SCTP_BUF_NEXT(ol); } if (llen % 4) { /* must add a pad to the param */ uint32_t cpthis = 0; int padlen; padlen = 4 - (llen % 4); m_copyback(op_err, llen, padlen, (caddr_t)&cpthis); } while (SCTP_BUF_NEXT(m_at) != NULL) { m_at = SCTP_BUF_NEXT(m_at); } SCTP_BUF_NEXT(m_at) = op_err; while (SCTP_BUF_NEXT(m_at) != NULL) { m_at = SCTP_BUF_NEXT(m_at); } } /* pre-calulate the size and update pkt header and chunk header */ p_len = 0; for (m_tmp = m; m_tmp; m_tmp = SCTP_BUF_NEXT(m_tmp)) { p_len += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { /* m_tmp should now point to last one */ break; } } /* Now we must build a cookie */ m_cookie = sctp_add_cookie(inp, init_pkt, offset, m, 0, &stc, &signature); if (m_cookie == NULL) { /* memory problem */ sctp_m_freem(m); return; } /* Now append the cookie to the end and update the space/size */ SCTP_BUF_NEXT(m_tmp) = m_cookie; for (m_tmp = m_cookie; m_tmp; m_tmp = SCTP_BUF_NEXT(m_tmp)) { p_len += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { /* m_tmp should now point to last one */ mp_last = m_tmp; break; } } /* * Place in the size, but we don't include the last pad (if any) in * the INIT-ACK. */ initack->ch.chunk_length = htons(p_len); /* * Time to sign the cookie, we don't sign over the cookie signature * though thus we set trailer. */ (void)sctp_hmac_m(SCTP_HMAC, (uint8_t *) inp->sctp_ep.secret_key[(int)(inp->sctp_ep.current_secret_number)], SCTP_SECRET_SIZE, m_cookie, sizeof(struct sctp_paramhdr), (uint8_t *) signature, SCTP_SIGNATURE_SIZE); /* * We sifa 0 here to NOT set IP_DF if its IPv4, we ignore the return * here since the timer will drive a retranmission. */ padval = p_len % 4; if ((padval) && (mp_last)) { /* see my previous comments on mp_last */ int ret; ret = sctp_add_pad_tombuf(mp_last, (4 - padval)); if (ret) { /* Houston we have a problem, no space */ sctp_m_freem(m); return; } p_len += padval; } if (stc.loopback_scope) { over_addr = &store1; } else { over_addr = NULL; } (void)sctp_lowlevel_chunk_output(inp, NULL, NULL, to, m, 0, NULL, 0, 0, 0, NULL, 0, inp->sctp_lport, sh->src_port, init_chk->init.initiate_tag, port, SCTP_SO_NOT_LOCKED, over_addr); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } static void sctp_prune_prsctp(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_sndrcvinfo *srcv, int dataout) { int freed_spc = 0; struct sctp_tmit_chunk *chk, *nchk; SCTP_TCB_LOCK_ASSERT(stcb); if ((asoc->peer_supports_prsctp) && (asoc->sent_queue_cnt_removeable > 0)) { TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { /* * Look for chunks marked with the PR_SCTP flag AND * the buffer space flag. If the one being sent is * equal or greater priority then purge the old one * and free some space. */ if (PR_SCTP_BUF_ENABLED(chk->flags)) { /* * This one is PR-SCTP AND buffer space * limited type */ if (chk->rec.data.timetodrop.tv_sec >= (long)srcv->sinfo_timetolive) { /* * Lower numbers equates to higher * priority so if the one we are * looking at has a larger or equal * priority we want to drop the data * and NOT retransmit it. */ if (chk->data) { /* * We release the book_size * if the mbuf is here */ int ret_spc; int cause; if (chk->sent > SCTP_DATAGRAM_UNSENT) cause = SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT; else cause = SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_UNSENT; ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, cause, SCTP_SO_LOCKED); freed_spc += ret_spc; if (freed_spc >= dataout) { return; } } /* if chunk was present */ } /* if of sufficent priority */ } /* if chunk has enabled */ } /* tailqforeach */ TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { /* Here we must move to the sent queue and mark */ if (PR_SCTP_BUF_ENABLED(chk->flags)) { if (chk->rec.data.timetodrop.tv_sec >= (long)srcv->sinfo_timetolive) { if (chk->data) { /* * We release the book_size * if the mbuf is here */ int ret_spc; ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_UNSENT, SCTP_SO_LOCKED); freed_spc += ret_spc; if (freed_spc >= dataout) { return; } } /* end if chk->data */ } /* end if right class */ } /* end if chk pr-sctp */ } /* tailqforeachsafe (chk) */ } /* if enabled in asoc */ } int sctp_get_frag_point(struct sctp_tcb *stcb, struct sctp_association *asoc) { int siz, ovh; /* * For endpoints that have both v6 and v4 addresses we must reserve * room for the ipv6 header, for those that are only dealing with V4 * we use a larger frag point. */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (stcb->asoc.sctp_frag_point > asoc->smallest_mtu) siz = asoc->smallest_mtu - ovh; else siz = (stcb->asoc.sctp_frag_point - ovh); /* * if (siz > (MCLBYTES-sizeof(struct sctp_data_chunk))) { */ /* A data chunk MUST fit in a cluster */ /* siz = (MCLBYTES - sizeof(struct sctp_data_chunk)); */ /* } */ /* adjust for an AUTH chunk if DATA requires auth */ if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) siz -= sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); if (siz % 4) { /* make it an even word boundary please */ siz -= (siz % 4); } return (siz); } static void sctp_set_prsctp_policy(struct sctp_stream_queue_pending *sp) { sp->pr_sctp_on = 0; /* * We assume that the user wants PR_SCTP_TTL if the user provides a * positive lifetime but does not specify any PR_SCTP policy. This * is a BAD assumption and causes problems at least with the * U-Vancovers MPI folks. I will change this to be no policy means * NO PR-SCTP. */ if (PR_SCTP_ENABLED(sp->sinfo_flags)) { sp->act_flags |= PR_SCTP_POLICY(sp->sinfo_flags); sp->pr_sctp_on = 1; } else { return; } switch (PR_SCTP_POLICY(sp->sinfo_flags)) { case CHUNK_FLAGS_PR_SCTP_BUF: /* * Time to live is a priority stored in tv_sec when doing * the buffer drop thing. */ sp->ts.tv_sec = sp->timetolive; sp->ts.tv_usec = 0; break; case CHUNK_FLAGS_PR_SCTP_TTL: { struct timeval tv; (void)SCTP_GETTIME_TIMEVAL(&sp->ts); tv.tv_sec = sp->timetolive / 1000; tv.tv_usec = (sp->timetolive * 1000) % 1000000; /* * TODO sctp_constants.h needs alternative time * macros when _KERNEL is undefined. */ timevaladd(&sp->ts, &tv); } break; case CHUNK_FLAGS_PR_SCTP_RTX: /* * Time to live is a the number or retransmissions stored in * tv_sec. */ sp->ts.tv_sec = sp->timetolive; sp->ts.tv_usec = 0; break; default: SCTPDBG(SCTP_DEBUG_USRREQ1, "Unknown PR_SCTP policy %u.\n", PR_SCTP_POLICY(sp->sinfo_flags)); break; } } static int sctp_msg_append(struct sctp_tcb *stcb, struct sctp_nets *net, struct mbuf *m, struct sctp_sndrcvinfo *srcv, int hold_stcb_lock) { int error = 0, holds_lock; struct mbuf *at; struct sctp_stream_queue_pending *sp = NULL; struct sctp_stream_out *strm; /* * Given an mbuf chain, put it into the association send queue and * place it on the wheel */ holds_lock = hold_stcb_lock; if (srcv->sinfo_stream >= stcb->asoc.streamoutcnt) { /* Invalid stream number */ SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_now; } if ((stcb->asoc.stream_locked) && (stcb->asoc.stream_locked_on != srcv->sinfo_stream)) { SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_now; } strm = &stcb->asoc.strmout[srcv->sinfo_stream]; /* Now can we send this? */ if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (stcb->asoc.state & SCTP_STATE_SHUTDOWN_PENDING)) { /* got data while shutting down */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_now; } sctp_alloc_a_strmoq(stcb, sp); if (sp == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); error = ENOMEM; goto out_now; } sp->sinfo_flags = srcv->sinfo_flags; sp->timetolive = srcv->sinfo_timetolive; sp->ppid = srcv->sinfo_ppid; sp->context = srcv->sinfo_context; sp->strseq = 0; if (sp->sinfo_flags & SCTP_ADDR_OVER) { sp->net = net; atomic_add_int(&sp->net->ref_count, 1); } else { sp->net = NULL; } (void)SCTP_GETTIME_TIMEVAL(&sp->ts); sp->stream = srcv->sinfo_stream; sp->msg_is_complete = 1; sp->sender_all_done = 1; sp->some_taken = 0; sp->data = m; sp->tail_mbuf = NULL; sp->length = 0; at = m; sctp_set_prsctp_policy(sp); /* * We could in theory (for sendall) sifa the length in, but we would * still have to hunt through the chain since we need to setup the * tail_mbuf */ while (at) { if (SCTP_BUF_NEXT(at) == NULL) sp->tail_mbuf = at; sp->length += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } SCTP_TCB_SEND_LOCK(stcb); sctp_snd_sb_alloc(stcb, sp->length); atomic_add_int(&stcb->asoc.stream_queue_cnt, 1); TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); if ((srcv->sinfo_flags & SCTP_UNORDERED) == 0) { sp->strseq = strm->next_sequence_sent; strm->next_sequence_sent++; } stcb->asoc.ss_functions.sctp_ss_add_to_stream(stcb, &stcb->asoc, strm, sp, 1); m = NULL; SCTP_TCB_SEND_UNLOCK(stcb); out_now: if (m) { sctp_m_freem(m); } return (error); } static struct mbuf * sctp_copy_mbufchain(struct mbuf *clonechain, struct mbuf *outchain, struct mbuf **endofchain, int can_take_mbuf, int sizeofcpy, uint8_t copy_by_ref) { struct mbuf *m; struct mbuf *appendchain; caddr_t cp; int len; if (endofchain == NULL) { /* error */ error_out: if (outchain) sctp_m_freem(outchain); return (NULL); } if (can_take_mbuf) { appendchain = clonechain; } else { if (!copy_by_ref && (sizeofcpy <= (int)((((SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count) - 1) * MLEN) + MHLEN))) ) { /* Its not in a cluster */ if (*endofchain == NULL) { /* lets get a mbuf cluster */ if (outchain == NULL) { /* This is the general case */ new_mbuf: outchain = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_HEADER); if (outchain == NULL) { goto error_out; } SCTP_BUF_LEN(outchain) = 0; *endofchain = outchain; /* get the prepend space */ SCTP_BUF_RESV_UF(outchain, (SCTP_FIRST_MBUF_RESV + 4)); } else { /* * We really should not get a NULL * in endofchain */ /* find end */ m = outchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } /* sanity */ if (*endofchain == NULL) { /* * huh, TSNH XXX maybe we * should panic */ sctp_m_freem(outchain); goto new_mbuf; } } /* get the new end of length */ len = M_TRAILINGSPACE(*endofchain); } else { /* how much is left at the end? */ len = M_TRAILINGSPACE(*endofchain); } /* Find the end of the data, for appending */ cp = (mtod((*endofchain), caddr_t)+SCTP_BUF_LEN((*endofchain))); /* Now lets copy it out */ if (len >= sizeofcpy) { /* It all fits, copy it in */ m_copydata(clonechain, 0, sizeofcpy, cp); SCTP_BUF_LEN((*endofchain)) += sizeofcpy; } else { /* fill up the end of the chain */ if (len > 0) { m_copydata(clonechain, 0, len, cp); SCTP_BUF_LEN((*endofchain)) += len; /* now we need another one */ sizeofcpy -= len; } m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_HEADER); if (m == NULL) { /* We failed */ goto error_out; } SCTP_BUF_NEXT((*endofchain)) = m; *endofchain = m; cp = mtod((*endofchain), caddr_t); m_copydata(clonechain, len, sizeofcpy, cp); SCTP_BUF_LEN((*endofchain)) += sizeofcpy; } return (outchain); } else { /* copy the old fashion way */ appendchain = SCTP_M_COPYM(clonechain, 0, M_COPYALL, M_DONTWAIT); #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = appendchain; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif } } if (appendchain == NULL) { /* error */ if (outchain) sctp_m_freem(outchain); return (NULL); } if (outchain) { /* tack on to the end */ if (*endofchain != NULL) { SCTP_BUF_NEXT(((*endofchain))) = appendchain; } else { m = outchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { SCTP_BUF_NEXT(m) = appendchain; break; } m = SCTP_BUF_NEXT(m); } } /* * save off the end and update the end-chain postion */ m = appendchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } return (outchain); } else { /* save off the end and update the end-chain postion */ m = appendchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } return (appendchain); } } int sctp_med_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *num_out, int *reason_code, int control_only, int from_where, struct timeval *now, int *now_filled, int frag_point, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ); static void sctp_sendall_iterator(struct sctp_inpcb *inp, struct sctp_tcb *stcb, void *ptr, uint32_t val) { struct sctp_copy_all *ca; struct mbuf *m; int ret = 0; int added_control = 0; int un_sent, do_chunk_output = 1; struct sctp_association *asoc; ca = (struct sctp_copy_all *)ptr; if (ca->m == NULL) { return; } if (ca->inp != inp) { /* TSNH */ return; } if ((ca->m) && ca->sndlen) { m = SCTP_M_COPYM(ca->m, 0, M_COPYALL, M_DONTWAIT); if (m == NULL) { /* can't copy so we are done */ ca->cnt_failed++; return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = m; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif } else { m = NULL; } SCTP_TCB_LOCK_ASSERT(stcb); if (ca->sndrcv.sinfo_flags & SCTP_ABORT) { /* Abort this assoc with m as the user defined reason */ if (m) { struct sctp_paramhdr *ph; SCTP_BUF_PREPEND(m, sizeof(struct sctp_paramhdr), M_DONTWAIT); if (m) { ph = mtod(m, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(ca->sndlen); } /* * We add one here to keep the assoc from * dis-appearing on us. */ atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(inp, stcb, SCTP_RESPONSE_TO_USER_REQ, m, SCTP_SO_NOT_LOCKED); /* * sctp_abort_an_association calls sctp_free_asoc() * free association will NOT free it since we * incremented the refcnt .. we do this to prevent * it being freed and things getting tricky since we * could end up (from free_asoc) calling inpcb_free * which would get a recursive lock call to the * iterator lock.. But as a consequence of that the * stcb will return to us un-locked.. since * free_asoc returns with either no TCB or the TCB * unlocked, we must relock.. to unlock in the * iterator timer :-0 */ SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); goto no_chunk_output; } } else { if (m) { ret = sctp_msg_append(stcb, stcb->asoc.primary_destination, m, &ca->sndrcv, 1); } asoc = &stcb->asoc; if (ca->sndrcv.sinfo_flags & SCTP_EOF) { /* shutdown this assoc */ int cnt; cnt = sctp_is_there_unsent_data(stcb); if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* * there is nothing queued to send, so I'm * done... */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* * only send SHUTDOWN the first time * through */ sctp_send_shutdown(stcb, stcb->asoc.primary_destination); if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); added_control = 1; do_chunk_output = 0; } } else { /* * we still got (or just got) data to send, * so set SHUTDOWN_PENDING */ /* * XXX sockets draft says that SCTP_EOF * should be sent with no data. currently, * we will allow user data to be sent first * and move to SHUTDOWN-PENDING */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { if (asoc->locked_on_sending) { /* * Locked to send out the * data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp) { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { abort_anyway: atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, NULL, SCTP_SO_NOT_LOCKED); atomic_add_int(&stcb->asoc.refcnt, -1); goto no_chunk_output; } sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } } } un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) ) { do_chunk_output = 0; } if (do_chunk_output) sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_NOT_LOCKED); else if (added_control) { int num_out = 0, reason = 0, now_filled = 0; struct timeval now; int frag_point; frag_point = sctp_get_frag_point(stcb, &stcb->asoc); (void)sctp_med_chunk_output(inp, stcb, &stcb->asoc, &num_out, &reason, 1, 1, &now, &now_filled, frag_point, SCTP_SO_NOT_LOCKED); } no_chunk_output: if (ret) { ca->cnt_failed++; } else { ca->cnt_sent++; } } static void sctp_sendall_completes(void *ptr, uint32_t val) { struct sctp_copy_all *ca; ca = (struct sctp_copy_all *)ptr; /* * Do a notify here? Kacheong suggests that the notify be done at * the send time.. so you would push up a notification if any send * failed. Don't know if this is feasable since the only failures we * have is "memory" related and if you cannot get an mbuf to send * the data you surely can't get an mbuf to send up to notify the * user you can't send the data :-> */ /* now free everything */ sctp_m_freem(ca->m); SCTP_FREE(ca, SCTP_M_COPYAL); } #define MC_ALIGN(m, len) do { \ SCTP_BUF_RESV_UF(m, ((MCLBYTES - (len)) & ~(sizeof(long) - 1)); \ } while (0) static struct mbuf * sctp_copy_out_all(struct uio *uio, int len) { struct mbuf *ret, *at; int left, willcpy, cancpy, error; ret = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_WAIT, 1, MT_DATA); if (ret == NULL) { /* TSNH */ return (NULL); } left = len; SCTP_BUF_LEN(ret) = 0; /* save space for the data chunk header */ cancpy = M_TRAILINGSPACE(ret); willcpy = min(cancpy, left); at = ret; while (left > 0) { /* Align data to the end */ error = uiomove(mtod(at, caddr_t), willcpy, uio); if (error) { err_out_now: sctp_m_freem(at); return (NULL); } SCTP_BUF_LEN(at) = willcpy; SCTP_BUF_NEXT_PKT(at) = SCTP_BUF_NEXT(at) = 0; left -= willcpy; if (left > 0) { SCTP_BUF_NEXT(at) = sctp_get_mbuf_for_msg(left, 0, M_WAIT, 1, MT_DATA); if (SCTP_BUF_NEXT(at) == NULL) { goto err_out_now; } at = SCTP_BUF_NEXT(at); SCTP_BUF_LEN(at) = 0; cancpy = M_TRAILINGSPACE(at); willcpy = min(cancpy, left); } } return (ret); } static int sctp_sendall(struct sctp_inpcb *inp, struct uio *uio, struct mbuf *m, struct sctp_sndrcvinfo *srcv) { int ret; struct sctp_copy_all *ca; SCTP_MALLOC(ca, struct sctp_copy_all *, sizeof(struct sctp_copy_all), SCTP_M_COPYAL); if (ca == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } memset(ca, 0, sizeof(struct sctp_copy_all)); ca->inp = inp; memcpy(&ca->sndrcv, srcv, sizeof(struct sctp_nonpad_sndrcvinfo)); /* * take off the sendall flag, it would be bad if we failed to do * this :-0 */ ca->sndrcv.sinfo_flags &= ~SCTP_SENDALL; /* get length and mbuf chain */ if (uio) { ca->sndlen = uio->uio_resid; ca->m = sctp_copy_out_all(uio, ca->sndlen); if (ca->m == NULL) { SCTP_FREE(ca, SCTP_M_COPYAL); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } } else { /* Gather the length of the send */ struct mbuf *mat; mat = m; ca->sndlen = 0; while (m) { ca->sndlen += SCTP_BUF_LEN(m); m = SCTP_BUF_NEXT(m); } ca->m = mat; } ret = sctp_initiate_iterator(NULL, sctp_sendall_iterator, NULL, SCTP_PCB_ANY_FLAGS, SCTP_PCB_ANY_FEATURES, SCTP_ASOC_ANY_STATE, (void *)ca, 0, sctp_sendall_completes, inp, 1); if (ret) { SCTP_PRINTF("Failed to initiate iterator for sendall\n"); SCTP_FREE(ca, SCTP_M_COPYAL); SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); return (EFAULT); } return (0); } void sctp_toss_old_cookies(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk, *nchk; TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); } } } void sctp_toss_old_asconf(struct sctp_tcb *stcb) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *nchk; struct sctp_asconf_chunk *acp; asoc = &stcb->asoc; TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { /* find SCTP_ASCONF chunk in queue */ if (chk->rec.chunk_id.id == SCTP_ASCONF) { if (chk->data) { acp = mtod(chk->data, struct sctp_asconf_chunk *); if (SCTP_TSN_GT(ntohl(acp->serial_number), asoc->asconf_seq_out_acked)) { /* Not Acked yet */ break; } } TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); } } } static void sctp_clean_up_datalist(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_tmit_chunk **data_list, int bundle_at, struct sctp_nets *net) { int i; struct sctp_tmit_chunk *tp1; for (i = 0; i < bundle_at; i++) { /* off of the send queue */ TAILQ_REMOVE(&asoc->send_queue, data_list[i], sctp_next); asoc->send_queue_cnt--; if (i > 0) { /* * Any chunk NOT 0 you zap the time chunk 0 gets * zapped or set based on if a RTO measurment is * needed. */ data_list[i]->do_rtt = 0; } /* record time */ data_list[i]->sent_rcv_time = net->last_sent_time; data_list[i]->rec.data.fast_retran_tsn = data_list[i]->rec.data.TSN_seq; if (data_list[i]->whoTo == NULL) { data_list[i]->whoTo = net; atomic_add_int(&net->ref_count, 1); } /* on to the sent queue */ tp1 = TAILQ_LAST(&asoc->sent_queue, sctpchunk_listhead); if ((tp1) && SCTP_TSN_GT(tp1->rec.data.TSN_seq, data_list[i]->rec.data.TSN_seq)) { struct sctp_tmit_chunk *tpp; /* need to move back */ back_up_more: tpp = TAILQ_PREV(tp1, sctpchunk_listhead, sctp_next); if (tpp == NULL) { TAILQ_INSERT_BEFORE(tp1, data_list[i], sctp_next); goto all_done; } tp1 = tpp; if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, data_list[i]->rec.data.TSN_seq)) { goto back_up_more; } TAILQ_INSERT_AFTER(&asoc->sent_queue, tp1, data_list[i], sctp_next); } else { TAILQ_INSERT_TAIL(&asoc->sent_queue, data_list[i], sctp_next); } all_done: /* This does not lower until the cum-ack passes it */ asoc->sent_queue_cnt++; if ((asoc->peers_rwnd <= 0) && (asoc->total_flight == 0) && (bundle_at == 1)) { /* Mark the chunk as being a window probe */ SCTP_STAT_INCR(sctps_windowprobed); } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC2, 3); #endif data_list[i]->sent = SCTP_DATAGRAM_SENT; data_list[i]->snd_count = 1; data_list[i]->rec.data.chunk_was_revoked = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP, data_list[i]->whoTo->flight_size, data_list[i]->book_size, (uintptr_t) data_list[i]->whoTo, data_list[i]->rec.data.TSN_seq); } sctp_flight_size_increase(data_list[i]); sctp_total_flight_increase(stcb, data_list[i]); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); } asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, (uint32_t) (data_list[i]->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } } } static void sctp_clean_up_ctl(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk, *nchk; TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK) || /* EY */ (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || (chk->rec.chunk_id.id == SCTP_ECN_CWR) || (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { /* Stray chunks must be cleaned up */ clean_up_anyway: TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) asoc->fwd_tsn_cnt--; sctp_free_a_chunk(stcb, chk); } else if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { /* special handling, we must look into the param */ if (chk != asoc->str_reset) { goto clean_up_anyway; } } } } static int sctp_can_we_split_this(struct sctp_tcb *stcb, uint32_t length, uint32_t goal_mtu, uint32_t frag_point, int eeor_on) { /* * Make a decision on if I should split a msg into multiple parts. * This is only asked of incomplete messages. */ if (eeor_on) { /* * If we are doing EEOR we need to always send it if its the * entire thing, since it might be all the guy is putting in * the hopper. */ if (goal_mtu >= length) { /*- * If we have data outstanding, * we get another chance when the sack * arrives to transmit - wait for more data */ if (stcb->asoc.total_flight == 0) { /* * If nothing is in flight, we zero the * packet counter. */ return (length); } return (0); } else { /* You can fill the rest */ return (goal_mtu); } } /*- * For those strange folk that make the send buffer * smaller than our fragmentation point, we can't * get a full msg in so we have to allow splitting. */ if (SCTP_SB_LIMIT_SND(stcb->sctp_socket) < frag_point) { return (length); } if ((length <= goal_mtu) || ((length - goal_mtu) < SCTP_BASE_SYSCTL(sctp_min_residual))) { /* Sub-optimial residual don't split in non-eeor mode. */ return (0); } /* * If we reach here length is larger than the goal_mtu. Do we wish * to split it for the sake of packet putting together? */ if (goal_mtu >= min(SCTP_BASE_SYSCTL(sctp_min_split_point), frag_point)) { /* Its ok to split it */ return (min(goal_mtu, frag_point)); } /* Nope, can't split */ return (0); } static uint32_t sctp_move_to_outqueue(struct sctp_tcb *stcb, struct sctp_stream_out *strq, uint32_t goal_mtu, uint32_t frag_point, int *locked, int *giveup, int eeor_mode, int *bail) { /* Move from the stream to the send_queue keeping track of the total */ struct sctp_association *asoc; struct sctp_stream_queue_pending *sp; struct sctp_tmit_chunk *chk; struct sctp_data_chunk *dchkh; uint32_t to_move, length; uint8_t rcv_flags = 0; uint8_t some_taken; uint8_t send_lock_up = 0; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; one_more_time: /* sa_ignore FREED_MEMORY */ sp = TAILQ_FIRST(&strq->outqueue); if (sp == NULL) { *locked = 0; if (send_lock_up == 0) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } sp = TAILQ_FIRST(&strq->outqueue); if (sp) { goto one_more_time; } if (strq->last_msg_incomplete) { SCTP_PRINTF("Huh? Stream:%d lm_in_c=%d but queue is NULL\n", strq->stream_no, strq->last_msg_incomplete); strq->last_msg_incomplete = 0; } to_move = 0; if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } goto out_of; } if ((sp->msg_is_complete) && (sp->length == 0)) { if (sp->sender_all_done) { /* * We are doing differed cleanup. Last time through * when we took all the data the sender_all_done was * not set. */ if ((sp->put_last_out == 0) && (sp->discard_rest == 0)) { SCTP_PRINTF("Gak, put out entire msg with NO end!-1\n"); SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d send_lock:%d\n", sp->sender_all_done, sp->length, sp->msg_is_complete, sp->put_last_out, send_lock_up); } if ((TAILQ_NEXT(sp, next) == NULL) && (send_lock_up == 0)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } atomic_subtract_int(&asoc->stream_queue_cnt, 1); TAILQ_REMOVE(&strq->outqueue, sp, next); stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, strq, sp, send_lock_up); if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } sctp_free_a_strmoq(stcb, sp); /* we can't be locked to it */ *locked = 0; stcb->asoc.locked_on_sending = NULL; if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } /* back to get the next msg */ goto one_more_time; } else { /* * sender just finished this but still holds a * reference */ *locked = 1; *giveup = 1; to_move = 0; goto out_of; } } else { /* is there some to get */ if (sp->length == 0) { /* no */ *locked = 1; *giveup = 1; to_move = 0; goto out_of; } else if (sp->discard_rest) { if (send_lock_up == 0) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } /* Whack down the size */ atomic_subtract_int(&stcb->asoc.total_output_queue_size, sp->length); if ((stcb->sctp_socket != NULL) && \ ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { atomic_subtract_int(&stcb->sctp_socket->so_snd.sb_cc, sp->length); } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; sp->tail_mbuf = NULL; } sp->length = 0; sp->some_taken = 1; *locked = 1; *giveup = 1; to_move = 0; goto out_of; } } some_taken = sp->some_taken; if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { sp->msg_is_complete = 1; } re_look: length = sp->length; if (sp->msg_is_complete) { /* The message is complete */ to_move = min(length, frag_point); if (to_move == length) { /* All of it fits in the MTU */ if (sp->some_taken) { rcv_flags |= SCTP_DATA_LAST_FRAG; sp->put_last_out = 1; } else { rcv_flags |= SCTP_DATA_NOT_FRAG; sp->put_last_out = 1; } } else { /* Not all of it fits, we fragment */ if (sp->some_taken == 0) { rcv_flags |= SCTP_DATA_FIRST_FRAG; } sp->some_taken = 1; } } else { to_move = sctp_can_we_split_this(stcb, length, goal_mtu, frag_point, eeor_mode); if (to_move) { /*- * We use a snapshot of length in case it * is expanding during the compare. */ uint32_t llen; llen = length; if (to_move >= llen) { to_move = llen; if (send_lock_up == 0) { /*- * We are taking all of an incomplete msg * thus we need a send lock. */ SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; if (sp->msg_is_complete) { /* * the sender finished the * msg */ goto re_look; } } } if (sp->some_taken == 0) { rcv_flags |= SCTP_DATA_FIRST_FRAG; sp->some_taken = 1; } } else { /* Nothing to take. */ if (sp->some_taken) { *locked = 1; } *giveup = 1; to_move = 0; goto out_of; } } /* If we reach here, we can copy out a chunk */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* No chunk memory */ *giveup = 1; to_move = 0; goto out_of; } /* * Setup for unordered if needed by looking at the user sent info * flags. */ if (sp->sinfo_flags & SCTP_UNORDERED) { rcv_flags |= SCTP_DATA_UNORDERED; } if ((SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) && ((sp->sinfo_flags & SCTP_EOF) == SCTP_EOF)) || ((sp->sinfo_flags & SCTP_SACK_IMMEDIATELY) == SCTP_SACK_IMMEDIATELY)) { rcv_flags |= SCTP_DATA_SACK_IMMEDIATELY; } /* clear out the chunk before setting up */ memset(chk, 0, sizeof(*chk)); chk->rec.data.rcv_flags = rcv_flags; if (to_move >= length) { /* we think we can steal the whole thing */ if ((sp->sender_all_done == 0) && (send_lock_up == 0)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } if (to_move < sp->length) { /* bail, it changed */ goto dont_do_it; } chk->data = sp->data; chk->last_mbuf = sp->tail_mbuf; /* register the stealing */ sp->data = sp->tail_mbuf = NULL; } else { struct mbuf *m; dont_do_it: chk->data = SCTP_M_COPYM(sp->data, 0, to_move, M_DONTWAIT); chk->last_mbuf = NULL; if (chk->data == NULL) { sp->some_taken = some_taken; sctp_free_a_chunk(stcb, chk); *bail = 1; to_move = 0; goto out_of; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = chk->data; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif /* Pull off the data */ m_adj(sp->data, to_move); /* Now lets work our way down and compact it */ m = sp->data; while (m && (SCTP_BUF_LEN(m) == 0)) { sp->data = SCTP_BUF_NEXT(m); SCTP_BUF_NEXT(m) = NULL; if (sp->tail_mbuf == m) { /*- * Freeing tail? TSNH since * we supposedly were taking less * than the sp->length. */ #ifdef INVARIANTS panic("Huh, freing tail? - TSNH"); #else SCTP_PRINTF("Huh, freeing tail? - TSNH\n"); sp->tail_mbuf = sp->data = NULL; sp->length = 0; #endif } sctp_m_free(m); m = sp->data; } } if (SCTP_BUF_IS_EXTENDED(chk->data)) { chk->copy_by_ref = 1; } else { chk->copy_by_ref = 0; } /* * get last_mbuf and counts of mb useage This is ugly but hopefully * its only one mbuf. */ if (chk->last_mbuf == NULL) { chk->last_mbuf = chk->data; while (SCTP_BUF_NEXT(chk->last_mbuf) != NULL) { chk->last_mbuf = SCTP_BUF_NEXT(chk->last_mbuf); } } if (to_move > length) { /*- This should not happen either * since we always lower to_move to the size * of sp->length if its larger. */ #ifdef INVARIANTS panic("Huh, how can to_move be larger?"); #else SCTP_PRINTF("Huh, how can to_move be larger?\n"); sp->length = 0; #endif } else { atomic_subtract_int(&sp->length, to_move); } if (M_LEADINGSPACE(chk->data) < (int)sizeof(struct sctp_data_chunk)) { /* Not enough room for a chunk header, get some */ struct mbuf *m; m = sctp_get_mbuf_for_msg(1, 0, M_DONTWAIT, 0, MT_DATA); if (m == NULL) { /* * we're in trouble here. _PREPEND below will free * all the data if there is no leading space, so we * must put the data back and restore. */ if (send_lock_up == 0) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } if (chk->data == NULL) { /* unsteal the data */ sp->data = chk->data; sp->tail_mbuf = chk->last_mbuf; } else { struct mbuf *m_tmp; /* reassemble the data */ m_tmp = sp->data; sp->data = chk->data; SCTP_BUF_NEXT(chk->last_mbuf) = m_tmp; } sp->some_taken = some_taken; atomic_add_int(&sp->length, to_move); chk->data = NULL; *bail = 1; sctp_free_a_chunk(stcb, chk); to_move = 0; goto out_of; } else { SCTP_BUF_LEN(m) = 0; SCTP_BUF_NEXT(m) = chk->data; chk->data = m; M_ALIGN(chk->data, 4); } } SCTP_BUF_PREPEND(chk->data, sizeof(struct sctp_data_chunk), M_DONTWAIT); if (chk->data == NULL) { /* HELP, TSNH since we assured it would not above? */ #ifdef INVARIANTS panic("prepend failes HELP?"); #else SCTP_PRINTF("prepend fails HELP?\n"); sctp_free_a_chunk(stcb, chk); #endif *bail = 1; to_move = 0; goto out_of; } sctp_snd_sb_alloc(stcb, sizeof(struct sctp_data_chunk)); chk->book_size = chk->send_size = (to_move + sizeof(struct sctp_data_chunk)); chk->book_size_scale = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->flags = 0; chk->asoc = &stcb->asoc; chk->pad_inplace = 0; chk->no_fr_allowed = 0; chk->rec.data.stream_seq = sp->strseq; chk->rec.data.stream_number = sp->stream; chk->rec.data.payloadtype = sp->ppid; chk->rec.data.context = sp->context; chk->rec.data.doing_fast_retransmit = 0; - chk->rec.data.ect_nonce = 0; /* ECN Nonce */ chk->rec.data.timetodrop = sp->ts; chk->flags = sp->act_flags; if (sp->net) { chk->whoTo = sp->net; atomic_add_int(&chk->whoTo->ref_count, 1); } else chk->whoTo = NULL; if (sp->holds_key_ref) { chk->auth_keyid = sp->auth_keyid; sctp_auth_key_acquire(stcb, chk->auth_keyid); chk->holds_key_ref = 1; } chk->rec.data.TSN_seq = atomic_fetchadd_int(&asoc->sending_seq, 1); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_AT_SEND_2_OUTQ) { sctp_misc_ints(SCTP_STRMOUT_LOG_SEND, (uintptr_t) stcb, sp->length, (uint32_t) ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq), chk->rec.data.TSN_seq); } dchkh = mtod(chk->data, struct sctp_data_chunk *); /* * Put the rest of the things in place now. Size was done earlier in * previous loop prior to padding. */ #ifdef SCTP_ASOCLOG_OF_TSNS SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->tsn_out_at >= SCTP_TSN_LOG_SIZE) { asoc->tsn_out_at = 0; asoc->tsn_out_wrapped = 1; } asoc->out_tsnlog[asoc->tsn_out_at].tsn = chk->rec.data.TSN_seq; asoc->out_tsnlog[asoc->tsn_out_at].strm = chk->rec.data.stream_number; asoc->out_tsnlog[asoc->tsn_out_at].seq = chk->rec.data.stream_seq; asoc->out_tsnlog[asoc->tsn_out_at].sz = chk->send_size; asoc->out_tsnlog[asoc->tsn_out_at].flgs = chk->rec.data.rcv_flags; asoc->out_tsnlog[asoc->tsn_out_at].stcb = (void *)stcb; asoc->out_tsnlog[asoc->tsn_out_at].in_pos = asoc->tsn_out_at; asoc->out_tsnlog[asoc->tsn_out_at].in_out = 2; asoc->tsn_out_at++; #endif dchkh->ch.chunk_type = SCTP_DATA; dchkh->ch.chunk_flags = chk->rec.data.rcv_flags; dchkh->dp.tsn = htonl(chk->rec.data.TSN_seq); dchkh->dp.stream_id = htons(strq->stream_no); dchkh->dp.stream_sequence = htons(chk->rec.data.stream_seq); dchkh->dp.protocol_id = chk->rec.data.payloadtype; dchkh->ch.chunk_length = htons(chk->send_size); /* Now advance the chk->send_size by the actual pad needed. */ if (chk->send_size < SCTP_SIZE32(chk->book_size)) { /* need a pad */ struct mbuf *lm; int pads; pads = SCTP_SIZE32(chk->book_size) - chk->send_size; if (sctp_pad_lastmbuf(chk->data, pads, chk->last_mbuf) == 0) { chk->pad_inplace = 1; } if ((lm = SCTP_BUF_NEXT(chk->last_mbuf)) != NULL) { /* pad added an mbuf */ chk->last_mbuf = lm; } chk->send_size += pads; } /* We only re-set the policy if it is on */ if (sp->pr_sctp_on) { sctp_set_prsctp_policy(sp); asoc->pr_sctp_cnt++; chk->pr_sctp_on = 1; } else { chk->pr_sctp_on = 0; } if (sp->msg_is_complete && (sp->length == 0) && (sp->sender_all_done)) { /* All done pull and kill the message */ atomic_subtract_int(&asoc->stream_queue_cnt, 1); if (sp->put_last_out == 0) { SCTP_PRINTF("Gak, put out entire msg with NO end!-2\n"); SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d send_lock:%d\n", sp->sender_all_done, sp->length, sp->msg_is_complete, sp->put_last_out, send_lock_up); } if ((send_lock_up == 0) && (TAILQ_NEXT(sp, next) == NULL)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } TAILQ_REMOVE(&strq->outqueue, sp, next); stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, strq, sp, send_lock_up); if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } sctp_free_a_strmoq(stcb, sp); /* we can't be locked to it */ *locked = 0; stcb->asoc.locked_on_sending = NULL; } else { /* more to go, we are locked */ *locked = 1; } asoc->chunks_on_out_queue++; TAILQ_INSERT_TAIL(&asoc->send_queue, chk, sctp_next); asoc->send_queue_cnt++; out_of: if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } return (to_move); } static void sctp_fill_outqueue(struct sctp_tcb *stcb, struct sctp_nets *net, int frag_point, int eeor_mode, int *quit_now) { struct sctp_association *asoc; struct sctp_stream_out *strq, *strqn; int goal_mtu, moved_how_much, total_moved = 0, bail = 0; int locked, giveup; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; #ifdef INET6 if (net->ro._l_addr.sin6.sin6_family == AF_INET6) { goal_mtu = net->mtu - SCTP_MIN_OVERHEAD; } else { /* ?? not sure what else to do */ goal_mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; } #else goal_mtu = net->mtu - SCTP_MIN_OVERHEAD; #endif /* Need an allowance for the data chunk header too */ goal_mtu -= sizeof(struct sctp_data_chunk); /* must make even word boundary */ goal_mtu &= 0xfffffffc; if (asoc->locked_on_sending) { /* We are stuck on one stream until the message completes. */ strq = asoc->locked_on_sending; locked = 1; } else { strq = stcb->asoc.ss_functions.sctp_ss_select_stream(stcb, net, asoc); locked = 0; } strqn = strq; while ((goal_mtu > 0) && strq) { giveup = 0; bail = 0; moved_how_much = sctp_move_to_outqueue(stcb, strq, goal_mtu, frag_point, &locked, &giveup, eeor_mode, &bail); if (moved_how_much) stcb->asoc.ss_functions.sctp_ss_scheduled(stcb, net, asoc, strq, moved_how_much); if (locked) { asoc->locked_on_sending = strq; if ((moved_how_much == 0) || (giveup) || bail) /* no more to move for now */ break; } else { asoc->locked_on_sending = NULL; if ((giveup) || bail) { break; } strq = stcb->asoc.ss_functions.sctp_ss_select_stream(stcb, net, asoc); if (strq == NULL) { break; } } total_moved += moved_how_much; goal_mtu -= (moved_how_much + sizeof(struct sctp_data_chunk)); goal_mtu &= 0xfffffffc; } if (bail) *quit_now = 1; stcb->asoc.ss_functions.sctp_ss_packet_done(stcb, net, asoc); if (total_moved == 0) { if ((stcb->asoc.sctp_cmt_on_off == 0) && (net == stcb->asoc.primary_destination)) { /* ran dry for primary network net */ SCTP_STAT_INCR(sctps_primary_randry); } else if (stcb->asoc.sctp_cmt_on_off > 0) { /* ran dry with CMT on */ SCTP_STAT_INCR(sctps_cmt_randry); } } } void sctp_fix_ecn_echo(struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { chk->sent = SCTP_DATAGRAM_UNSENT; } } } void sctp_move_chunks_from_net(struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk; struct sctp_stream_queue_pending *sp; unsigned int i; if (net == NULL) { return; } asoc = &stcb->asoc; for (i = 0; i < stcb->asoc.streamoutcnt; i++) { TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) { if (sp->net == net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } } } TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { if (chk->whoTo == net) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = NULL; } } } int sctp_med_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *num_out, int *reason_code, int control_only, int from_where, struct timeval *now, int *now_filled, int frag_point, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /* * Ok this is the generic chunk service queue. we must do the * following: - Service the stream queue that is next, moving any * message (note I must get a complete message i.e. FIRST/MIDDLE and * LAST to the out queue in one pass) and assigning TSN's - Check to * see if the cwnd/rwnd allows any output, if so we go ahead and * fomulate and send the low level chunks. Making sure to combine * any control in the control chunk queue also. */ struct sctp_nets *net, *start_at, *sack_goes_to = NULL, *old_start_at = NULL; struct mbuf *outchain, *endoutchain; struct sctp_tmit_chunk *chk, *nchk; /* temp arrays for unlinking */ struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; int no_fragmentflg, error; unsigned int max_rwnd_per_dest, max_send_per_dest; int one_chunk, hbflag, skip_data_for_this_net; int asconf, cookie, no_out_cnt; int bundle_at, ctl_cnt, no_data_chunks, eeor_mode; unsigned int mtu, r_mtu, omtu, mx_mtu, to_out; int tsns_sent = 0; uint32_t auth_offset = 0; struct sctp_auth_chunk *auth = NULL; uint16_t auth_keyid; int override_ok = 1; int data_auth_reqd = 0; /* * JRS 5/14/07 - Add flag for whether a heartbeat is sent to the * destination. */ int pf_hbflag = 0; int quit_now = 0; *num_out = 0; auth_keyid = stcb->asoc.authinfo.active_keyid; if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED) || (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR))) { eeor_mode = 1; } else { eeor_mode = 0; } ctl_cnt = no_out_cnt = asconf = cookie = 0; /* * First lets prime the pump. For each destination, if there is room * in the flight size, attempt to pull an MTU's worth out of the * stream queues into the general send_queue */ #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC2, 2); #endif SCTP_TCB_LOCK_ASSERT(stcb); hbflag = 0; if ((control_only) || (asoc->stream_reset_outstanding)) no_data_chunks = 1; else no_data_chunks = 0; /* Nothing to possible to send? */ if ((TAILQ_EMPTY(&asoc->control_send_queue) || (asoc->ctrl_queue_cnt == stcb->asoc.ecn_echo_cnt_onq)) && TAILQ_EMPTY(&asoc->asconf_send_queue) && TAILQ_EMPTY(&asoc->send_queue) && stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { nothing_to_send: *reason_code = 9; return (0); } if (asoc->peers_rwnd == 0) { /* No room in peers rwnd */ *reason_code = 1; if (asoc->total_flight > 0) { /* we are allowed one chunk in flight */ no_data_chunks = 1; } } if (stcb->asoc.ecn_echo_cnt_onq) { /* Record where a sack goes, if any */ if (no_data_chunks && (asoc->ctrl_queue_cnt == stcb->asoc.ecn_echo_cnt_onq)) { /* Nothing but ECNe to send - we don't do that */ goto nothing_to_send; } TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK)) { sack_goes_to = chk->whoTo; break; } } } max_rwnd_per_dest = ((asoc->peers_rwnd + asoc->total_flight) / asoc->numnets); if (stcb->sctp_socket) max_send_per_dest = SCTP_SB_LIMIT_SND(stcb->sctp_socket) / asoc->numnets; else max_send_per_dest = 0; if ((no_data_chunks == 0) && (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc))) { TAILQ_FOREACH(net, &asoc->nets, sctp_next) { /* * This for loop we are in takes in each net, if * its's got space in cwnd and has data sent to it * (when CMT is off) then it calls * sctp_fill_outqueue for the net. This gets data on * the send queue for that network. * * In sctp_fill_outqueue TSN's are assigned and data is * copied out of the stream buffers. Note mostly * copy by reference (we hope). */ net->window_probe = 0; if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) || (net->dest_state & SCTP_ADDR_UNCONFIRMED)) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 1, SCTP_CWND_LOG_FILL_OUTQ_CALLED); } continue; } if ((asoc->sctp_cmt_on_off == 0) && (asoc->primary_destination != net) && (net->ref_count < 2)) { /* nothing can be in queue for this guy */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 2, SCTP_CWND_LOG_FILL_OUTQ_CALLED); } continue; } if (net->flight_size >= net->cwnd) { /* skip this network, no room - can't fill */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 3, SCTP_CWND_LOG_FILL_OUTQ_CALLED); } continue; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 4, SCTP_CWND_LOG_FILL_OUTQ_CALLED); } sctp_fill_outqueue(stcb, net, frag_point, eeor_mode, &quit_now); if (quit_now) { /* memory alloc failure */ no_data_chunks = 1; break; } } } /* now service each destination and send out what we can for it */ /* Nothing to send? */ if (TAILQ_EMPTY(&asoc->control_send_queue) && TAILQ_EMPTY(&asoc->asconf_send_queue) && TAILQ_EMPTY(&asoc->send_queue)) { *reason_code = 8; return (0); } if (asoc->sctp_cmt_on_off > 0) { /* get the last start point */ start_at = asoc->last_net_cmt_send_started; if (start_at == NULL) { /* null so to beginning */ start_at = TAILQ_FIRST(&asoc->nets); } else { start_at = TAILQ_NEXT(asoc->last_net_cmt_send_started, sctp_next); if (start_at == NULL) { start_at = TAILQ_FIRST(&asoc->nets); } } asoc->last_net_cmt_send_started = start_at; } else { start_at = TAILQ_FIRST(&asoc->nets); } old_start_at = NULL; again_one_more_time: for (net = start_at; net != NULL; net = TAILQ_NEXT(net, sctp_next)) { /* how much can we send? */ /* SCTPDBG("Examine for sending net:%x\n", (uint32_t)net); */ if (old_start_at && (old_start_at == net)) { /* through list ocmpletely. */ break; } tsns_sent = 0xa; if ((asoc->sctp_cmt_on_off == 0) && (asoc->primary_destination != net) && (net->ref_count < 2)) { /* * Ref-count of 1 so we cannot have data or control * queued to this address. Skip it (non-CMT). */ continue; } if (TAILQ_EMPTY(&asoc->control_send_queue) && TAILQ_EMPTY(&asoc->asconf_send_queue) && (net->flight_size >= net->cwnd)) { /* * Nothing on control or asconf and flight is full, * we can skip even in the CMT case. */ continue; } ctl_cnt = bundle_at = 0; endoutchain = outchain = NULL; no_fragmentflg = 1; one_chunk = 0; if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { skip_data_for_this_net = 1; } else { skip_data_for_this_net = 0; } if ((net->ro.ro_rt) && (net->ro.ro_rt->rt_ifp)) { /* * if we have a route and an ifp check to see if we * have room to send to this guy */ struct ifnet *ifp; ifp = net->ro.ro_rt->rt_ifp; if ((ifp->if_snd.ifq_len + 2) >= ifp->if_snd.ifq_maxlen) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, net, ifp->if_snd.ifq_len, ifp->if_snd.ifq_maxlen, SCTP_MAX_IFP_APPLIED); } continue; } } switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { case AF_INET: mtu = net->mtu - (sizeof(struct ip) + sizeof(struct sctphdr)); break; #ifdef INET6 case AF_INET6: mtu = net->mtu - (sizeof(struct ip6_hdr) + sizeof(struct sctphdr)); break; #endif default: /* TSNH */ mtu = net->mtu; break; } mx_mtu = mtu; to_out = 0; if (mtu > asoc->peers_rwnd) { if (asoc->total_flight > 0) { /* We have a packet in flight somewhere */ r_mtu = asoc->peers_rwnd; } else { /* We are always allowed to send one MTU out */ one_chunk = 1; r_mtu = mtu; } } else { r_mtu = mtu; } /************************/ /* ASCONF transmission */ /************************/ /* Now first lets go through the asconf queue */ TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { if (chk->rec.chunk_id.id != SCTP_ASCONF) { continue; } if (chk->whoTo != net) { /* * No, not sent to the network we are * looking at */ break; } if (chk->data == NULL) { break; } if (chk->sent != SCTP_DATAGRAM_UNSENT && chk->sent != SCTP_DATAGRAM_RESEND) { break; } /* * if no AUTH is yet included and this chunk * requires it, make sure to account for it. We * don't apply the size until the AUTH chunk is * actually added below in case there is no room for * this chunk. NOTE: we overload the use of "omtu" * here */ if ((auth == NULL) && sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks)) { omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else omtu = 0; /* Here we do NOT factor the r_mtu */ if ((chk->send_size < (int)(mtu - omtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* * We probably should glom the mbuf chain * from the chk->data for control but the * problem is it becomes yet one more level * of tracking to do if for some reason * output fails. Then I have got to * reconstruct the merged control chain.. el * yucko.. for now we take the easy way and * do the copy */ /* * Add an AUTH chunk, if chunk requires it * save the offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, (int)chk->rec.chunk_id.can_take_data, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { *reason_code = 8; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); /* update our MTU size */ if (mtu > (chk->send_size + omtu)) mtu -= (chk->send_size + omtu); else mtu = 0; to_out += (chk->send_size + omtu); /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } if (chk->rec.chunk_id.can_take_data) chk->data = NULL; /* * set hb flag since we can use these for * RTO */ hbflag = 1; asconf = 1; /* * should sysctl this: don't bundle data * with ASCONF since it requires AUTH */ no_data_chunks = 1; chk->sent = SCTP_DATAGRAM_SENT; chk->snd_count++; if (mtu == 0) { /* * Ok we are out of room but we can * output without effecting the * flight size since this little guy * is a control only packet. */ sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf flag as * it is used to do appropriate * source address selection. */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, stcb->asoc.authinfo.active_keyid, no_fragmentflg, 0, NULL, asconf, inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), net->port, so_locked, NULL))) { if (error == ENOBUFS) { asoc->ifp_had_enobuf = 1; SCTP_STAT_INCR(sctps_lowlevelerr); } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; /* error, could not output */ if (error == EHOSTUNREACH) { /* * Destination went * unreachable * during this send */ sctp_move_chunks_from_net(stcb, net); } *reason_code = 7; continue; } else asoc->ifp_had_enobuf = 0; if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; /* * increase the number we sent, if a * cookie is sent we don't tell them * any was sent out. */ outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (!no_out_cnt) *num_out += ctl_cnt; /* recalc a clean slate and setup */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = (net->mtu - SCTP_MIN_V4_OVERHEAD); } to_out = 0; no_fragmentflg = 1; } } } /************************/ /* Control transmission */ /************************/ /* Now first lets go through the control queue */ TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { if ((sack_goes_to) && (chk->rec.chunk_id.id == SCTP_ECN_ECHO) && (chk->whoTo != sack_goes_to)) { /* * if we have a sack in queue, and we are * looking at an ecn echo that is NOT queued * to where the sack is going.. */ if (chk->whoTo == net) { /* * Don't transmit it to where its * going (current net) */ continue; } else if (sack_goes_to == net) { /* * But do transmit it to this * address */ goto skip_net_check; } } if (chk->whoTo != net) { /* * No, not sent to the network we are * looking at */ continue; } skip_net_check: if (chk->data == NULL) { continue; } if (chk->sent != SCTP_DATAGRAM_UNSENT) { /* * It must be unsent. Cookies and ASCONF's * hang around but there timers will force * when marked for resend. */ continue; } /* * if no AUTH is yet included and this chunk * requires it, make sure to account for it. We * don't apply the size until the AUTH chunk is * actually added below in case there is no room for * this chunk. NOTE: we overload the use of "omtu" * here */ if ((auth == NULL) && sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks)) { omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else omtu = 0; /* Here we do NOT factor the r_mtu */ if ((chk->send_size <= (int)(mtu - omtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* * We probably should glom the mbuf chain * from the chk->data for control but the * problem is it becomes yet one more level * of tracking to do if for some reason * output fails. Then I have got to * reconstruct the merged control chain.. el * yucko.. for now we take the easy way and * do the copy */ /* * Add an AUTH chunk, if chunk requires it * save the offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, (int)chk->rec.chunk_id.can_take_data, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { *reason_code = 8; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); /* update our MTU size */ if (mtu > (chk->send_size + omtu)) mtu -= (chk->send_size + omtu); else mtu = 0; to_out += (chk->send_size + omtu); /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } if (chk->rec.chunk_id.can_take_data) chk->data = NULL; /* Mark things to be removed, if needed */ if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK) || /* EY */ (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || (chk->rec.chunk_id.id == SCTP_ECN_CWR) || (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { if (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) { hbflag = 1; /* * JRS 5/14/07 - Set the * flag to say a heartbeat * is being sent. */ pf_hbflag = 1; } /* remove these chunks at the end */ if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK)) { /* turn off the timer */ if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_1); } } ctl_cnt++; } else { /* * Other chunks, since they have * timers running (i.e. COOKIE) we * just "trust" that it gets sent or * retransmitted. */ ctl_cnt++; if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { cookie = 1; no_out_cnt = 1; } else if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { /* * Increment ecne send count * here this means we may be * over-zealous in our * counting if the send * fails, but its the best * place to do it (we used * to do it in the queue of * the chunk, but that did * not tell how many times * it was sent. */ SCTP_STAT_INCR(sctps_sendecne); } chk->sent = SCTP_DATAGRAM_SENT; chk->snd_count++; } if (mtu == 0) { /* * Ok we are out of room but we can * output without effecting the * flight size since this little guy * is a control only packet. */ if (asconf) { sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf * flag as it is used to do * appropriate source * address selection. */ } if (cookie) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); cookie = 0; } if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, stcb->asoc.authinfo.active_keyid, no_fragmentflg, 0, NULL, asconf, inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), net->port, so_locked, NULL))) { if (error == ENOBUFS) { asoc->ifp_had_enobuf = 1; SCTP_STAT_INCR(sctps_lowlevelerr); } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } /* error, could not output */ if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } if (error == EHOSTUNREACH) { /* * Destination went * unreachable * during this send */ sctp_move_chunks_from_net(stcb, net); } *reason_code = 7; continue; } else asoc->ifp_had_enobuf = 0; /* Only HB or ASCONF advances time */ if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } /* * increase the number we sent, if a * cookie is sent we don't tell them * any was sent out. */ outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (!no_out_cnt) *num_out += ctl_cnt; /* recalc a clean slate and setup */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = (net->mtu - SCTP_MIN_V4_OVERHEAD); } to_out = 0; no_fragmentflg = 1; } } } /* JRI: if dest is in PF state, do not send data to it */ if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && (net->dest_state & SCTP_ADDR_PF)) { goto no_data_fill; } if (net->flight_size >= net->cwnd) { goto no_data_fill; } if ((asoc->sctp_cmt_on_off > 0) && (SCTP_BASE_SYSCTL(sctp_buffer_splitting) & SCTP_RECV_BUFFER_SPLITTING) && (net->flight_size > max_rwnd_per_dest)) { goto no_data_fill; } /* * We need a specific accounting for the usage of the send * buffer. We also need to check the number of messages per * net. For now, this is better than nothing and it disabled * by default... */ if ((asoc->sctp_cmt_on_off > 0) && (SCTP_BASE_SYSCTL(sctp_buffer_splitting) & SCTP_SEND_BUFFER_SPLITTING) && (max_send_per_dest > 0) && (net->flight_size > max_send_per_dest)) { goto no_data_fill; } /*********************/ /* Data transmission */ /*********************/ /* * if AUTH for DATA is required and no AUTH has been added * yet, account for this in the mtu now... if no data can be * bundled, this adjustment won't matter anyways since the * packet will be going out... */ data_auth_reqd = sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks); if (data_auth_reqd && (auth == NULL)) { mtu -= sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } /* now lets add any data within the MTU constraints */ switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { case AF_INET: if (net->mtu > (sizeof(struct ip) + sizeof(struct sctphdr))) omtu = net->mtu - (sizeof(struct ip) + sizeof(struct sctphdr)); else omtu = 0; break; #ifdef INET6 case AF_INET6: if (net->mtu > (sizeof(struct ip6_hdr) + sizeof(struct sctphdr))) omtu = net->mtu - (sizeof(struct ip6_hdr) + sizeof(struct sctphdr)); else omtu = 0; break; #endif default: /* TSNH */ omtu = 0; break; } if ((((asoc->state & SCTP_STATE_OPEN) == SCTP_STATE_OPEN) && (skip_data_for_this_net == 0)) || (cookie)) { TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { if (no_data_chunks) { /* let only control go out */ *reason_code = 1; break; } if (net->flight_size >= net->cwnd) { /* skip this net, no room for data */ *reason_code = 2; break; } if ((chk->whoTo != NULL) && (chk->whoTo != net)) { /* Don't send the chunk on this net */ continue; } if ((chk->send_size > omtu) && ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) == 0)) { /*- * strange, we have a chunk that is * to big for its destination and * yet no fragment ok flag. * Something went wrong when the * PMTU changed...we did not mark * this chunk for some reason?? I * will fix it here by letting IP * fragment it for now and printing * a warning. This really should not * happen ... */ SCTP_PRINTF("Warning chunk of %d bytes > mtu:%d and yet PMTU disc missed\n", chk->send_size, mtu); chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } if (SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) && ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) == SCTP_STATE_SHUTDOWN_PENDING)) { struct sctp_data_chunk *dchkh; dchkh = mtod(chk->data, struct sctp_data_chunk *); dchkh->ch.chunk_flags |= SCTP_DATA_SACK_IMMEDIATELY; } if (((chk->send_size <= mtu) && (chk->send_size <= r_mtu)) || ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) && (chk->send_size <= asoc->peers_rwnd))) { /* ok we will add this one */ /* * Add an AUTH chunk, if chunk * requires it, save the offset into * the chain for AUTH */ if (data_auth_reqd) { if (auth == NULL) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, SCTP_DATA); auth_keyid = chk->auth_keyid; override_ok = 0; SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } else if (override_ok) { /* * use this data's * keyid */ auth_keyid = chk->auth_keyid; override_ok = 0; } else if (auth_keyid != chk->auth_keyid) { /* * different keyid, * so done bundling */ break; } } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, 0, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "No memory?\n"); if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } *reason_code = 3; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* upate our MTU size */ /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* unsigned subtraction of mtu */ if (mtu > chk->send_size) mtu -= chk->send_size; else mtu = 0; /* unsigned subtraction of r_mtu */ if (r_mtu > chk->send_size) r_mtu -= chk->send_size; else r_mtu = 0; to_out += chk->send_size; if ((to_out > mx_mtu) && no_fragmentflg) { #ifdef INVARIANTS panic("Exceeding mtu of %d out size is %d", mx_mtu, to_out); #else SCTP_PRINTF("Exceeding mtu of %d out size is %d\n", mx_mtu, to_out); #endif } chk->window_probe = 0; data_list[bundle_at++] = chk; if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { mtu = 0; break; } if (chk->sent == SCTP_DATAGRAM_UNSENT) { if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) { SCTP_STAT_INCR_COUNTER64(sctps_outorderchunks); } else { SCTP_STAT_INCR_COUNTER64(sctps_outunorderchunks); } if (((chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) == SCTP_DATA_LAST_FRAG) && ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0)) /* * Count number of * user msg's that * were fragmented * we do this by * counting when we * see a LAST * fragment only. */ SCTP_STAT_INCR_COUNTER64(sctps_fragusrmsgs); } if ((mtu == 0) || (r_mtu == 0) || (one_chunk)) { if ((one_chunk) && (stcb->asoc.total_flight == 0)) { data_list[0]->window_probe = 1; net->window_probe = 1; } break; } } else { /* * Must be sent in order of the * TSN's (on a network) */ break; } } /* for (chunk gather loop for this net) */ } /* if asoc.state OPEN */ no_data_fill: /* Is there something to send for this destination? */ if (outchain) { /* We may need to start a control timer or two */ if (asconf) { sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf flag as it is * used to do appropriate source address * selection. */ } if (cookie) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); cookie = 0; } /* must start a send timer if data is being sent */ if (bundle_at && (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer))) { /* * no timer running on this destination * restart it. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } else if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && pf_hbflag && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF) && (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer))) { /* * JRS 5/14/07 - If a HB has been sent to a * PF destination and no T3 timer is * currently running, start the T3 timer to * track the HBs that were sent. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } /* Now send it, if there is anything to send :> */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, auth_keyid, no_fragmentflg, bundle_at, data_list[0], asconf, inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), net->port, so_locked, NULL))) { /* error, we could not output */ if (error == ENOBUFS) { SCTP_STAT_INCR(sctps_lowlevelerr); asoc->ifp_had_enobuf = 1; } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } if (error == EHOSTUNREACH) { /* * Destination went unreachable * during this send */ sctp_move_chunks_from_net(stcb, net); } *reason_code = 6; /*- * I add this line to be paranoid. As far as * I can tell the continue, takes us back to * the top of the for, but just to make sure * I will reset these again here. */ ctl_cnt = bundle_at = 0; continue; /* This takes us back to the * for() for the nets. */ } else { asoc->ifp_had_enobuf = 0; } outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (bundle_at || hbflag) { /* For data/asconf and hb set time */ if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } } if (!no_out_cnt) { *num_out += (ctl_cnt + bundle_at); } if (bundle_at) { /* setup for a RTO measurement */ tsns_sent = data_list[0]->rec.data.TSN_seq; /* fill time if not already filled */ if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); *now_filled = 1; *now = asoc->time_last_sent; } else { asoc->time_last_sent = *now; } data_list[0]->do_rtt = 1; SCTP_STAT_INCR_BY(sctps_senddata, bundle_at); sctp_clean_up_datalist(stcb, asoc, data_list, bundle_at, net); if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (net->flight_size < net->cwnd) { /* start or restart it */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_2); } SCTP_STAT_INCR(sctps_earlyfrstrout); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net); } else { /* stop it if its running */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpout); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_3); } } } } if (one_chunk) { break; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_SEND); } } if (old_start_at == NULL) { old_start_at = start_at; start_at = TAILQ_FIRST(&asoc->nets); if (old_start_at) goto again_one_more_time; } /* * At the end there should be no NON timed chunks hanging on this * queue. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, *num_out, SCTP_CWND_LOG_FROM_SEND); } if ((*num_out == 0) && (*reason_code == 0)) { *reason_code = 4; } else { *reason_code = 5; } sctp_clean_up_ctl(stcb, asoc); return (0); } void sctp_queue_op_err(struct sctp_tcb *stcb, struct mbuf *op_err) { /*- * Prepend a OPERATIONAL_ERROR chunk header and put on the end of * the control chunk queue. */ struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; struct mbuf *mat; SCTP_TCB_LOCK_ASSERT(stcb); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(op_err); return; } chk->copy_by_ref = 0; SCTP_BUF_PREPEND(op_err, sizeof(struct sctp_chunkhdr), M_DONTWAIT); if (op_err == NULL) { sctp_free_a_chunk(stcb, chk); return; } chk->send_size = 0; mat = op_err; while (mat != NULL) { chk->send_size += SCTP_BUF_LEN(mat); mat = SCTP_BUF_NEXT(mat); } chk->rec.chunk_id.id = SCTP_OPERATION_ERROR; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = op_err; chk->whoTo = chk->asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); hdr = mtod(op_err, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_OPERATION_ERROR; hdr->chunk_flags = 0; hdr->chunk_length = htons(chk->send_size); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } int sctp_send_cookie_echo(struct mbuf *m, int offset, struct sctp_tcb *stcb, struct sctp_nets *net) { /*- * pull out the cookie and put it at the front of the control chunk * queue. */ int at; struct mbuf *cookie; struct sctp_paramhdr parm, *phdr; struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; uint16_t ptype, plen; /* First find the cookie in the param area */ cookie = NULL; at = offset + sizeof(struct sctp_init_chunk); SCTP_TCB_LOCK_ASSERT(stcb); do { phdr = sctp_get_next_param(m, at, &parm, sizeof(parm)); if (phdr == NULL) { return (-3); } ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (ptype == SCTP_STATE_COOKIE) { int pad; /* found the cookie */ if ((pad = (plen % 4))) { plen += 4 - pad; } cookie = SCTP_M_COPYM(m, at, plen, M_DONTWAIT); if (cookie == NULL) { /* No memory */ return (-2); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = cookie; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif break; } at += SCTP_SIZE32(plen); } while (phdr); if (cookie == NULL) { /* Did not find the cookie */ return (-3); } /* ok, we got the cookie lets change it into a cookie echo chunk */ /* first the change from param to cookie */ hdr = mtod(cookie, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_COOKIE_ECHO; hdr->chunk_flags = 0; /* get the chunk stuff now and place it in the FRONT of the queue */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(cookie); return (-5); } chk->copy_by_ref = 0; chk->send_size = plen; chk->rec.chunk_id.id = SCTP_COOKIE_ECHO; chk->rec.chunk_id.can_take_data = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = CHUNK_FLAGS_FRAGMENT_OK; chk->asoc = &stcb->asoc; chk->data = cookie; chk->whoTo = chk->asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_HEAD(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return (0); } void sctp_send_heartbeat_ack(struct sctp_tcb *stcb, struct mbuf *m, int offset, int chk_length, struct sctp_nets *net) { /* * take a HB request and make it into a HB ack and send it. */ struct mbuf *outchain; struct sctp_chunkhdr *chdr; struct sctp_tmit_chunk *chk; if (net == NULL) /* must have a net pointer */ return; outchain = SCTP_M_COPYM(m, offset, chk_length, M_DONTWAIT); if (outchain == NULL) { /* gak out of memory */ return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = outchain; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif chdr = mtod(outchain, struct sctp_chunkhdr *); chdr->chunk_type = SCTP_HEARTBEAT_ACK; chdr->chunk_flags = 0; if (chk_length % 4) { /* need pad */ uint32_t cpthis = 0; int padlen; padlen = 4 - (chk_length % 4); m_copyback(outchain, chk_length, padlen, (caddr_t)&cpthis); } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(outchain); return; } chk->copy_by_ref = 0; chk->send_size = chk_length; chk->rec.chunk_id.id = SCTP_HEARTBEAT_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = outchain; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } void sctp_send_cookie_ack(struct sctp_tcb *stcb) { /* formulate and queue a cookie-ack back to sender */ struct mbuf *cookie_ack; struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; cookie_ack = NULL; SCTP_TCB_LOCK_ASSERT(stcb); cookie_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_chunkhdr), 0, M_DONTWAIT, 1, MT_HEADER); if (cookie_ack == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(cookie_ack, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(cookie_ack); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_chunkhdr); chk->rec.chunk_id.id = SCTP_COOKIE_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = cookie_ack; if (chk->asoc->last_control_chunk_from != NULL) { chk->whoTo = chk->asoc->last_control_chunk_from; } else { chk->whoTo = chk->asoc->primary_destination; } atomic_add_int(&chk->whoTo->ref_count, 1); hdr = mtod(cookie_ack, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_COOKIE_ACK; hdr->chunk_flags = 0; hdr->chunk_length = htons(chk->send_size); SCTP_BUF_LEN(cookie_ack) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_shutdown_ack(struct sctp_tcb *stcb, struct sctp_nets *net) { /* formulate and queue a SHUTDOWN-ACK back to the sender */ struct mbuf *m_shutdown_ack; struct sctp_shutdown_ack_chunk *ack_cp; struct sctp_tmit_chunk *chk; m_shutdown_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_ack_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown_ack == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(m_shutdown_ack, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_shutdown_ack); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_chunkhdr); chk->rec.chunk_id.id = SCTP_SHUTDOWN_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = m_shutdown_ack; chk->whoTo = net; atomic_add_int(&net->ref_count, 1); ack_cp = mtod(m_shutdown_ack, struct sctp_shutdown_ack_chunk *); ack_cp->ch.chunk_type = SCTP_SHUTDOWN_ACK; ack_cp->ch.chunk_flags = 0; ack_cp->ch.chunk_length = htons(chk->send_size); SCTP_BUF_LEN(m_shutdown_ack) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_shutdown(struct sctp_tcb *stcb, struct sctp_nets *net) { /* formulate and queue a SHUTDOWN to the sender */ struct mbuf *m_shutdown; struct sctp_shutdown_chunk *shutdown_cp; struct sctp_tmit_chunk *chk; m_shutdown = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(m_shutdown, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_shutdown); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_shutdown_chunk); chk->rec.chunk_id.id = SCTP_SHUTDOWN; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = m_shutdown; chk->whoTo = net; atomic_add_int(&net->ref_count, 1); shutdown_cp = mtod(m_shutdown, struct sctp_shutdown_chunk *); shutdown_cp->ch.chunk_type = SCTP_SHUTDOWN; shutdown_cp->ch.chunk_flags = 0; shutdown_cp->ch.chunk_length = htons(chk->send_size); shutdown_cp->cumulative_tsn_ack = htonl(stcb->asoc.cumulative_tsn); SCTP_BUF_LEN(m_shutdown) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_asconf(struct sctp_tcb *stcb, struct sctp_nets *net, int addr_locked) { /* * formulate and queue an ASCONF to the peer. ASCONF parameters * should be queued on the assoc queue. */ struct sctp_tmit_chunk *chk; struct mbuf *m_asconf; int len; SCTP_TCB_LOCK_ASSERT(stcb); if ((!TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) && (!sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS))) { /* can't send a new one if there is one in flight already */ return; } /* compose an ASCONF chunk, maximum length is PMTU */ m_asconf = sctp_compose_asconf(stcb, &len, addr_locked); if (m_asconf == NULL) { return; } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_asconf); return; } chk->copy_by_ref = 0; chk->data = m_asconf; chk->send_size = len; chk->rec.chunk_id.id = SCTP_ASCONF; chk->rec.chunk_id.can_take_data = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = CHUNK_FLAGS_FRAGMENT_OK; chk->asoc = &stcb->asoc; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->asconf_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_asconf_ack(struct sctp_tcb *stcb) { /* * formulate and queue a asconf-ack back to sender. the asconf-ack * must be stored in the tcb. */ struct sctp_tmit_chunk *chk; struct sctp_asconf_ack *ack, *latest_ack; struct mbuf *m_ack, *m; struct sctp_nets *net = NULL; SCTP_TCB_LOCK_ASSERT(stcb); /* Get the latest ASCONF-ACK */ latest_ack = TAILQ_LAST(&stcb->asoc.asconf_ack_sent, sctp_asconf_ackhead); if (latest_ack == NULL) { return; } if (latest_ack->last_sent_to != NULL && latest_ack->last_sent_to == stcb->asoc.last_control_chunk_from) { /* we're doing a retransmission */ net = sctp_find_alternate_net(stcb, stcb->asoc.last_control_chunk_from, 0); if (net == NULL) { /* no alternate */ if (stcb->asoc.last_control_chunk_from == NULL) net = stcb->asoc.primary_destination; else net = stcb->asoc.last_control_chunk_from; } } else { /* normal case */ if (stcb->asoc.last_control_chunk_from == NULL) net = stcb->asoc.primary_destination; else net = stcb->asoc.last_control_chunk_from; } latest_ack->last_sent_to = net; TAILQ_FOREACH(ack, &stcb->asoc.asconf_ack_sent, next) { if (ack->data == NULL) { continue; } /* copy the asconf_ack */ m_ack = SCTP_M_COPYM(ack->data, 0, M_COPYALL, M_DONTWAIT); if (m_ack == NULL) { /* couldn't copy it */ return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = m_ack; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ if (m_ack) sctp_m_freem(m_ack); return; } chk->copy_by_ref = 0; chk->whoTo = net; chk->data = m_ack; chk->send_size = 0; /* Get size */ m = m_ack; chk->send_size = ack->len; chk->rec.chunk_id.id = SCTP_ASCONF_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; /* XXX */ chk->asoc = &stcb->asoc; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } return; } static int sctp_chunk_retransmission(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *cnt_out, struct timeval *now, int *now_filled, int *fr_done, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /*- * send out one MTU of retransmission. If fast_retransmit is * happening we ignore the cwnd. Otherwise we obey the cwnd and * rwnd. For a Cookie or Asconf in the control chunk queue we * retransmit them by themselves. * * For data chunks we will pick out the lowest TSN's in the sent_queue * marked for resend and bundle them all together (up to a MTU of * destination). The address to send to should have been * selected/changed where the retransmission was marked (i.e. in FR * or t3-timeout routines). */ struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; struct sctp_tmit_chunk *chk, *fwd; struct mbuf *m, *endofchain; struct sctp_nets *net = NULL; uint32_t tsns_sent = 0; int no_fragmentflg, bundle_at, cnt_thru; unsigned int mtu; int error, i, one_chunk, fwd_tsn, ctl_cnt, tmr_started; struct sctp_auth_chunk *auth = NULL; uint32_t auth_offset = 0; uint16_t auth_keyid; int override_ok = 1; int data_auth_reqd = 0; uint32_t dmtu = 0; SCTP_TCB_LOCK_ASSERT(stcb); tmr_started = ctl_cnt = bundle_at = error = 0; no_fragmentflg = 1; fwd_tsn = 0; *cnt_out = 0; fwd = NULL; endofchain = m = NULL; auth_keyid = stcb->asoc.authinfo.active_keyid; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC3, 1); #endif if ((TAILQ_EMPTY(&asoc->sent_queue)) && (TAILQ_EMPTY(&asoc->control_send_queue))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "SCTP hits empty queue with cnt set to %d?\n", asoc->sent_queue_retran_cnt); asoc->sent_queue_cnt = 0; asoc->sent_queue_cnt_removeable = 0; /* send back 0/0 so we enter normal transmission */ *cnt_out = 0; return (0); } TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if ((chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) || (chk->rec.chunk_id.id == SCTP_STREAM_RESET) || (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN)) { if (chk->sent != SCTP_DATAGRAM_RESEND) { continue; } if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { if (chk != asoc->str_reset) { /* * not eligible for retran if its * not ours */ continue; } } ctl_cnt++; if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { fwd_tsn = 1; fwd = chk; } /* * Add an AUTH chunk, if chunk requires it save the * offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); break; } } one_chunk = 0; cnt_thru = 0; /* do we have control chunks to retransmit? */ if (m != NULL) { /* Start a timer no matter if we suceed or fail */ if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, chk->whoTo); } else if (chk->rec.chunk_id.id == SCTP_ASCONF) sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, chk->whoTo); chk->snd_count++; /* update our count */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, chk->whoTo, (struct sockaddr *)&chk->whoTo->ro._l_addr, m, auth_offset, auth, stcb->asoc.authinfo.active_keyid, no_fragmentflg, 0, NULL, 0, inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), chk->whoTo->port, so_locked, NULL))) { SCTP_STAT_INCR(sctps_lowlevelerr); return (error); } m = endofchain = NULL; auth = NULL; auth_offset = 0; /* * We don't want to mark the net->sent time here since this * we use this for HB and retrans cannot measure RTT */ /* (void)SCTP_GETTIME_TIMEVAL(&chk->whoTo->last_sent_time); */ *cnt_out += 1; chk->sent = SCTP_DATAGRAM_SENT; sctp_ucount_decr(stcb->asoc.sent_queue_retran_cnt); if (fwd_tsn == 0) { return (0); } else { /* Clean up the fwd-tsn list */ sctp_clean_up_ctl(stcb, asoc); return (0); } } /* * Ok, it is just data retransmission we need to do or that and a * fwd-tsn with it all. */ if (TAILQ_EMPTY(&asoc->sent_queue)) { return (SCTP_RETRAN_DONE); } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT)) { /* not yet open, resend the cookie and that is it */ return (1); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(20, inp, stcb, NULL); #endif data_auth_reqd = sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks); TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->sent != SCTP_DATAGRAM_RESEND) { /* No, not sent to this net or not ready for rtx */ continue; } if (chk->data == NULL) { printf("TSN:%x chk->snd_count:%d chk->sent:%d can't retran - no data\n", chk->rec.data.TSN_seq, chk->snd_count, chk->sent); continue; } if ((SCTP_BASE_SYSCTL(sctp_max_retran_chunk)) && (chk->snd_count >= SCTP_BASE_SYSCTL(sctp_max_retran_chunk))) { /* Gak, we have exceeded max unlucky retran, abort! */ SCTP_PRINTF("Gak, chk->snd_count:%d >= max:%d - send abort\n", chk->snd_count, SCTP_BASE_SYSCTL(sctp_max_retran_chunk)); atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(stcb->sctp_ep, stcb, 0, NULL, so_locked); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); return (SCTP_RETRAN_EXIT); } /* pick up the net */ net = chk->whoTo; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; } if ((asoc->peers_rwnd < mtu) && (asoc->total_flight > 0)) { /* No room in peers rwnd */ uint32_t tsn; tsn = asoc->last_acked_seq + 1; if (tsn == chk->rec.data.TSN_seq) { /* * we make a special exception for this * case. The peer has no rwnd but is missing * the lowest chunk.. which is probably what * is holding up the rwnd. */ goto one_chunk_around; } return (1); } one_chunk_around: if (asoc->peers_rwnd < mtu) { one_chunk = 1; if ((asoc->peers_rwnd == 0) && (asoc->total_flight == 0)) { chk->window_probe = 1; chk->whoTo->window_probe = 1; } } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC3, 2); #endif bundle_at = 0; m = NULL; net->fast_retran_ip = 0; if (chk->rec.data.doing_fast_retransmit == 0) { /* * if no FR in progress skip destination that have * flight_size > cwnd. */ if (net->flight_size >= net->cwnd) { continue; } } else { /* * Mark the destination net to have FR recovery * limits put on it. */ *fr_done = 1; net->fast_retran_ip = 1; } /* * if no AUTH is yet included and this chunk requires it, * make sure to account for it. We don't apply the size * until the AUTH chunk is actually added below in case * there is no room for this chunk. */ if (data_auth_reqd && (auth == NULL)) { dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else dmtu = 0; if ((chk->send_size <= (mtu - dmtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* ok we will add this one */ if (data_auth_reqd) { if (auth == NULL) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, SCTP_DATA); auth_keyid = chk->auth_keyid; override_ok = 0; SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } else if (override_ok) { auth_keyid = chk->auth_keyid; override_ok = 0; } else if (chk->auth_keyid != auth_keyid) { /* different keyid, so done bundling */ break; } } m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); if (m == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* upate our MTU size */ if (mtu > (chk->send_size + dmtu)) mtu -= (chk->send_size + dmtu); else mtu = 0; data_list[bundle_at++] = chk; if (one_chunk && (asoc->total_flight <= 0)) { SCTP_STAT_INCR(sctps_windowprobed); } } if (one_chunk == 0) { /* * now are there anymore forward from chk to pick * up? */ fwd = TAILQ_NEXT(chk, sctp_next); while (fwd) { if (fwd->sent != SCTP_DATAGRAM_RESEND) { /* Nope, not for retran */ fwd = TAILQ_NEXT(fwd, sctp_next); continue; } if (fwd->whoTo != net) { /* Nope, not the net in question */ fwd = TAILQ_NEXT(fwd, sctp_next); continue; } if (data_auth_reqd && (auth == NULL)) { dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else dmtu = 0; if (fwd->send_size <= (mtu - dmtu)) { if (data_auth_reqd) { if (auth == NULL) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, SCTP_DATA); auth_keyid = fwd->auth_keyid; override_ok = 0; SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } else if (override_ok) { auth_keyid = fwd->auth_keyid; override_ok = 0; } else if (fwd->auth_keyid != auth_keyid) { /* * different keyid, * so done bundling */ break; } } m = sctp_copy_mbufchain(fwd->data, m, &endofchain, 0, fwd->send_size, fwd->copy_by_ref); if (m == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* Do clear IP_DF ? */ if (fwd->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* upate our MTU size */ if (mtu > (fwd->send_size + dmtu)) mtu -= (fwd->send_size + dmtu); else mtu = 0; data_list[bundle_at++] = fwd; if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { break; } fwd = TAILQ_NEXT(fwd, sctp_next); } else { /* can't fit so we are done */ break; } } } /* Is there something to send for this destination? */ if (m) { /* * No matter if we fail/or suceed we should start a * timer. A failure is like a lost IP packet :-) */ if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* * no timer running on this destination * restart it. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); tmr_started = 1; } /* Now lets send it, if there is anything to send :> */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, m, auth_offset, auth, auth_keyid, no_fragmentflg, 0, NULL, 0, inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), net->port, so_locked, NULL))) { /* error, we could not output */ SCTP_STAT_INCR(sctps_lowlevelerr); return (error); } m = endofchain = NULL; auth = NULL; auth_offset = 0; /* For HB's */ /* * We don't want to mark the net->sent time here * since this we use this for HB and retrans cannot * measure RTT */ /* (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); */ /* For auto-close */ cnt_thru++; if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); *now = asoc->time_last_sent; *now_filled = 1; } else { asoc->time_last_sent = *now; } *cnt_out += bundle_at; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC4, bundle_at); #endif if (bundle_at) { tsns_sent = data_list[0]->rec.data.TSN_seq; } for (i = 0; i < bundle_at; i++) { SCTP_STAT_INCR(sctps_sendretransdata); data_list[i]->sent = SCTP_DATAGRAM_SENT; /* * When we have a revoked data, and we * retransmit it, then we clear the revoked * flag since this flag dictates if we * subtracted from the fs */ if (data_list[i]->rec.data.chunk_was_revoked) { /* Deflate the cwnd */ data_list[i]->whoTo->cwnd -= data_list[i]->book_size; data_list[i]->rec.data.chunk_was_revoked = 0; } data_list[i]->snd_count++; sctp_ucount_decr(asoc->sent_queue_retran_cnt); /* record the time */ data_list[i]->sent_rcv_time = asoc->time_last_sent; if (data_list[i]->book_size_scale) { /* * need to double the book size on * this one */ data_list[i]->book_size_scale = 0; /* * Since we double the booksize, we * must also double the output queue * size, since this get shrunk when * we free by this amount. */ atomic_add_int(&((asoc)->total_output_queue_size), data_list[i]->book_size); data_list[i]->book_size *= 2; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); } asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, (uint32_t) (data_list[i]->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP_RSND, data_list[i]->whoTo->flight_size, data_list[i]->book_size, (uintptr_t) data_list[i]->whoTo, data_list[i]->rec.data.TSN_seq); } sctp_flight_size_increase(data_list[i]); sctp_total_flight_increase(stcb, data_list[i]); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } if ((i == 0) && (data_list[i]->rec.data.doing_fast_retransmit)) { SCTP_STAT_INCR(sctps_sendfastretrans); if ((data_list[i] == TAILQ_FIRST(&asoc->sent_queue)) && (tmr_started == 0)) { /*- * ok we just fast-retrans'd * the lowest TSN, i.e the * first on the list. In * this case we want to give * some more time to get a * SACK back without a * t3-expiring. */ sctp_timer_stop(SCTP_TIMER_TYPE_SEND, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_4); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_RESEND); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(21, inp, stcb, NULL); #endif } else { /* None will fit */ return (1); } if (asoc->sent_queue_retran_cnt <= 0) { /* all done we have no more to retran */ asoc->sent_queue_retran_cnt = 0; break; } if (one_chunk) { /* No more room in rwnd */ return (1); } /* stop the for loop here. we sent out a packet */ break; } return (0); } static int sctp_timer_validation(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int ret) { struct sctp_nets *net; /* Validate that a timer is running somewhere */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* Here is a timer */ return (ret); } } SCTP_TCB_LOCK_ASSERT(stcb); /* Gak, we did not have a timer somewhere */ SCTPDBG(SCTP_DEBUG_OUTPUT3, "Deadlock avoided starting timer on a dest at retran\n"); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, asoc->primary_destination); return (ret); } void sctp_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_where, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /*- * Ok this is the generic chunk service queue. we must do the * following: * - See if there are retransmits pending, if so we must * do these first. * - Service the stream queue that is next, moving any * message (note I must get a complete message i.e. * FIRST/MIDDLE and LAST to the out queue in one pass) and assigning * TSN's * - Check to see if the cwnd/rwnd allows any output, if so we * go ahead and fomulate and send the low level chunks. Making sure * to combine any control in the control chunk queue also. */ struct sctp_association *asoc; struct sctp_nets *net; int error = 0, num_out = 0, tot_out = 0, ret = 0, reason_code = 0, burst_cnt = 0; struct timeval now; int now_filled = 0; int nagle_on = 0; int frag_point = sctp_get_frag_point(stcb, &stcb->asoc); int un_sent = 0; int fr_done, tot_frs = 0; asoc = &stcb->asoc; if (from_where == SCTP_OUTPUT_FROM_USR_SEND) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY)) { nagle_on = 0; } else { nagle_on = 1; } } SCTP_TCB_LOCK_ASSERT(stcb); un_sent = (stcb->asoc.total_output_queue_size - stcb->asoc.total_flight); if ((un_sent <= 0) && (TAILQ_EMPTY(&asoc->control_send_queue)) && (TAILQ_EMPTY(&asoc->asconf_send_queue)) && (asoc->sent_queue_retran_cnt == 0)) { /* Nothing to do unless there is something to be sent left */ return; } /* * Do we have something to send, data or control AND a sack timer * running, if so piggy-back the sack. */ if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_send_sack(stcb); (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); } while (asoc->sent_queue_retran_cnt) { /*- * Ok, it is retransmission time only, we send out only ONE * packet with a single call off to the retran code. */ if (from_where == SCTP_OUTPUT_FROM_COOKIE_ACK) { /*- * Special hook for handling cookiess discarded * by peer that carried data. Send cookie-ack only * and then the next call with get the retran's. */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, from_where, &now, &now_filled, frag_point, so_locked); return; } else if (from_where != SCTP_OUTPUT_FROM_HB_TMR) { /* if its not from a HB then do it */ fr_done = 0; ret = sctp_chunk_retransmission(inp, stcb, asoc, &num_out, &now, &now_filled, &fr_done, so_locked); if (fr_done) { tot_frs++; } } else { /* * its from any other place, we don't allow retran * output (only control) */ ret = 1; } if (ret > 0) { /* Can't send anymore */ /*- * now lets push out control by calling med-level * output once. this assures that we WILL send HB's * if queued too. */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, from_where, &now, &now_filled, frag_point, so_locked); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(8, inp, stcb, NULL); #endif (void)sctp_timer_validation(inp, stcb, asoc, ret); return; } if (ret < 0) { /*- * The count was off.. retran is not happening so do * the normal retransmission. */ #ifdef SCTP_AUDITING_ENABLED sctp_auditing(9, inp, stcb, NULL); #endif if (ret == SCTP_RETRAN_EXIT) { return; } break; } if (from_where == SCTP_OUTPUT_FROM_T3) { /* Only one transmission allowed out of a timeout */ #ifdef SCTP_AUDITING_ENABLED sctp_auditing(10, inp, stcb, NULL); #endif /* Push out any control */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, from_where, &now, &now_filled, frag_point, so_locked); return; } if ((asoc->max_burst > 0) && (tot_frs > asoc->max_burst)) { /* Hit FR burst limit */ return; } if ((num_out == 0) && (ret == 0)) { /* No more retrans to send */ break; } } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(12, inp, stcb, NULL); #endif /* Check for bad destinations, if they exist move chunks around. */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) == SCTP_ADDR_NOT_REACHABLE) { /*- * if possible move things off of this address we * still may send below due to the dormant state but * we try to find an alternate address to send to * and if we have one we move all queued data on the * out wheel to this alternate address. */ if (net->ref_count > 1) sctp_move_chunks_from_net(stcb, net); } else if ((asoc->sctp_cmt_on_off > 0) && (asoc->sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { /* * JRS 5/14/07 - If CMT PF is on and the current * destination is in PF state, move all queued data * to an alternate desination. */ if (net->ref_count > 1) sctp_move_chunks_from_net(stcb, net); } else { /*- * if ((asoc->sat_network) || (net->addr_is_local)) * { burst_limit = asoc->max_burst * * SCTP_SAT_NETWORK_BURST_INCR; } */ if (asoc->max_burst > 0) { if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst)) { if ((net->flight_size + (asoc->max_burst * net->mtu)) < net->cwnd) { /* * JRS - Use the congestion * control given in the * congestion control module */ asoc->cc_functions.sctp_cwnd_update_after_output(stcb, net, asoc->max_burst); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, net, 0, asoc->max_burst, SCTP_MAX_BURST_APPLIED); } SCTP_STAT_INCR(sctps_maxburstqueued); } net->fast_retran_ip = 0; } else { if (net->flight_size == 0) { /* * Should be decaying the * cwnd here */ ; } } } } } burst_cnt = 0; do { error = sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 0, from_where, &now, &now_filled, frag_point, so_locked); if (error) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Error %d was returned from med-c-op\n", error); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, asoc->primary_destination, error, burst_cnt, SCTP_MAX_BURST_ERROR_STOP); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, error, SCTP_SEND_NOW_COMPLETES); sctp_log_cwnd(stcb, NULL, 0xdeadbeef, SCTP_SEND_NOW_COMPLETES); } break; } SCTPDBG(SCTP_DEBUG_OUTPUT3, "m-c-o put out %d\n", num_out); tot_out += num_out; burst_cnt++; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, num_out, SCTP_SEND_NOW_COMPLETES); if (num_out == 0) { sctp_log_cwnd(stcb, NULL, reason_code, SCTP_SEND_NOW_COMPLETES); } } if (nagle_on) { /*- * When nagle is on, we look at how much is un_sent, then * if its smaller than an MTU and we have data in * flight we stop. */ un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) && (stcb->asoc.total_flight > 0)) { break; } } if (TAILQ_EMPTY(&asoc->control_send_queue) && TAILQ_EMPTY(&asoc->send_queue) && stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { /* Nothing left to send */ break; } if ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) <= 0) { /* Nothing left to send */ break; } } while (num_out && ((asoc->max_burst == 0) || SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) || (burst_cnt < asoc->max_burst))); if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) == 0) { if ((asoc->max_burst > 0) && (burst_cnt >= asoc->max_burst)) { SCTP_STAT_INCR(sctps_maxburstqueued); asoc->burst_limit_applied = 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, asoc->primary_destination, 0, burst_cnt, SCTP_MAX_BURST_APPLIED); } } else { asoc->burst_limit_applied = 0; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, tot_out, SCTP_SEND_NOW_COMPLETES); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, we have put out %d chunks\n", tot_out); /*- * Now we need to clean up the control chunk chain if a ECNE is on * it. It must be marked as UNSENT again so next call will continue * to send it until such time that we get a CWR, to remove it. */ if (stcb->asoc.ecn_echo_cnt_onq) sctp_fix_ecn_echo(asoc); return; } int sctp_output(inp, m, addr, control, p, flags) struct sctp_inpcb *inp; struct mbuf *m; struct sockaddr *addr; struct mbuf *control; struct thread *p; int flags; { if (inp == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (inp->sctp_socket == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } return (sctp_sosend(inp->sctp_socket, addr, (struct uio *)NULL, m, control, flags, p )); } void send_forward_tsn(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; struct sctp_forward_tsn_chunk *fwdtsn; uint32_t advance_peer_ack_point; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { /* mark it to unsent */ chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; /* Do we correct its output location? */ if (chk->whoTo != asoc->primary_destination) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); } goto sctp_fill_in_rest; } } /* Ok if we reach here we must build one */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } asoc->fwd_tsn_cnt++; chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_FORWARD_CUM_TSN; chk->rec.chunk_id.can_take_data = 0; chk->asoc = asoc; chk->whoTo = NULL; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; sctp_fill_in_rest: /*- * Here we go through and fill out the part that deals with * stream/seq of the ones we skip. */ SCTP_BUF_LEN(chk->data) = 0; { struct sctp_tmit_chunk *at, *tp1, *last; struct sctp_strseq *strseq; unsigned int cnt_of_space, i, ovh; unsigned int space_needed; unsigned int cnt_of_skipped = 0; TAILQ_FOREACH(at, &asoc->sent_queue, sctp_next) { if (at->sent != SCTP_FORWARD_TSN_SKIP) { /* no more to look at */ break; } if (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED) { /* We don't report these */ continue; } cnt_of_skipped++; } space_needed = (sizeof(struct sctp_forward_tsn_chunk) + (cnt_of_skipped * sizeof(struct sctp_strseq))); cnt_of_space = M_TRAILINGSPACE(chk->data); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MIN_OVERHEAD; } else { ovh = SCTP_MIN_V4_OVERHEAD; } if (cnt_of_space > (asoc->smallest_mtu - ovh)) { /* trim to a mtu size */ cnt_of_space = asoc->smallest_mtu - ovh; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { sctp_misc_ints(SCTP_FWD_TSN_CHECK, 0xff, 0, cnt_of_skipped, asoc->advanced_peer_ack_point); } advance_peer_ack_point = asoc->advanced_peer_ack_point; if (cnt_of_space < space_needed) { /*- * ok we must trim down the chunk by lowering the * advance peer ack point. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { sctp_misc_ints(SCTP_FWD_TSN_CHECK, 0xff, 0xff, cnt_of_space, space_needed); } cnt_of_skipped = cnt_of_space - sizeof(struct sctp_forward_tsn_chunk); cnt_of_skipped /= sizeof(struct sctp_strseq); /*- * Go through and find the TSN that will be the one * we report. */ at = TAILQ_FIRST(&asoc->sent_queue); for (i = 0; i < cnt_of_skipped; i++) { tp1 = TAILQ_NEXT(at, sctp_next); if (tp1 == NULL) { break; } at = tp1; } if (at && SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { sctp_misc_ints(SCTP_FWD_TSN_CHECK, 0xff, cnt_of_skipped, at->rec.data.TSN_seq, asoc->advanced_peer_ack_point); } last = at; /*- * last now points to last one I can report, update * peer ack point */ if (last) advance_peer_ack_point = last->rec.data.TSN_seq; space_needed = sizeof(struct sctp_forward_tsn_chunk) + cnt_of_skipped * sizeof(struct sctp_strseq); } chk->send_size = space_needed; /* Setup the chunk */ fwdtsn = mtod(chk->data, struct sctp_forward_tsn_chunk *); fwdtsn->ch.chunk_length = htons(chk->send_size); fwdtsn->ch.chunk_flags = 0; fwdtsn->ch.chunk_type = SCTP_FORWARD_CUM_TSN; fwdtsn->new_cumulative_tsn = htonl(advance_peer_ack_point); SCTP_BUF_LEN(chk->data) = chk->send_size; fwdtsn++; /*- * Move pointer to after the fwdtsn and transfer to the * strseq pointer. */ strseq = (struct sctp_strseq *)fwdtsn; /*- * Now populate the strseq list. This is done blindly * without pulling out duplicate stream info. This is * inefficent but won't harm the process since the peer will * look at these in sequence and will thus release anything. * It could mean we exceed the PMTU and chop off some that * we could have included.. but this is unlikely (aka 1432/4 * would mean 300+ stream seq's would have to be reported in * one FWD-TSN. With a bit of work we can later FIX this to * optimize and pull out duplcates.. but it does add more * overhead. So for now... not! */ at = TAILQ_FIRST(&asoc->sent_queue); for (i = 0; i < cnt_of_skipped; i++) { tp1 = TAILQ_NEXT(at, sctp_next); if (tp1 == NULL) break; if (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED) { /* We don't report these */ i--; at = tp1; continue; } if (at->rec.data.TSN_seq == advance_peer_ack_point) { at->rec.data.fwd_tsn_cnt = 0; } strseq->stream = ntohs(at->rec.data.stream_number); strseq->sequence = ntohs(at->rec.data.stream_seq); strseq++; at = tp1; } } return; } void sctp_send_sack(struct sctp_tcb *stcb) { /*- * Queue up a SACK or NR-SACK in the control queue. * We must first check to see if a SACK or NR-SACK is * somehow on the control queue. * If so, we will take and and remove the old one. */ struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *a_chk; struct sctp_sack_chunk *sack; struct sctp_nr_sack_chunk *nr_sack; struct sctp_gap_ack_block *gap_descriptor; struct sack_track *selector; int mergeable = 0; int offset; caddr_t limit; uint32_t *dup; int limit_reached = 0; unsigned int i, siz, j; unsigned int num_gap_blocks = 0, num_nr_gap_blocks = 0, space; int num_dups = 0; int space_req; uint32_t highest_tsn; uint8_t flags; uint8_t type; uint8_t tsn_map; if ((stcb->asoc.sctp_nr_sack_on_off == 1) && (stcb->asoc.peer_supports_nr_sack == 1)) { type = SCTP_NR_SELECTIVE_ACK; } else { type = SCTP_SELECTIVE_ACK; } a_chk = NULL; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->last_data_chunk_from == NULL) { /* Hmm we never received anything */ return; } sctp_slide_mapping_arrays(stcb); sctp_set_rwnd(stcb, asoc); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == type) { /* Hmm, found a sack already on queue, remove it */ TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt--; a_chk = chk; if (a_chk->data) { sctp_m_freem(a_chk->data); a_chk->data = NULL; } sctp_free_remote_addr(a_chk->whoTo); a_chk->whoTo = NULL; break; } } if (a_chk == NULL) { sctp_alloc_a_chunk(stcb, a_chk); if (a_chk == NULL) { /* No memory so we drop the idea, and set a timer */ if (stcb->asoc.delayed_ack) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_5); sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } else { stcb->asoc.send_sack = 1; } return; } a_chk->copy_by_ref = 0; a_chk->rec.chunk_id.id = type; a_chk->rec.chunk_id.can_take_data = 1; } /* Clear our pkt counts */ asoc->data_pkts_seen = 0; a_chk->asoc = asoc; a_chk->snd_count = 0; a_chk->send_size = 0; /* fill in later */ a_chk->sent = SCTP_DATAGRAM_UNSENT; a_chk->whoTo = NULL; if ((asoc->numduptsns) || (asoc->last_data_chunk_from->dest_state & SCTP_ADDR_NOT_REACHABLE)) { /*- * Ok, we have some duplicates or the destination for the * sack is unreachable, lets see if we can select an * alternate than asoc->last_data_chunk_from */ if ((!(asoc->last_data_chunk_from->dest_state & SCTP_ADDR_NOT_REACHABLE)) && (asoc->used_alt_onsack > asoc->numnets)) { /* We used an alt last time, don't this time */ a_chk->whoTo = NULL; } else { asoc->used_alt_onsack++; a_chk->whoTo = sctp_find_alternate_net(stcb, asoc->last_data_chunk_from, 0); } if (a_chk->whoTo == NULL) { /* Nope, no alternate */ a_chk->whoTo = asoc->last_data_chunk_from; asoc->used_alt_onsack = 0; } } else { /* * No duplicates so we use the last place we received data * from. */ asoc->used_alt_onsack = 0; a_chk->whoTo = asoc->last_data_chunk_from; } if (a_chk->whoTo) { atomic_add_int(&a_chk->whoTo->ref_count, 1); } if (SCTP_TSN_GT(asoc->highest_tsn_inside_map, asoc->highest_tsn_inside_nr_map)) { highest_tsn = asoc->highest_tsn_inside_map; } else { highest_tsn = asoc->highest_tsn_inside_nr_map; } if (highest_tsn == asoc->cumulative_tsn) { /* no gaps */ if (type == SCTP_SELECTIVE_ACK) { space_req = sizeof(struct sctp_sack_chunk); } else { space_req = sizeof(struct sctp_nr_sack_chunk); } } else { /* gaps get a cluster */ space_req = MCLBYTES; } /* Ok now lets formulate a MBUF with our sack */ a_chk->data = sctp_get_mbuf_for_msg(space_req, 0, M_DONTWAIT, 1, MT_DATA); if ((a_chk->data == NULL) || (a_chk->whoTo == NULL)) { /* rats, no mbuf memory */ if (a_chk->data) { /* was a problem with the destination */ sctp_m_freem(a_chk->data); a_chk->data = NULL; } sctp_free_a_chunk(stcb, a_chk); /* sa_ignore NO_NULL_CHK */ if (stcb->asoc.delayed_ack) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_6); sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } else { stcb->asoc.send_sack = 1; } return; } /* ok, lets go through and fill it in */ SCTP_BUF_RESV_UF(a_chk->data, SCTP_MIN_OVERHEAD); space = M_TRAILINGSPACE(a_chk->data); if (space > (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD)) { space = (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD); } limit = mtod(a_chk->data, caddr_t); limit += space; - /* 0x01 is used by nonce for ecn */ - if ((SCTP_BASE_SYSCTL(sctp_ecn_enable)) && - (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) && - (asoc->peer_supports_ecn_nonce)) - flags = (asoc->receiver_nonce_sum & SCTP_SACK_NONCE_SUM); - else - flags = 0; + flags = 0; if ((asoc->sctp_cmt_on_off > 0) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { /*- * CMT DAC algorithm: If 2 (i.e., 0x10) packets have been * received, then set high bit to 1, else 0. Reset * pkts_rcvd. */ flags |= (asoc->cmt_dac_pkts_rcvd << 6); asoc->cmt_dac_pkts_rcvd = 0; } #ifdef SCTP_ASOCLOG_OF_TSNS stcb->asoc.cumack_logsnt[stcb->asoc.cumack_log_atsnt] = asoc->cumulative_tsn; stcb->asoc.cumack_log_atsnt++; if (stcb->asoc.cumack_log_atsnt >= SCTP_TSN_LOG_SIZE) { stcb->asoc.cumack_log_atsnt = 0; } #endif /* reset the readers interpretation */ stcb->freed_by_sorcv_sincelast = 0; if (type == SCTP_SELECTIVE_ACK) { sack = mtod(a_chk->data, struct sctp_sack_chunk *); nr_sack = NULL; gap_descriptor = (struct sctp_gap_ack_block *)((caddr_t)sack + sizeof(struct sctp_sack_chunk)); if (highest_tsn > asoc->mapping_array_base_tsn) { siz = (((highest_tsn - asoc->mapping_array_base_tsn) + 1) + 7) / 8; } else { siz = (((MAX_TSN - highest_tsn) + 1) + highest_tsn + 7) / 8; } } else { sack = NULL; nr_sack = mtod(a_chk->data, struct sctp_nr_sack_chunk *); gap_descriptor = (struct sctp_gap_ack_block *)((caddr_t)nr_sack + sizeof(struct sctp_nr_sack_chunk)); if (asoc->highest_tsn_inside_map > asoc->mapping_array_base_tsn) { siz = (((asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn) + 1) + 7) / 8; } else { siz = (((MAX_TSN - asoc->mapping_array_base_tsn) + 1) + asoc->highest_tsn_inside_map + 7) / 8; } } if (SCTP_TSN_GT(asoc->mapping_array_base_tsn, asoc->cumulative_tsn)) { offset = 1; } else { offset = asoc->mapping_array_base_tsn - asoc->cumulative_tsn; } if (((type == SCTP_SELECTIVE_ACK) && SCTP_TSN_GT(highest_tsn, asoc->cumulative_tsn)) || ((type == SCTP_NR_SELECTIVE_ACK) && SCTP_TSN_GT(asoc->highest_tsn_inside_map, asoc->cumulative_tsn))) { /* we have a gap .. maybe */ for (i = 0; i < siz; i++) { tsn_map = asoc->mapping_array[i]; if (type == SCTP_SELECTIVE_ACK) { tsn_map |= asoc->nr_mapping_array[i]; } if (i == 0) { /* * Clear all bits corresponding to TSNs * smaller or equal to the cumulative TSN. */ tsn_map &= (~0 << (1 - offset)); } selector = &sack_array[tsn_map]; if (mergeable && selector->right_edge) { /* * Backup, left and right edges were ok to * merge. */ num_gap_blocks--; gap_descriptor--; } if (selector->num_entries == 0) mergeable = 0; else { for (j = 0; j < selector->num_entries; j++) { if (mergeable && selector->right_edge) { /* * do a merge by NOT setting * the left side */ mergeable = 0; } else { /* * no merge, set the left * side */ mergeable = 0; gap_descriptor->start = htons((selector->gaps[j].start + offset)); } gap_descriptor->end = htons((selector->gaps[j].end + offset)); num_gap_blocks++; gap_descriptor++; if (((caddr_t)gap_descriptor + sizeof(struct sctp_gap_ack_block)) > limit) { /* no more room */ limit_reached = 1; break; } } if (selector->left_edge) { mergeable = 1; } } if (limit_reached) { /* Reached the limit stop */ break; } offset += 8; } } if ((type == SCTP_NR_SELECTIVE_ACK) && (limit_reached == 0)) { mergeable = 0; if (asoc->highest_tsn_inside_nr_map > asoc->mapping_array_base_tsn) { siz = (((asoc->highest_tsn_inside_nr_map - asoc->mapping_array_base_tsn) + 1) + 7) / 8; } else { siz = (((MAX_TSN - asoc->mapping_array_base_tsn) + 1) + asoc->highest_tsn_inside_nr_map + 7) / 8; } if (SCTP_TSN_GT(asoc->mapping_array_base_tsn, asoc->cumulative_tsn)) { offset = 1; } else { offset = asoc->mapping_array_base_tsn - asoc->cumulative_tsn; } if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->cumulative_tsn)) { /* we have a gap .. maybe */ for (i = 0; i < siz; i++) { tsn_map = asoc->nr_mapping_array[i]; if (i == 0) { /* * Clear all bits corresponding to * TSNs smaller or equal to the * cumulative TSN. */ tsn_map &= (~0 << (1 - offset)); } selector = &sack_array[tsn_map]; if (mergeable && selector->right_edge) { /* * Backup, left and right edges were * ok to merge. */ num_nr_gap_blocks--; gap_descriptor--; } if (selector->num_entries == 0) mergeable = 0; else { for (j = 0; j < selector->num_entries; j++) { if (mergeable && selector->right_edge) { /* * do a merge by NOT * setting the left * side */ mergeable = 0; } else { /* * no merge, set the * left side */ mergeable = 0; gap_descriptor->start = htons((selector->gaps[j].start + offset)); } gap_descriptor->end = htons((selector->gaps[j].end + offset)); num_nr_gap_blocks++; gap_descriptor++; if (((caddr_t)gap_descriptor + sizeof(struct sctp_gap_ack_block)) > limit) { /* no more room */ limit_reached = 1; break; } } if (selector->left_edge) { mergeable = 1; } } if (limit_reached) { /* Reached the limit stop */ break; } offset += 8; } } } /* now we must add any dups we are going to report. */ if ((limit_reached == 0) && (asoc->numduptsns)) { dup = (uint32_t *) gap_descriptor; for (i = 0; i < asoc->numduptsns; i++) { *dup = htonl(asoc->dup_tsns[i]); dup++; num_dups++; if (((caddr_t)dup + sizeof(uint32_t)) > limit) { /* no more room */ break; } } asoc->numduptsns = 0; } /* * now that the chunk is prepared queue it to the control chunk * queue. */ if (type == SCTP_SELECTIVE_ACK) { a_chk->send_size = sizeof(struct sctp_sack_chunk) + (num_gap_blocks + num_nr_gap_blocks) * sizeof(struct sctp_gap_ack_block) + num_dups * sizeof(int32_t); SCTP_BUF_LEN(a_chk->data) = a_chk->send_size; sack->sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); sack->sack.a_rwnd = htonl(asoc->my_rwnd); sack->sack.num_gap_ack_blks = htons(num_gap_blocks); sack->sack.num_dup_tsns = htons(num_dups); sack->ch.chunk_type = type; sack->ch.chunk_flags = flags; sack->ch.chunk_length = htons(a_chk->send_size); } else { a_chk->send_size = sizeof(struct sctp_nr_sack_chunk) + (num_gap_blocks + num_nr_gap_blocks) * sizeof(struct sctp_gap_ack_block) + num_dups * sizeof(int32_t); SCTP_BUF_LEN(a_chk->data) = a_chk->send_size; nr_sack->nr_sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); nr_sack->nr_sack.a_rwnd = htonl(asoc->my_rwnd); nr_sack->nr_sack.num_gap_ack_blks = htons(num_gap_blocks); nr_sack->nr_sack.num_nr_gap_ack_blks = htons(num_nr_gap_blocks); nr_sack->nr_sack.num_dup_tsns = htons(num_dups); nr_sack->nr_sack.reserved = 0; nr_sack->ch.chunk_type = type; nr_sack->ch.chunk_flags = flags; nr_sack->ch.chunk_length = htons(a_chk->send_size); } TAILQ_INSERT_TAIL(&asoc->control_send_queue, a_chk, sctp_next); asoc->my_last_reported_rwnd = asoc->my_rwnd; asoc->ctrl_queue_cnt++; asoc->send_sack = 0; SCTP_STAT_INCR(sctps_sendsacks); return; } void sctp_send_abort_tcb(struct sctp_tcb *stcb, struct mbuf *operr, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_abort; struct mbuf *m_out = NULL, *m_end = NULL; struct sctp_abort_chunk *abort = NULL; int sz; uint32_t auth_offset = 0; struct sctp_auth_chunk *auth = NULL; /*- * Add an AUTH chunk, if chunk requires it and save the offset into * the chain for AUTH */ if (sctp_auth_is_required_chunk(SCTP_ABORT_ASSOCIATION, stcb->asoc.peer_auth_chunks)) { m_out = sctp_add_auth_chunk(m_out, &m_end, &auth, &auth_offset, stcb, SCTP_ABORT_ASSOCIATION); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } SCTP_TCB_LOCK_ASSERT(stcb); m_abort = sctp_get_mbuf_for_msg(sizeof(struct sctp_abort_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_abort == NULL) { /* no mbuf's */ if (m_out) sctp_m_freem(m_out); return; } /* link in any error */ SCTP_BUF_NEXT(m_abort) = operr; sz = 0; if (operr) { struct mbuf *n; n = operr; while (n) { sz += SCTP_BUF_LEN(n); n = SCTP_BUF_NEXT(n); } } SCTP_BUF_LEN(m_abort) = sizeof(*abort); if (m_out == NULL) { /* NO Auth chunk prepended, so reserve space in front */ SCTP_BUF_RESV_UF(m_abort, SCTP_MIN_OVERHEAD); m_out = m_abort; } else { /* Put AUTH chunk at the front of the chain */ SCTP_BUF_NEXT(m_end) = m_abort; } /* fill in the ABORT chunk */ abort = mtod(m_abort, struct sctp_abort_chunk *); abort->ch.chunk_type = SCTP_ABORT_ASSOCIATION; abort->ch.chunk_flags = 0; abort->ch.chunk_length = htons(sizeof(*abort) + sz); (void)sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, stcb->asoc.primary_destination, (struct sockaddr *)&stcb->asoc.primary_destination->ro._l_addr, m_out, auth_offset, auth, stcb->asoc.authinfo.active_keyid, 1, 0, NULL, 0, stcb->sctp_ep->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), stcb->asoc.primary_destination->port, so_locked, NULL); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } void sctp_send_shutdown_complete(struct sctp_tcb *stcb, struct sctp_nets *net, int reflect_vtag) { /* formulate and SEND a SHUTDOWN-COMPLETE */ struct mbuf *m_shutdown_comp; struct sctp_shutdown_complete_chunk *shutdown_complete; uint32_t vtag; uint8_t flags; m_shutdown_comp = sctp_get_mbuf_for_msg(sizeof(struct sctp_chunkhdr), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown_comp == NULL) { /* no mbuf's */ return; } if (reflect_vtag) { flags = SCTP_HAD_NO_TCB; vtag = stcb->asoc.my_vtag; } else { flags = 0; vtag = stcb->asoc.peer_vtag; } shutdown_complete = mtod(m_shutdown_comp, struct sctp_shutdown_complete_chunk *); shutdown_complete->ch.chunk_type = SCTP_SHUTDOWN_COMPLETE; shutdown_complete->ch.chunk_flags = flags; shutdown_complete->ch.chunk_length = htons(sizeof(struct sctp_shutdown_complete_chunk)); SCTP_BUF_LEN(m_shutdown_comp) = sizeof(struct sctp_shutdown_complete_chunk); (void)sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, net, (struct sockaddr *)&net->ro._l_addr, m_shutdown_comp, 0, NULL, 0, 1, 0, NULL, 0, stcb->sctp_ep->sctp_lport, stcb->rport, htonl(vtag), net->port, SCTP_SO_NOT_LOCKED, NULL); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); return; } void sctp_send_shutdown_complete2(struct mbuf *m, int iphlen, struct sctphdr *sh, uint32_t vrf_id, uint16_t port) { /* formulate and SEND a SHUTDOWN-COMPLETE */ struct mbuf *o_pak; struct mbuf *mout; struct ip *iph, *iph_out; struct udphdr *udp = NULL; #ifdef INET6 struct ip6_hdr *ip6, *ip6_out; #endif int offset_out, len, mlen; struct sctp_shutdown_complete_msg *comp_cp; iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: len = (sizeof(struct ip) + sizeof(struct sctp_shutdown_complete_msg)); break; #ifdef INET6 case IPV6_VERSION >> 4: len = (sizeof(struct ip6_hdr) + sizeof(struct sctp_shutdown_complete_msg)); break; #endif default: return; } if (port) { len += sizeof(struct udphdr); } mout = sctp_get_mbuf_for_msg(len + max_linkhdr, 1, M_DONTWAIT, 1, MT_DATA); if (mout == NULL) { return; } SCTP_BUF_RESV_UF(mout, max_linkhdr); SCTP_BUF_LEN(mout) = len; SCTP_BUF_NEXT(mout) = NULL; iph_out = NULL; #ifdef INET6 ip6_out = NULL; #endif offset_out = 0; switch (iph->ip_v) { case IPVERSION: iph_out = mtod(mout, struct ip *); /* Fill in the IP header for the ABORT */ iph_out->ip_v = IPVERSION; iph_out->ip_hl = (sizeof(struct ip) / 4); iph_out->ip_tos = (u_char)0; iph_out->ip_id = 0; iph_out->ip_off = 0; iph_out->ip_ttl = MAXTTL; if (port) { iph_out->ip_p = IPPROTO_UDP; } else { iph_out->ip_p = IPPROTO_SCTP; } iph_out->ip_src.s_addr = iph->ip_dst.s_addr; iph_out->ip_dst.s_addr = iph->ip_src.s_addr; /* let IP layer calculate this */ iph_out->ip_sum = 0; offset_out += sizeof(*iph_out); comp_cp = (struct sctp_shutdown_complete_msg *)( (caddr_t)iph_out + offset_out); break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = (struct ip6_hdr *)iph; ip6_out = mtod(mout, struct ip6_hdr *); /* Fill in the IPv6 header for the ABORT */ ip6_out->ip6_flow = ip6->ip6_flow; ip6_out->ip6_hlim = MODULE_GLOBAL(ip6_defhlim); if (port) { ip6_out->ip6_nxt = IPPROTO_UDP; } else { ip6_out->ip6_nxt = IPPROTO_SCTP; } ip6_out->ip6_src = ip6->ip6_dst; ip6_out->ip6_dst = ip6->ip6_src; /* * ?? The old code had both the iph len + payload, I think * this is wrong and would never have worked */ ip6_out->ip6_plen = sizeof(struct sctp_shutdown_complete_msg); offset_out += sizeof(*ip6_out); comp_cp = (struct sctp_shutdown_complete_msg *)( (caddr_t)ip6_out + offset_out); break; #endif /* INET6 */ default: /* Currently not supported. */ sctp_m_freem(mout); return; } if (port) { udp = (struct udphdr *)comp_cp; udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(sizeof(struct sctp_shutdown_complete_msg) + sizeof(struct udphdr)); if (iph_out) udp->uh_sum = in_pseudo(iph_out->ip_src.s_addr, iph_out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); offset_out += sizeof(struct udphdr); comp_cp = (struct sctp_shutdown_complete_msg *)((caddr_t)comp_cp + sizeof(struct udphdr)); } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* no mbuf's */ sctp_m_freem(mout); return; } /* Now copy in and fill in the ABORT tags etc. */ comp_cp->sh.src_port = sh->dest_port; comp_cp->sh.dest_port = sh->src_port; comp_cp->sh.checksum = 0; comp_cp->sh.v_tag = sh->v_tag; comp_cp->shut_cmp.ch.chunk_flags = SCTP_HAD_NO_TCB; comp_cp->shut_cmp.ch.chunk_type = SCTP_SHUTDOWN_COMPLETE; comp_cp->shut_cmp.ch.chunk_length = htons(sizeof(struct sctp_shutdown_complete_chunk)); if (iph_out != NULL) { sctp_route_t ro; int ret; mlen = SCTP_BUF_LEN(mout); bzero(&ro, sizeof ro); /* set IPv4 length */ iph_out->ip_len = mlen; #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, mlen); #endif if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else comp_cp->sh.checksum = sctp_calculate_cksum(mout, offset_out); SCTP_STAT_INCR(sctps_sendswcrc); #endif SCTP_ENABLE_UDP_CSUM(mout); } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_ATTACH_CHAIN(o_pak, mout, mlen); /* out it goes */ SCTP_IP_OUTPUT(ret, o_pak, &ro, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #ifdef INET6 if (ip6_out != NULL) { struct route_in6 ro; int ret; struct ifnet *ifp = NULL; bzero(&ro, sizeof(ro)); mlen = SCTP_BUF_LEN(mout); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, mlen); #endif SCTP_ATTACH_CHAIN(o_pak, mout, mlen); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else comp_cp->sh.checksum = sctp_calculate_cksum(mout, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); SCTP_STAT_INCR(sctps_sendswcrc); #endif if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), mlen - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); return; } static struct sctp_nets * sctp_select_hb_destination(struct sctp_tcb *stcb, struct timeval *now) { struct sctp_nets *net, *hnet; int ms_goneby, highest_ms, state_overide = 0; (void)SCTP_GETTIME_TIMEVAL(now); highest_ms = 0; hnet = NULL; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ( ((net->dest_state & SCTP_ADDR_NOHB) && ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) || (net->dest_state & SCTP_ADDR_OUT_OF_SCOPE) ) { /* * Skip this guy from consideration if HB is off AND * its confirmed */ continue; } if (sctp_destination_is_reachable(stcb, (struct sockaddr *)&net->ro._l_addr) == 0) { /* skip this dest net from consideration */ continue; } if (net->last_sent_time.tv_sec) { /* Sent to so we subtract */ ms_goneby = (now->tv_sec - net->last_sent_time.tv_sec) * 1000; } else /* Never been sent to */ ms_goneby = 0x7fffffff; /*- * When the address state is unconfirmed but still * considered reachable, we HB at a higher rate. Once it * goes confirmed OR reaches the "unreachable" state, thenw * we cut it back to HB at a more normal pace. */ if ((net->dest_state & (SCTP_ADDR_UNCONFIRMED | SCTP_ADDR_NOT_REACHABLE)) == SCTP_ADDR_UNCONFIRMED) { state_overide = 1; } else { state_overide = 0; } if ((((unsigned int)ms_goneby >= net->RTO) || (state_overide)) && (ms_goneby > highest_ms)) { highest_ms = ms_goneby; hnet = net; } } if (hnet && ((hnet->dest_state & (SCTP_ADDR_UNCONFIRMED | SCTP_ADDR_NOT_REACHABLE)) == SCTP_ADDR_UNCONFIRMED)) { state_overide = 1; } else { state_overide = 0; } if (hnet && highest_ms && (((unsigned int)highest_ms >= hnet->RTO) || state_overide)) { /*- * Found the one with longest delay bounds OR it is * unconfirmed and still not marked unreachable. */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "net:%p is the hb winner -", hnet); #ifdef SCTP_DEBUG if (hnet) { SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT4, (struct sockaddr *)&hnet->ro._l_addr); } else { SCTPDBG(SCTP_DEBUG_OUTPUT4, " none\n"); } #endif /* update the timer now */ hnet->last_sent_time = *now; return (hnet); } /* Nothing to HB */ return (NULL); } int sctp_send_hb(struct sctp_tcb *stcb, int user_req, struct sctp_nets *u_net) { struct sctp_tmit_chunk *chk; struct sctp_nets *net; struct sctp_heartbeat_chunk *hb; struct timeval now; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; SCTP_TCB_LOCK_ASSERT(stcb); if (user_req == 0) { net = sctp_select_hb_destination(stcb, &now); if (net == NULL) { /*- * All our busy none to send to, just start the * timer again. */ if (stcb->asoc.state == 0) { return (0); } sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); return (0); } } else { net = u_net; if (net == NULL) { return (0); } (void)SCTP_GETTIME_TIMEVAL(&now); } sin = (struct sockaddr_in *)&net->ro._l_addr; if (sin->sin_family != AF_INET) { if (sin->sin_family != AF_INET6) { /* huh */ return (0); } } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT4, "Gak, can't get a chunk for hb\n"); return (0); } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_HEARTBEAT_REQUEST; chk->rec.chunk_id.can_take_data = 1; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_heartbeat_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return (0); } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); /* Now we have a mbuf that we can fill in with the details */ hb = mtod(chk->data, struct sctp_heartbeat_chunk *); memset(hb, 0, sizeof(struct sctp_heartbeat_chunk)); /* fill out chunk header */ hb->ch.chunk_type = SCTP_HEARTBEAT_REQUEST; hb->ch.chunk_flags = 0; hb->ch.chunk_length = htons(chk->send_size); /* Fill out hb parameter */ hb->heartbeat.hb_info.ph.param_type = htons(SCTP_HEARTBEAT_INFO); hb->heartbeat.hb_info.ph.param_length = htons(sizeof(struct sctp_heartbeat_info_param)); hb->heartbeat.hb_info.time_value_1 = now.tv_sec; hb->heartbeat.hb_info.time_value_2 = now.tv_usec; /* Did our user request this one, put it in */ hb->heartbeat.hb_info.user_req = user_req; hb->heartbeat.hb_info.addr_family = sin->sin_family; hb->heartbeat.hb_info.addr_len = sin->sin_len; if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* * we only take from the entropy pool if the address is not * confirmed. */ net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); } else { net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = 0; net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = 0; } if (sin->sin_family == AF_INET) { memcpy(hb->heartbeat.hb_info.address, &sin->sin_addr, sizeof(sin->sin_addr)); } else if (sin->sin_family == AF_INET6) { /* We leave the scope the way it is in our lookup table. */ sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; memcpy(hb->heartbeat.hb_info.address, &sin6->sin6_addr, sizeof(sin6->sin6_addr)); } else { /* huh compiler bug */ return (0); } /* * JRS 5/14/07 - In CMT PF, the T3 timer is used to track * PF-heartbeats. Because of this, threshold management is done by * the t3 timer handler, and does not need to be done upon the send * of a PF-heartbeat. If CMT PF is on and the destination to which a * heartbeat is being sent is in PF state, do NOT do threshold * management. */ if ((stcb->asoc.sctp_cmt_pf == 0) || ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF)) { /* ok we have a destination that needs a beat */ /* lets do the theshold management Qiaobing style */ if (sctp_threshold_management(stcb->sctp_ep, stcb, net, stcb->asoc.max_send_times)) { /*- * we have lost the association, in a way this is * quite bad since we really are one less time since * we really did not send yet. This is the down side * to the Q's style as defined in the RFC and not my * alternate style defined in the RFC. */ if (chk->data != NULL) { sctp_m_freem(chk->data); chk->data = NULL; } /* * Here we do NOT use the macro since the * association is now gone. */ if (chk->whoTo) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = NULL; } sctp_free_a_chunk((struct sctp_tcb *)NULL, chk); return (-1); } } net->hb_responded = 0; TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); stcb->asoc.ctrl_queue_cnt++; SCTP_STAT_INCR(sctps_sendheartbeat); /*- * Call directly med level routine to put out the chunk. It will * always tumble out control chunks aka HB but it may even tumble * out data too. */ return (1); } void sctp_send_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t high_tsn) { struct sctp_association *asoc; struct sctp_ecne_chunk *ecne; struct sctp_tmit_chunk *chk; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if ((chk->rec.chunk_id.id == SCTP_ECN_ECHO) && (net == chk->whoTo)) { /* found a previous ECN_ECHO update it if needed */ uint32_t cnt, ctsn; ecne = mtod(chk->data, struct sctp_ecne_chunk *); ctsn = ntohl(ecne->tsn); if (SCTP_TSN_GT(high_tsn, ctsn)) { ecne->tsn = htonl(high_tsn); - cnt = ntohl(ecne->num_pkts_since_cwr); - cnt++; - ecne->num_pkts_since_cwr = htonl(cnt); SCTP_STAT_INCR(sctps_queue_upd_ecne); } + cnt = ntohl(ecne->num_pkts_since_cwr); + cnt++; + ecne->num_pkts_since_cwr = htonl(cnt); return; } } /* nope could not find one to update so we must build one */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; SCTP_STAT_INCR(sctps_queue_upd_ecne); chk->rec.chunk_id.id = SCTP_ECN_ECHO; chk->rec.chunk_id.can_take_data = 0; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_ecne_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); stcb->asoc.ecn_echo_cnt_onq++; ecne = mtod(chk->data, struct sctp_ecne_chunk *); ecne->ch.chunk_type = SCTP_ECN_ECHO; ecne->ch.chunk_flags = 0; ecne->ch.chunk_length = htons(sizeof(struct sctp_ecne_chunk)); ecne->tsn = htonl(high_tsn); ecne->num_pkts_since_cwr = htonl(1); TAILQ_INSERT_HEAD(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_send_packet_dropped(struct sctp_tcb *stcb, struct sctp_nets *net, struct mbuf *m, int iphlen, int bad_crc) { struct sctp_association *asoc; struct sctp_pktdrop_chunk *drp; struct sctp_tmit_chunk *chk; uint8_t *datap; int len; int was_trunc = 0; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6h; #endif int fullsz = 0, extra = 0; long spc; int offset; struct sctp_chunkhdr *ch, chunk_buf; unsigned int chk_length; if (!stcb) { return; } asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->peer_supports_pktdrop == 0) { /*- * peer must declare support before I send one. */ return; } if (stcb->sctp_socket == NULL) { return; } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; iph = mtod(m, struct ip *); if (iph == NULL) { sctp_free_a_chunk(stcb, chk); return; } switch (iph->ip_v) { case IPVERSION: /* IPv4 */ len = chk->send_size = iph->ip_len; break; #ifdef INET6 case IPV6_VERSION >> 4: /* IPv6 */ ip6h = mtod(m, struct ip6_hdr *); len = chk->send_size = htons(ip6h->ip6_plen); break; #endif default: return; } /* Validate that we do not have an ABORT in here. */ offset = iphlen + sizeof(struct sctphdr); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* break to abort land */ break; } switch (ch->chunk_type) { case SCTP_PACKET_DROPPED: case SCTP_ABORT_ASSOCIATION: case SCTP_INITIATION_ACK: /** * We don't respond with an PKT-DROP to an ABORT * or PKT-DROP. We also do not respond to an * INIT-ACK, because we can't know if the initiation * tag is correct or not. */ sctp_free_a_chunk(stcb, chk); return; default: break; } offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } if ((len + SCTP_MAX_OVERHEAD + sizeof(struct sctp_pktdrop_chunk)) > min(stcb->asoc.smallest_mtu, MCLBYTES)) { /* * only send 1 mtu worth, trim off the excess on the end. */ fullsz = len - extra; len = min(stcb->asoc.smallest_mtu, MCLBYTES) - SCTP_MAX_OVERHEAD; was_trunc = 1; } chk->asoc = &stcb->asoc; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { jump_out: sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); drp = mtod(chk->data, struct sctp_pktdrop_chunk *); if (drp == NULL) { sctp_m_freem(chk->data); chk->data = NULL; goto jump_out; } chk->book_size = SCTP_SIZE32((chk->send_size + sizeof(struct sctp_pktdrop_chunk) + sizeof(struct sctphdr) + SCTP_MED_OVERHEAD)); chk->book_size_scale = 0; if (was_trunc) { drp->ch.chunk_flags = SCTP_PACKET_TRUNCATED; drp->trunc_len = htons(fullsz); /* * Len is already adjusted to size minus overhead above take * out the pkt_drop chunk itself from it. */ chk->send_size = len - sizeof(struct sctp_pktdrop_chunk); len = chk->send_size; } else { /* no truncation needed */ drp->ch.chunk_flags = 0; drp->trunc_len = htons(0); } if (bad_crc) { drp->ch.chunk_flags |= SCTP_BADCRC; } chk->send_size += sizeof(struct sctp_pktdrop_chunk); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; if (net) { /* we should hit here */ chk->whoTo = net; } else { chk->whoTo = asoc->primary_destination; } atomic_add_int(&chk->whoTo->ref_count, 1); chk->rec.chunk_id.id = SCTP_PACKET_DROPPED; chk->rec.chunk_id.can_take_data = 1; drp->ch.chunk_type = SCTP_PACKET_DROPPED; drp->ch.chunk_length = htons(chk->send_size); spc = SCTP_SB_LIMIT_RCV(stcb->sctp_socket); if (spc < 0) { spc = 0; } drp->bottle_bw = htonl(spc); if (asoc->my_rwnd) { drp->current_onq = htonl(asoc->size_on_reasm_queue + asoc->size_on_all_streams + asoc->my_rwnd_control_len + stcb->sctp_socket->so_rcv.sb_cc); } else { /*- * If my rwnd is 0, possibly from mbuf depletion as well as * space used, tell the peer there is NO space aka onq == bw */ drp->current_onq = htonl(spc); } drp->reserved = 0; datap = drp->data; m_copydata(m, iphlen, len, (caddr_t)datap); TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_send_cwr(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t high_tsn, uint8_t override) { struct sctp_association *asoc; struct sctp_cwr_chunk *cwr; struct sctp_tmit_chunk *chk; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_ECN_CWR; chk->rec.chunk_id.can_take_data = 1; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_cwr_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); cwr = mtod(chk->data, struct sctp_cwr_chunk *); cwr->ch.chunk_type = SCTP_ECN_CWR; cwr->ch.chunk_flags = override; cwr->ch.chunk_length = htons(sizeof(struct sctp_cwr_chunk)); cwr->tsn = htonl(high_tsn); TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_add_stream_reset_out(struct sctp_tmit_chunk *chk, int number_entries, uint16_t * list, uint32_t seq, uint32_t resp_seq, uint32_t last_sent) { int len, old_len, i; struct sctp_stream_reset_out_request *req_out; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_out = (struct sctp_stream_reset_out_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = (sizeof(struct sctp_stream_reset_out_request) + (sizeof(uint16_t) * number_entries)); req_out->ph.param_type = htons(SCTP_STR_RESET_OUT_REQUEST); req_out->ph.param_length = htons(len); req_out->request_seq = htonl(seq); req_out->response_seq = htonl(resp_seq); req_out->send_reset_at_tsn = htonl(last_sent); if (number_entries) { for (i = 0; i < number_entries; i++) { req_out->list_of_streams[i] = htons(list[i]); } } if (SCTP_SIZE32(len) > len) { /*- * Need to worry about the pad we may end up adding to the * end. This is easy since the struct is either aligned to 4 * bytes or 2 bytes off. */ req_out->list_of_streams[number_entries] = 0; } /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_in(struct sctp_tmit_chunk *chk, int number_entries, uint16_t * list, uint32_t seq) { int len, old_len, i; struct sctp_stream_reset_in_request *req_in; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_in = (struct sctp_stream_reset_in_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = (sizeof(struct sctp_stream_reset_in_request) + (sizeof(uint16_t) * number_entries)); req_in->ph.param_type = htons(SCTP_STR_RESET_IN_REQUEST); req_in->ph.param_length = htons(len); req_in->request_seq = htonl(seq); if (number_entries) { for (i = 0; i < number_entries; i++) { req_in->list_of_streams[i] = htons(list[i]); } } if (SCTP_SIZE32(len) > len) { /*- * Need to worry about the pad we may end up adding to the * end. This is easy since the struct is either aligned to 4 * bytes or 2 bytes off. */ req_in->list_of_streams[number_entries] = 0; } /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_tsn(struct sctp_tmit_chunk *chk, uint32_t seq) { int len, old_len; struct sctp_stream_reset_tsn_request *req_tsn; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_tsn = (struct sctp_stream_reset_tsn_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_tsn_request); req_tsn->ph.param_type = htons(SCTP_STR_RESET_TSN_REQUEST); req_tsn->ph.param_length = htons(len); req_tsn->request_seq = htonl(seq); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->send_size = len + old_len; chk->book_size = SCTP_SIZE32(chk->send_size); chk->book_size_scale = 0; SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); return; } void sctp_add_stream_reset_result(struct sctp_tmit_chunk *chk, uint32_t resp_seq, uint32_t result) { int len, old_len; struct sctp_stream_reset_response *resp; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ resp = (struct sctp_stream_reset_response *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_response); resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); resp->ph.param_length = htons(len); resp->response_seq = htonl(resp_seq); resp->result = ntohl(result); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_result_tsn(struct sctp_tmit_chunk *chk, uint32_t resp_seq, uint32_t result, uint32_t send_una, uint32_t recv_next) { int len, old_len; struct sctp_stream_reset_response_tsn *resp; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ resp = (struct sctp_stream_reset_response_tsn *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_response_tsn); resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); resp->ph.param_length = htons(len); resp->response_seq = htonl(resp_seq); resp->result = htonl(result); resp->senders_next_tsn = htonl(send_una); resp->receivers_next_tsn = htonl(recv_next); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->send_size = SCTP_SIZE32(chk->book_size); chk->book_size_scale = 0; SCTP_BUF_LEN(chk->data) = chk->send_size; return; } static void sctp_add_a_stream(struct sctp_tmit_chunk *chk, uint32_t seq, uint16_t adding) { int len, old_len; struct sctp_chunkhdr *ch; struct sctp_stream_reset_add_strm *addstr; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ addstr = (struct sctp_stream_reset_add_strm *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_add_strm); /* Fill it out. */ addstr->ph.param_type = htons(SCTP_STR_RESET_ADD_STREAMS); addstr->ph.param_length = htons(len); addstr->request_seq = htonl(seq); addstr->number_of_streams = htons(adding); addstr->reserved = 0; /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->send_size = len + old_len; chk->book_size = SCTP_SIZE32(chk->send_size); chk->book_size_scale = 0; SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); return; } int sctp_send_str_reset_req(struct sctp_tcb *stcb, int number_entries, uint16_t * list, uint8_t send_out_req, uint32_t resp_seq, uint8_t send_in_req, uint8_t send_tsn_req, uint8_t add_stream, uint16_t adding ) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk; struct sctp_chunkhdr *ch; uint32_t seq; asoc = &stcb->asoc; if (asoc->stream_reset_outstanding) { /*- * Already one pending, must get ACK back to clear the flag. */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EBUSY); return (EBUSY); } if ((send_out_req == 0) && (send_in_req == 0) && (send_tsn_req == 0) && (add_stream == 0)) { /* nothing to do */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (send_tsn_req && (send_out_req || send_in_req)) { /* error, can't do that */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_STREAM_RESET; chk->rec.chunk_id.can_take_data = 0; chk->asoc = &stcb->asoc; chk->book_size = sizeof(struct sctp_chunkhdr); chk->send_size = SCTP_SIZE32(chk->book_size); chk->book_size_scale = 0; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); /* setup chunk parameters */ chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); ch = mtod(chk->data, struct sctp_chunkhdr *); ch->chunk_type = SCTP_STREAM_RESET; ch->chunk_flags = 0; ch->chunk_length = htons(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; seq = stcb->asoc.str_reset_seq_out; if (send_out_req) { sctp_add_stream_reset_out(chk, number_entries, list, seq, resp_seq, (stcb->asoc.sending_seq - 1)); asoc->stream_reset_out_is_outstanding = 1; seq++; asoc->stream_reset_outstanding++; } if (add_stream) { sctp_add_a_stream(chk, seq, adding); seq++; asoc->stream_reset_outstanding++; } if (send_in_req) { sctp_add_stream_reset_in(chk, number_entries, list, seq); asoc->stream_reset_outstanding++; } if (send_tsn_req) { sctp_add_stream_reset_tsn(chk, seq); asoc->stream_reset_outstanding++; } asoc->str_reset = chk; /* insert the chunk for sending */ TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo); return (0); } void sctp_send_abort(struct mbuf *m, int iphlen, struct sctphdr *sh, uint32_t vtag, struct mbuf *err_cause, uint32_t vrf_id, uint16_t port) { /*- * Formulate the abort message, and send it back down. */ struct mbuf *o_pak; struct mbuf *mout; struct sctp_abort_msg *abm; struct ip *iph, *iph_out; struct udphdr *udp; #ifdef INET6 struct ip6_hdr *ip6, *ip6_out; #endif int iphlen_out, len; /* don't respond to ABORT with ABORT */ if (sctp_is_there_an_abort_here(m, iphlen, &vtag)) { if (err_cause) sctp_m_freem(err_cause); return; } iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: len = (sizeof(struct ip) + sizeof(struct sctp_abort_msg)); break; #ifdef INET6 case IPV6_VERSION >> 4: len = (sizeof(struct ip6_hdr) + sizeof(struct sctp_abort_msg)); break; #endif default: if (err_cause) { sctp_m_freem(err_cause); } return; } if (port) { len += sizeof(struct udphdr); } mout = sctp_get_mbuf_for_msg(len + max_linkhdr, 1, M_DONTWAIT, 1, MT_DATA); if (mout == NULL) { if (err_cause) { sctp_m_freem(err_cause); } return; } SCTP_BUF_RESV_UF(mout, max_linkhdr); SCTP_BUF_LEN(mout) = len; SCTP_BUF_NEXT(mout) = err_cause; iph_out = NULL; #ifdef INET6 ip6_out = NULL; #endif switch (iph->ip_v) { case IPVERSION: iph_out = mtod(mout, struct ip *); /* Fill in the IP header for the ABORT */ iph_out->ip_v = IPVERSION; iph_out->ip_hl = (sizeof(struct ip) / 4); iph_out->ip_tos = (u_char)0; iph_out->ip_id = 0; iph_out->ip_off = 0; iph_out->ip_ttl = MAXTTL; if (port) { iph_out->ip_p = IPPROTO_UDP; } else { iph_out->ip_p = IPPROTO_SCTP; } iph_out->ip_src.s_addr = iph->ip_dst.s_addr; iph_out->ip_dst.s_addr = iph->ip_src.s_addr; /* let IP layer calculate this */ iph_out->ip_sum = 0; iphlen_out = sizeof(*iph_out); abm = (struct sctp_abort_msg *)((caddr_t)iph_out + iphlen_out); break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = (struct ip6_hdr *)iph; ip6_out = mtod(mout, struct ip6_hdr *); /* Fill in the IP6 header for the ABORT */ ip6_out->ip6_flow = ip6->ip6_flow; ip6_out->ip6_hlim = MODULE_GLOBAL(ip6_defhlim); if (port) { ip6_out->ip6_nxt = IPPROTO_UDP; } else { ip6_out->ip6_nxt = IPPROTO_SCTP; } ip6_out->ip6_src = ip6->ip6_dst; ip6_out->ip6_dst = ip6->ip6_src; iphlen_out = sizeof(*ip6_out); abm = (struct sctp_abort_msg *)((caddr_t)ip6_out + iphlen_out); break; #endif /* INET6 */ default: /* Currently not supported */ sctp_m_freem(mout); return; } udp = (struct udphdr *)abm; if (port) { udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; /* set udp->uh_ulen later */ udp->uh_sum = 0; iphlen_out += sizeof(struct udphdr); abm = (struct sctp_abort_msg *)((caddr_t)abm + sizeof(struct udphdr)); } abm->sh.src_port = sh->dest_port; abm->sh.dest_port = sh->src_port; abm->sh.checksum = 0; if (vtag == 0) { abm->sh.v_tag = sh->v_tag; abm->msg.ch.chunk_flags = SCTP_HAD_NO_TCB; } else { abm->sh.v_tag = htonl(vtag); abm->msg.ch.chunk_flags = 0; } abm->msg.ch.chunk_type = SCTP_ABORT_ASSOCIATION; if (err_cause) { struct mbuf *m_tmp = err_cause; int err_len = 0; /* get length of the err_cause chain */ while (m_tmp != NULL) { err_len += SCTP_BUF_LEN(m_tmp); m_tmp = SCTP_BUF_NEXT(m_tmp); } len = SCTP_BUF_LEN(mout) + err_len; if (err_len % 4) { /* need pad at end of chunk */ uint32_t cpthis = 0; int padlen; padlen = 4 - (len % 4); m_copyback(mout, len, padlen, (caddr_t)&cpthis); len += padlen; } abm->msg.ch.chunk_length = htons(sizeof(abm->msg.ch) + err_len); } else { len = SCTP_BUF_LEN(mout); abm->msg.ch.chunk_length = htons(sizeof(abm->msg.ch)); } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* no mbuf's */ sctp_m_freem(mout); return; } if (iph_out != NULL) { sctp_route_t ro; int ret; /* zap the stack pointer to the route */ bzero(&ro, sizeof ro); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip)); udp->uh_sum = in_pseudo(iph_out->ip_src.s_addr, iph_out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "sctp_send_abort calling ip_output:\n"); SCTPDBG_PKT(SCTP_DEBUG_OUTPUT2, iph_out, &abm->sh); /* set IPv4 length */ iph_out->ip_len = len; /* out it goes */ #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else abm->sh.checksum = sctp_calculate_cksum(mout, iphlen_out); SCTP_STAT_INCR(sctps_sendswcrc); #endif SCTP_ENABLE_UDP_CSUM(o_pak); } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_IP_OUTPUT(ret, o_pak, &ro, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #ifdef INET6 if (ip6_out != NULL) { struct route_in6 ro; int ret; struct ifnet *ifp = NULL; /* zap the stack pointer to the route */ bzero(&ro, sizeof(ro)); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip6_hdr)); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "sctp_send_abort calling ip6_output:\n"); SCTPDBG_PKT(SCTP_DEBUG_OUTPUT2, (struct ip *)ip6_out, &abm->sh); ip6_out->ip6_plen = len - sizeof(*ip6_out); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else abm->sh.checksum = sctp_calculate_cksum(mout, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); SCTP_STAT_INCR(sctps_sendswcrc); #endif if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), len - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } void sctp_send_operr_to(struct mbuf *m, int iphlen, struct mbuf *scm, uint32_t vtag, uint32_t vrf_id, uint16_t port) { struct mbuf *o_pak; struct sctphdr *sh, *sh_out; struct sctp_chunkhdr *ch; struct ip *iph, *iph_out; struct udphdr *udp = NULL; struct mbuf *mout; #ifdef INET6 struct ip6_hdr *ip6, *ip6_out; #endif int iphlen_out, len; iph = mtod(m, struct ip *); sh = (struct sctphdr *)((caddr_t)iph + iphlen); switch (iph->ip_v) { case IPVERSION: len = (sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr)); break; #ifdef INET6 case IPV6_VERSION >> 4: len = (sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr)); break; #endif default: if (scm) { sctp_m_freem(scm); } return; } if (port) { len += sizeof(struct udphdr); } mout = sctp_get_mbuf_for_msg(len + max_linkhdr, 1, M_DONTWAIT, 1, MT_DATA); if (mout == NULL) { if (scm) { sctp_m_freem(scm); } return; } SCTP_BUF_RESV_UF(mout, max_linkhdr); SCTP_BUF_LEN(mout) = len; SCTP_BUF_NEXT(mout) = scm; iph_out = NULL; #ifdef INET6 ip6_out = NULL; #endif switch (iph->ip_v) { case IPVERSION: iph_out = mtod(mout, struct ip *); /* Fill in the IP header for the ABORT */ iph_out->ip_v = IPVERSION; iph_out->ip_hl = (sizeof(struct ip) / 4); iph_out->ip_tos = (u_char)0; iph_out->ip_id = 0; iph_out->ip_off = 0; iph_out->ip_ttl = MAXTTL; if (port) { iph_out->ip_p = IPPROTO_UDP; } else { iph_out->ip_p = IPPROTO_SCTP; } iph_out->ip_src.s_addr = iph->ip_dst.s_addr; iph_out->ip_dst.s_addr = iph->ip_src.s_addr; /* let IP layer calculate this */ iph_out->ip_sum = 0; iphlen_out = sizeof(struct ip); sh_out = (struct sctphdr *)((caddr_t)iph_out + iphlen_out); break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = (struct ip6_hdr *)iph; ip6_out = mtod(mout, struct ip6_hdr *); /* Fill in the IP6 header for the ABORT */ ip6_out->ip6_flow = ip6->ip6_flow; ip6_out->ip6_hlim = MODULE_GLOBAL(ip6_defhlim); if (port) { ip6_out->ip6_nxt = IPPROTO_UDP; } else { ip6_out->ip6_nxt = IPPROTO_SCTP; } ip6_out->ip6_src = ip6->ip6_dst; ip6_out->ip6_dst = ip6->ip6_src; iphlen_out = sizeof(struct ip6_hdr); sh_out = (struct sctphdr *)((caddr_t)ip6_out + iphlen_out); break; #endif /* INET6 */ default: /* Currently not supported */ sctp_m_freem(mout); return; } udp = (struct udphdr *)sh_out; if (port) { udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; /* set udp->uh_ulen later */ udp->uh_sum = 0; iphlen_out += sizeof(struct udphdr); sh_out = (struct sctphdr *)((caddr_t)udp + sizeof(struct udphdr)); } sh_out->src_port = sh->dest_port; sh_out->dest_port = sh->src_port; sh_out->v_tag = vtag; sh_out->checksum = 0; ch = (struct sctp_chunkhdr *)((caddr_t)sh_out + sizeof(struct sctphdr)); ch->chunk_type = SCTP_OPERATION_ERROR; ch->chunk_flags = 0; if (scm) { struct mbuf *m_tmp = scm; int cause_len = 0; /* get length of the err_cause chain */ while (m_tmp != NULL) { cause_len += SCTP_BUF_LEN(m_tmp); m_tmp = SCTP_BUF_NEXT(m_tmp); } len = SCTP_BUF_LEN(mout) + cause_len; if (cause_len % 4) { /* need pad at end of chunk */ uint32_t cpthis = 0; int padlen; padlen = 4 - (len % 4); m_copyback(mout, len, padlen, (caddr_t)&cpthis); len += padlen; } ch->chunk_length = htons(sizeof(struct sctp_chunkhdr) + cause_len); } else { len = SCTP_BUF_LEN(mout); ch->chunk_length = htons(sizeof(struct sctp_chunkhdr)); } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* no mbuf's */ sctp_m_freem(mout); return; } if (iph_out != NULL) { sctp_route_t ro; int ret; /* zap the stack pointer to the route */ bzero(&ro, sizeof ro); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip)); udp->uh_sum = in_pseudo(iph_out->ip_src.s_addr, iph_out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); } /* set IPv4 length */ iph_out->ip_len = len; /* out it goes */ #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else sh_out->checksum = sctp_calculate_cksum(mout, iphlen_out); SCTP_STAT_INCR(sctps_sendswcrc); #endif SCTP_ENABLE_UDP_CSUM(o_pak); } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_IP_OUTPUT(ret, o_pak, &ro, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #ifdef INET6 if (ip6_out != NULL) { struct route_in6 ro; int ret; struct ifnet *ifp = NULL; /* zap the stack pointer to the route */ bzero(&ro, sizeof(ro)); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip6_hdr)); } ip6_out->ip6_plen = len - sizeof(*ip6_out); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else sh_out->checksum = sctp_calculate_cksum(mout, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); SCTP_STAT_INCR(sctps_sendswcrc); #endif if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), len - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } else { #if defined(SCTP_WITH_NO_CSUM) SCTP_STAT_INCR(sctps_sendnocrc); #else mout->m_pkthdr.csum_flags = CSUM_SCTP; mout->m_pkthdr.csum_data = 0; SCTP_STAT_INCR(sctps_sendhwcrc); #endif } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, NULL, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } static struct mbuf * sctp_copy_resume(struct sctp_stream_queue_pending *sp, struct uio *uio, struct sctp_sndrcvinfo *srcv, int max_send_len, int user_marks_eor, int *error, uint32_t * sndout, struct mbuf **new_tail) { struct mbuf *m; m = m_uiotombuf(uio, M_WAITOK, max_send_len, 0, (M_PKTHDR | (user_marks_eor ? M_EOR : 0))); if (m == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); *error = ENOMEM; } else { *sndout = m_length(m, NULL); *new_tail = m_last(m); } return (m); } static int sctp_copy_one(struct sctp_stream_queue_pending *sp, struct uio *uio, int resv_upfront) { int left; left = sp->length; sp->data = m_uiotombuf(uio, M_WAITOK, sp->length, resv_upfront, 0); if (sp->data == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } sp->tail_mbuf = m_last(sp->data); return (0); } static struct sctp_stream_queue_pending * sctp_copy_it_in(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_sndrcvinfo *srcv, struct uio *uio, struct sctp_nets *net, int max_send_len, int user_marks_eor, int *error, int non_blocking) { /*- * This routine must be very careful in its work. Protocol * processing is up and running so care must be taken to spl...() * when you need to do something that may effect the stcb/asoc. The * sb is locked however. When data is copied the protocol processing * should be enabled since this is a slower operation... */ struct sctp_stream_queue_pending *sp = NULL; int resv_in_first; *error = 0; /* Now can we send this? */ if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { /* got data while shutting down */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); *error = ECONNRESET; goto out_now; } sctp_alloc_a_strmoq(stcb, sp); if (sp == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOMEM); *error = ENOMEM; goto out_now; } sp->act_flags = 0; sp->sender_all_done = 0; sp->sinfo_flags = srcv->sinfo_flags; sp->timetolive = srcv->sinfo_timetolive; sp->ppid = srcv->sinfo_ppid; sp->context = srcv->sinfo_context; sp->strseq = 0; (void)SCTP_GETTIME_TIMEVAL(&sp->ts); sp->stream = srcv->sinfo_stream; sp->length = min(uio->uio_resid, max_send_len); if ((sp->length == (uint32_t) uio->uio_resid) && ((user_marks_eor == 0) || (srcv->sinfo_flags & SCTP_EOF) || (user_marks_eor && (srcv->sinfo_flags & SCTP_EOR)))) { sp->msg_is_complete = 1; } else { sp->msg_is_complete = 0; } sp->sender_all_done = 0; sp->some_taken = 0; sp->put_last_out = 0; resv_in_first = sizeof(struct sctp_data_chunk); sp->data = sp->tail_mbuf = NULL; if (sp->length == 0) { *error = 0; goto skip_copy; } sp->auth_keyid = stcb->asoc.authinfo.active_keyid; if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { sctp_auth_key_acquire(stcb, stcb->asoc.authinfo.active_keyid); sp->holds_key_ref = 1; } *error = sctp_copy_one(sp, uio, resv_in_first); skip_copy: if (*error) { sctp_free_a_strmoq(stcb, sp); sp = NULL; } else { if (sp->sinfo_flags & SCTP_ADDR_OVER) { sp->net = net; atomic_add_int(&sp->net->ref_count, 1); } else { sp->net = NULL; } sctp_set_prsctp_policy(sp); } out_now: return (sp); } int sctp_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *p ) { int error, use_rcvinfo = 0; struct sctp_sndrcvinfo srcv; struct sockaddr *addr_to_use; #if defined(INET) && defined(INET6) struct sockaddr_in sin; #endif if (control) { /* process cmsg snd/rcv info (maybe a assoc-id) */ if (sctp_find_cmsg(SCTP_SNDRCV, (void *)&srcv, control, sizeof(srcv))) { /* got one */ use_rcvinfo = 1; } } addr_to_use = addr; #if defined(INET) && defined(INET6) if ((addr) && (addr->sa_family == AF_INET6)) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin, sin6); addr_to_use = (struct sockaddr *)&sin; } } #endif error = sctp_lower_sosend(so, addr_to_use, uio, top, control, flags, use_rcvinfo ? &srcv : NULL ,p ); return (error); } int sctp_lower_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *i_pak, struct mbuf *control, int flags, struct sctp_sndrcvinfo *srcv , struct thread *p ) { unsigned int sndlen = 0, max_len; int error, len; struct mbuf *top = NULL; int queue_only = 0, queue_only_for_init = 0; int free_cnt_applied = 0; int un_sent; int now_filled = 0; unsigned int inqueue_bytes = 0; struct sctp_block_entry be; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; struct timeval now; struct sctp_nets *net; struct sctp_association *asoc; struct sctp_inpcb *t_inp; int user_marks_eor; int create_lock_applied = 0; int nagle_applies = 0; int some_on_control = 0; int got_all_of_the_send = 0; int hold_tcblock = 0; int non_blocking = 0; uint32_t local_add_more, local_soresv = 0; uint16_t port; uint16_t sinfo_flags; sctp_assoc_t sinfo_assoc_id; error = 0; net = NULL; stcb = NULL; asoc = NULL; t_inp = inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; if (i_pak) { SCTP_RELEASE_PKT(i_pak); } return (error); } if ((uio == NULL) && (i_pak == NULL)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } user_marks_eor = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); atomic_add_int(&inp->total_sends, 1); if (uio) { if (uio->uio_resid < 0) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } sndlen = uio->uio_resid; } else { top = SCTP_HEADER_TO_CHAIN(i_pak); sndlen = SCTP_HEADER_LEN(i_pak); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "Send called addr:%p send length %d\n", addr, sndlen); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_socket->so_qlimit)) { /* The listener can NOT send */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } /** * Pre-screen address, if one is given the sin-len * must be set correctly! */ if (addr) { union sctp_sockstore *raddr = (union sctp_sockstore *)addr; switch (raddr->sa.sa_family) { #if defined(INET) case AF_INET: if (raddr->sin.sin_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } port = raddr->sin.sin_port; break; #endif #if defined(INET6) case AF_INET6: if (raddr->sin6.sin6_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } port = raddr->sin6.sin6_port; break; #endif default: SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EAFNOSUPPORT); error = EAFNOSUPPORT; goto out_unlocked; } } else port = 0; if (srcv) { sinfo_flags = srcv->sinfo_flags; sinfo_assoc_id = srcv->sinfo_assoc_id; if (INVALID_SINFO_FLAG(sinfo_flags) || PR_SCTP_INVALID_POLICY(sinfo_flags)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (srcv->sinfo_flags) SCTP_STAT_INCR(sctps_sends_with_flags); } else { sinfo_flags = inp->def_send.sinfo_flags; sinfo_assoc_id = inp->def_send.sinfo_assoc_id; } if (sinfo_flags & SCTP_SENDALL) { /* its a sendall */ error = sctp_sendall(inp, uio, top, srcv); top = NULL; goto out_unlocked; } if ((sinfo_flags & SCTP_ADDR_OVER) && (addr == NULL)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } /* now we must find the assoc */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } SCTP_TCB_LOCK(stcb); hold_tcblock = 1; SCTP_INP_RUNLOCK(inp); } else if (sinfo_assoc_id) { stcb = sctp_findassociation_ep_asocid(inp, sinfo_assoc_id, 0); } else if (addr) { /*- * Since we did not use findep we must * increment it, and if we don't find a tcb * decrement it. */ SCTP_INP_WLOCK(inp); SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } else { hold_tcblock = 1; } } if ((stcb == NULL) && (addr)) { /* Possible implicit send? */ SCTP_ASOC_CREATE_LOCK(inp); create_lock_applied = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* Should I really unlock ? */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (addr->sa_family == AF_INET6)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } SCTP_INP_WLOCK(inp); SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); /* With the lock applied look again */ stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } else { hold_tcblock = 1; } if (t_inp != inp) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } } if (stcb == NULL) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } if (addr == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOENT); error = ENOENT; goto out_unlocked; } else { /* * UDP style, we must go ahead and start the INIT * process */ uint32_t vrf_id; if ((sinfo_flags & SCTP_ABORT) || ((sinfo_flags & SCTP_EOF) && (sndlen == 0))) { /*- * User asks to abort a non-existant assoc, * or EOF a non-existant assoc with no data */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOENT); error = ENOENT; goto out_unlocked; } /* get an asoc/stcb struct */ vrf_id = inp->def_vrf_id; #ifdef INVARIANTS if (create_lock_applied == 0) { panic("Error, should hold create lock and I don't?"); } #endif stcb = sctp_aloc_assoc(inp, addr, &error, 0, vrf_id, p ); if (stcb == NULL) { /* Error is setup for us in the call */ goto out_unlocked; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } else { SCTP_PRINTF("Huh-3? create lock should have been on??\n"); } /* * Turn on queue only flag to prevent data from * being sent */ queue_only = 1; asoc = &stcb->asoc; SCTP_SET_STATE(asoc, SCTP_STATE_COOKIE_WAIT); (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); /* initialize authentication params for the assoc */ sctp_initialize_auth_params(inp, stcb); if (control) { /* * see if a init structure exists in cmsg * headers */ struct sctp_initmsg initm; int i; if (sctp_find_cmsg(SCTP_INIT, (void *)&initm, control, sizeof(initm))) { /* * we have an INIT override of the * default */ if (initm.sinit_max_attempts) asoc->max_init_times = initm.sinit_max_attempts; if (initm.sinit_num_ostreams) asoc->pre_open_streams = initm.sinit_num_ostreams; if (initm.sinit_max_instreams) asoc->max_inbound_streams = initm.sinit_max_instreams; if (initm.sinit_max_init_timeo) asoc->initial_init_rto_max = initm.sinit_max_init_timeo; if (asoc->streamoutcnt < asoc->pre_open_streams) { struct sctp_stream_out *tmp_str; int had_lock = 0; /* Default is NOT correct */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, defout:%d pre_open:%d\n", asoc->streamoutcnt, asoc->pre_open_streams); /* * What happens if this * fails? we panic ... */ if (hold_tcblock) { had_lock = 1; SCTP_TCB_UNLOCK(stcb); } SCTP_MALLOC(tmp_str, struct sctp_stream_out *, (asoc->pre_open_streams * sizeof(struct sctp_stream_out)), SCTP_M_STRMO); if (had_lock) { SCTP_TCB_LOCK(stcb); } if (tmp_str != NULL) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = tmp_str; asoc->strm_realoutsize = asoc->streamoutcnt = asoc->pre_open_streams; } else { asoc->pre_open_streams = asoc->streamoutcnt; } for (i = 0; i < asoc->streamoutcnt; i++) { /*- * inbound side must be set * to 0xffff, also NOTE when * we get the INIT-ACK back * (for INIT sender) we MUST * reduce the count * (streamoutcnt) but first * check if we sent to any * of the upper streams that * were dropped (if some * were). Those that were * dropped must be notified * to the upper layer as * failed to send. */ asoc->strmout[i].next_sequence_sent = 0x0; TAILQ_INIT(&asoc->strmout[i].outqueue); asoc->strmout[i].stream_no = i; asoc->strmout[i].last_msg_incomplete = 0; asoc->ss_functions.sctp_ss_init_stream(&asoc->strmout[i]); } } } } hold_tcblock = 1; /* out with the INIT */ queue_only_for_init = 1; /*- * we may want to dig in after this call and adjust the MTU * value. It defaulted to 1500 (constant) but the ro * structure may now have an update and thus we may need to * change it BEFORE we append the message. */ } } else asoc = &stcb->asoc; if (srcv == NULL) srcv = (struct sctp_sndrcvinfo *)&asoc->def_send; if (srcv->sinfo_flags & SCTP_ADDR_OVER) { if (addr) net = sctp_findnet(stcb, addr); else net = NULL; if ((net == NULL) || ((port != 0) && (port != stcb->rport))) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } } else { net = stcb->asoc.primary_destination; } atomic_add_int(&stcb->total_sends, 1); /* Keep the stcb from being freed under our feet */ atomic_add_int(&asoc->refcnt, 1); free_cnt_applied = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT)) { if (sndlen > asoc->smallest_mtu) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out_unlocked; } } if ((SCTP_SO_IS_NBIO(so) || (flags & MSG_NBIO) )) { non_blocking = 1; } /* would we block? */ if (non_blocking) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if ((SCTP_SB_LIMIT_SND(so) < (sndlen + inqueue_bytes + stcb->asoc.sb_send_resv)) || (stcb->asoc.chunks_on_out_queue >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EWOULDBLOCK); if (sndlen > SCTP_SB_LIMIT_SND(so)) error = EMSGSIZE; else error = EWOULDBLOCK; goto out_unlocked; } stcb->asoc.sb_send_resv += sndlen; SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } else { atomic_add_int(&stcb->asoc.sb_send_resv, sndlen); } local_soresv = sndlen; if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_unlocked; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } if (asoc->stream_reset_outstanding) { /* * Can't queue any data while stream reset is underway. */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EAGAIN); error = EAGAIN; goto out_unlocked; } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { queue_only = 1; } /* we are now done with all control */ if (control) { sctp_m_freem(control); control = NULL; } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { if (srcv->sinfo_flags & SCTP_ABORT) { ; } else { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_unlocked; } } /* Ok, we will attempt a msgsnd :> */ if (p) { p->td_ru.ru_msgsnd++; } /* Are we aborting? */ if (srcv->sinfo_flags & SCTP_ABORT) { struct mbuf *mm; int tot_demand, tot_out = 0, max_out; SCTP_STAT_INCR(sctps_sends_with_abort); if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { /* It has to be up before we abort */ /* how big is the user initiated abort? */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } if (top) { struct mbuf *cntm = NULL; mm = sctp_get_mbuf_for_msg(1, 0, M_WAIT, 1, MT_DATA); if (sndlen != 0) { cntm = top; while (cntm) { tot_out += SCTP_BUF_LEN(cntm); cntm = SCTP_BUF_NEXT(cntm); } } tot_demand = (tot_out + sizeof(struct sctp_paramhdr)); } else { /* Must fit in a MTU */ tot_out = sndlen; tot_demand = (tot_out + sizeof(struct sctp_paramhdr)); if (tot_demand > SCTP_DEFAULT_ADD_MORE) { /* To big */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out; } mm = sctp_get_mbuf_for_msg(tot_demand, 0, M_WAIT, 1, MT_DATA); } if (mm == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOMEM); error = ENOMEM; goto out; } max_out = asoc->smallest_mtu - sizeof(struct sctp_paramhdr); max_out -= sizeof(struct sctp_abort_msg); if (tot_out > max_out) { tot_out = max_out; } if (mm) { struct sctp_paramhdr *ph; /* now move forward the data pointer */ ph = mtod(mm, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons((sizeof(struct sctp_paramhdr) + tot_out)); ph++; SCTP_BUF_LEN(mm) = tot_out + sizeof(struct sctp_paramhdr); if (top == NULL) { error = uiomove((caddr_t)ph, (int)tot_out, uio); if (error) { /*- * Here if we can't get his data we * still abort we just don't get to * send the users note :-0 */ sctp_m_freem(mm); mm = NULL; } } else { if (sndlen != 0) { SCTP_BUF_NEXT(mm) = top; } } } if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } atomic_add_int(&stcb->asoc.refcnt, -1); free_cnt_applied = 0; /* release this lock, otherwise we hang on ourselves */ sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, mm, SCTP_SO_LOCKED); /* now relock the stcb so everything is sane */ hold_tcblock = 0; stcb = NULL; /* * In this case top is already chained to mm avoid double * free, since we free it below if top != NULL and driver * would free it after sending the packet out */ if (sndlen != 0) { top = NULL; } goto out_unlocked; } /* Calculate the maximum we can send */ inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { if (non_blocking) { /* we already checked for non-blocking above. */ max_len = sndlen; } else { max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; } } else { max_len = 0; } if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } /* Is the stream no. valid? */ if (srcv->sinfo_stream >= asoc->streamoutcnt) { /* Invalid stream number */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (asoc->strmout == NULL) { /* huh? software error */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); error = EFAULT; goto out_unlocked; } /* Unless E_EOR mode is on, we must make a send FIT in one call. */ if ((user_marks_eor == 0) && (sndlen > SCTP_SB_LIMIT_SND(stcb->sctp_socket))) { /* It will NEVER fit */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out_unlocked; } if ((uio == NULL) && user_marks_eor) { /*- * We do not support eeor mode for * sending with mbuf chains (like sendfile). */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (user_marks_eor) { local_add_more = min(SCTP_SB_LIMIT_SND(so), SCTP_BASE_SYSCTL(sctp_add_more_threshold)); } else { /*- * For non-eeor the whole message must fit in * the socket send buffer. */ local_add_more = sndlen; } len = 0; if (non_blocking) { goto skip_preblock; } if (((max_len <= local_add_more) && (SCTP_SB_LIMIT_SND(so) >= local_add_more)) || (max_len == 0) || ((stcb->asoc.chunks_on_out_queue + stcb->asoc.stream_queue_cnt) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { /* No room right now ! */ SOCKBUF_LOCK(&so->so_snd); inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); while ((SCTP_SB_LIMIT_SND(so) < (inqueue_bytes + local_add_more)) || ((stcb->asoc.stream_queue_cnt + stcb->asoc.chunks_on_out_queue) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "pre_block limit:%u <(inq:%d + %d) || (%d+%d > %d)\n", (unsigned int)SCTP_SB_LIMIT_SND(so), inqueue_bytes, local_add_more, stcb->asoc.stream_queue_cnt, stcb->asoc.chunks_on_out_queue, SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_INTO_BLKA, so, asoc, sndlen); } be.error = 0; stcb->block_entry = &be; error = sbwait(&so->so_snd); stcb->block_entry = NULL; if (error || so->so_error || be.error) { if (error == 0) { if (so->so_error) error = so->so_error; if (be.error) { error = be.error; } } SOCKBUF_UNLOCK(&so->so_snd); goto out_unlocked; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, so, asoc, stcb->asoc.total_output_queue_size); } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); } if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; } else { max_len = 0; } SOCKBUF_UNLOCK(&so->so_snd); } skip_preblock: if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } /* * sndlen covers for mbuf case uio_resid covers for the non-mbuf * case NOTE: uio will be null when top/mbuf is passed */ if (sndlen == 0) { if (srcv->sinfo_flags & SCTP_EOF) { got_all_of_the_send = 1; goto dataless_eof; } else { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } } if (top == NULL) { struct sctp_stream_queue_pending *sp; struct sctp_stream_out *strm; uint32_t sndout; SCTP_TCB_SEND_LOCK(stcb); if ((asoc->stream_locked) && (asoc->stream_locked_on != srcv->sinfo_stream)) { SCTP_TCB_SEND_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } SCTP_TCB_SEND_UNLOCK(stcb); strm = &stcb->asoc.strmout[srcv->sinfo_stream]; if (strm->last_msg_incomplete == 0) { do_a_copy_in: sp = sctp_copy_it_in(stcb, asoc, srcv, uio, net, max_len, user_marks_eor, &error, non_blocking); if ((sp == NULL) || (error)) { goto out; } SCTP_TCB_SEND_LOCK(stcb); if (sp->msg_is_complete) { strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } else { /* * Just got locked to this guy in case of an * interrupt. */ strm->last_msg_incomplete = 1; asoc->stream_locked = 1; asoc->stream_locked_on = srcv->sinfo_stream; sp->sender_all_done = 0; } sctp_snd_sb_alloc(stcb, sp->length); atomic_add_int(&asoc->stream_queue_cnt, 1); if ((srcv->sinfo_flags & SCTP_UNORDERED) == 0) { sp->strseq = strm->next_sequence_sent; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_AT_SEND_2_SCTP) { sctp_misc_ints(SCTP_STRMOUT_LOG_ASSIGN, (uintptr_t) stcb, sp->length, (uint32_t) ((srcv->sinfo_stream << 16) | sp->strseq), 0); } strm->next_sequence_sent++; } else { SCTP_STAT_INCR(sctps_sends_with_unord); } TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); stcb->asoc.ss_functions.sctp_ss_add_to_stream(stcb, asoc, strm, sp, 1); SCTP_TCB_SEND_UNLOCK(stcb); } else { SCTP_TCB_SEND_LOCK(stcb); sp = TAILQ_LAST(&strm->outqueue, sctp_streamhead); SCTP_TCB_SEND_UNLOCK(stcb); if (sp == NULL) { /* ???? Huh ??? last msg is gone */ #ifdef INVARIANTS panic("Warning: Last msg marked incomplete, yet nothing left?"); #else SCTP_PRINTF("Warning: Last msg marked incomplete, yet nothing left?\n"); strm->last_msg_incomplete = 0; #endif goto do_a_copy_in; } } while (uio->uio_resid > 0) { /* How much room do we have? */ struct mbuf *new_tail, *mm; if (SCTP_SB_LIMIT_SND(so) > stcb->asoc.total_output_queue_size) max_len = SCTP_SB_LIMIT_SND(so) - stcb->asoc.total_output_queue_size; else max_len = 0; if ((max_len > SCTP_BASE_SYSCTL(sctp_add_more_threshold)) || (max_len && (SCTP_SB_LIMIT_SND(so) < SCTP_BASE_SYSCTL(sctp_add_more_threshold))) || (uio->uio_resid && (uio->uio_resid <= (int)max_len))) { sndout = 0; new_tail = NULL; if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } mm = sctp_copy_resume(sp, uio, srcv, max_len, user_marks_eor, &error, &sndout, &new_tail); if ((mm == NULL) || error) { if (mm) { sctp_m_freem(mm); } goto out; } /* Update the mbuf and count */ SCTP_TCB_SEND_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* * we need to get out. Peer probably * aborted. */ sctp_m_freem(mm); if (stcb->asoc.state & SCTP_PCB_FLAGS_WAS_ABORTED) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; } SCTP_TCB_SEND_UNLOCK(stcb); goto out; } if (sp->tail_mbuf) { /* tack it to the end */ SCTP_BUF_NEXT(sp->tail_mbuf) = mm; sp->tail_mbuf = new_tail; } else { /* A stolen mbuf */ sp->data = mm; sp->tail_mbuf = new_tail; } sctp_snd_sb_alloc(stcb, sndout); atomic_add_int(&sp->length, sndout); len += sndout; /* Did we reach EOR? */ if ((uio->uio_resid == 0) && ((user_marks_eor == 0) || (srcv->sinfo_flags & SCTP_EOF) || (user_marks_eor && (srcv->sinfo_flags & SCTP_EOR)))) { sp->msg_is_complete = 1; } else { sp->msg_is_complete = 0; } SCTP_TCB_SEND_UNLOCK(stcb); } if (uio->uio_resid == 0) { /* got it all? */ continue; } /* PR-SCTP? */ if ((asoc->peer_supports_prsctp) && (asoc->sent_queue_cnt_removeable > 0)) { /* * This is ugly but we must assure locking * order */ if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } sctp_prune_prsctp(stcb, asoc, srcv, sndlen); inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if (SCTP_SB_LIMIT_SND(so) > stcb->asoc.total_output_queue_size) max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; else max_len = 0; if (max_len > 0) { continue; } SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } /* wait for space now */ if (non_blocking) { /* Non-blocking io in place out */ goto skip_out_eof; } /* What about the INIT, send it maybe */ if (queue_only_for_init) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) { /* a collision took us forward? */ queue_only = 0; } else { sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_SET_STATE(asoc, SCTP_STATE_COOKIE_WAIT); queue_only = 1; } } if ((net->flight_size > net->cwnd) && (asoc->sctp_cmt_on_off == 0)) { SCTP_STAT_INCR(sctps_send_cwnd_avoid); queue_only = 1; } else if (asoc->ifp_had_enobuf) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (net->flight_size > (2 * net->mtu)) { queue_only = 1; } asoc->ifp_had_enobuf = 0; } un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (stcb->asoc.stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD))) { /*- * Ok, Nagle is set on and we have data outstanding. * Don't send anything and let SACKs drive out the * data unless wen have a "full" segment to send. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); } SCTP_STAT_INCR(sctps_naglequeued); nagle_applies = 1; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); } SCTP_STAT_INCR(sctps_naglesent); nagle_applies = 0; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_misc_ints(SCTP_CWNDLOG_PRESEND, queue_only_for_init, queue_only, nagle_applies, un_sent); sctp_misc_ints(SCTP_CWNDLOG_PRESEND, stcb->asoc.total_output_queue_size, stcb->asoc.total_flight, stcb->asoc.chunks_on_out_queue, stcb->asoc.total_flight_count); } if (queue_only_for_init) queue_only_for_init = 0; if ((queue_only == 0) && (nagle_applies == 0)) { /*- * need to start chunk output * before blocking.. note that if * a lock is already applied, then * the input via the net is happening * and I don't need to start output :-D */ if (hold_tcblock == 0) { if (SCTP_TCB_TRYLOCK(stcb)) { hold_tcblock = 1; sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } if (hold_tcblock == 1) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } } SOCKBUF_LOCK(&so->so_snd); /*- * This is a bit strange, but I think it will * work. The total_output_queue_size is locked and * protected by the TCB_LOCK, which we just released. * There is a race that can occur between releasing it * above, and me getting the socket lock, where sacks * come in but we have not put the SB_WAIT on the * so_snd buffer to get the wakeup. After the LOCK * is applied the sack_processing will also need to * LOCK the so->so_snd to do the actual sowwakeup(). So * once we have the socket buffer lock if we recheck the * size we KNOW we will get to sleep safely with the * wakeup flag in place. */ if (SCTP_SB_LIMIT_SND(so) <= (stcb->asoc.total_output_queue_size + min(SCTP_BASE_SYSCTL(sctp_add_more_threshold), SCTP_SB_LIMIT_SND(so)))) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_INTO_BLK, so, asoc, uio->uio_resid); } be.error = 0; stcb->block_entry = &be; error = sbwait(&so->so_snd); stcb->block_entry = NULL; if (error || so->so_error || be.error) { if (error == 0) { if (so->so_error) error = so->so_error; if (be.error) { error = be.error; } } SOCKBUF_UNLOCK(&so->so_snd); goto out_unlocked; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, so, asoc, stcb->asoc.total_output_queue_size); } } SOCKBUF_UNLOCK(&so->so_snd); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } } SCTP_TCB_SEND_LOCK(stcb); if (sp) { if (sp->msg_is_complete == 0) { strm->last_msg_incomplete = 1; asoc->stream_locked = 1; asoc->stream_locked_on = srcv->sinfo_stream; } else { sp->sender_all_done = 1; strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } } else { SCTP_PRINTF("Huh no sp TSNH?\n"); strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } SCTP_TCB_SEND_UNLOCK(stcb); if (uio->uio_resid == 0) { got_all_of_the_send = 1; } } else { /* We send in a 0, since we do NOT have any locks */ error = sctp_msg_append(stcb, net, top, srcv, 0); top = NULL; if (srcv->sinfo_flags & SCTP_EOF) { /* * This should only happen for Panda for the mbuf * send case, which does NOT yet support EEOR mode. * Thus, we can just set this flag to do the proper * EOF handling. */ got_all_of_the_send = 1; } } if (error) { goto out; } dataless_eof: /* EOF thing ? */ if ((srcv->sinfo_flags & SCTP_EOF) && (got_all_of_the_send == 1) && (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { int cnt; SCTP_STAT_INCR(sctps_sends_with_eof); error = 0; if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } cnt = sctp_is_there_unsent_data(stcb); if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* there is nothing queued to send, so I'm done... */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* only send SHUTDOWN the first time through */ sctp_send_shutdown(stcb, stcb->asoc.primary_destination); if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } else { /*- * we still got (or just got) data to send, so set * SHUTDOWN_PENDING */ /*- * XXX sockets draft says that SCTP_EOF should be * sent with no data. currently, we will allow user * data to be sent first and move to * SHUTDOWN-PENDING */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp) { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { abort_anyway: if (free_cnt_applied) { atomic_add_int(&stcb->asoc.refcnt, -1); free_cnt_applied = 0; } sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, NULL, SCTP_SO_LOCKED); /* * now relock the stcb so everything * is sane */ hold_tcblock = 0; stcb = NULL; goto out; } sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_feature_off(inp, SCTP_PCB_FLAGS_NODELAY); } } } skip_out_eof: if (!TAILQ_EMPTY(&stcb->asoc.control_send_queue)) { some_on_control = 1; } if (queue_only_for_init) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) { /* a collision took us forward? */ queue_only = 0; } else { sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); queue_only = 1; } } if ((net->flight_size > net->cwnd) && (stcb->asoc.sctp_cmt_on_off == 0)) { SCTP_STAT_INCR(sctps_send_cwnd_avoid); queue_only = 1; } else if (asoc->ifp_had_enobuf) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (net->flight_size > (2 * net->mtu)) { queue_only = 1; } asoc->ifp_had_enobuf = 0; } un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (stcb->asoc.stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD))) { /*- * Ok, Nagle is set on and we have data outstanding. * Don't send anything and let SACKs drive out the * data unless wen have a "full" segment to send. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); } SCTP_STAT_INCR(sctps_naglequeued); nagle_applies = 1; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); } SCTP_STAT_INCR(sctps_naglesent); nagle_applies = 0; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_misc_ints(SCTP_CWNDLOG_PRESEND, queue_only_for_init, queue_only, nagle_applies, un_sent); sctp_misc_ints(SCTP_CWNDLOG_PRESEND, stcb->asoc.total_output_queue_size, stcb->asoc.total_flight, stcb->asoc.chunks_on_out_queue, stcb->asoc.total_flight_count); } if (queue_only_for_init) queue_only_for_init = 0; if ((queue_only == 0) && (nagle_applies == 0) && (stcb->asoc.peers_rwnd && un_sent)) { /* we can attempt to send too. */ if (hold_tcblock == 0) { /* * If there is activity recv'ing sacks no need to * send */ if (SCTP_TCB_TRYLOCK(stcb)) { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); hold_tcblock = 1; } } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } } else if ((queue_only == 0) && (stcb->asoc.peers_rwnd == 0) && (stcb->asoc.total_flight == 0)) { /* We get to have a probe outstanding */ if (hold_tcblock == 0) { hold_tcblock = 1; SCTP_TCB_LOCK(stcb); } sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } else if (some_on_control) { int num_out, reason, frag_point; /* Here we do control only */ if (hold_tcblock == 0) { hold_tcblock = 1; SCTP_TCB_LOCK(stcb); } frag_point = sctp_get_frag_point(stcb, &stcb->asoc); (void)sctp_med_chunk_output(inp, stcb, &stcb->asoc, &num_out, &reason, 1, 1, &now, &now_filled, frag_point, SCTP_SO_LOCKED); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "USR Send complete qo:%d prw:%d unsent:%d tf:%d cooq:%d toqs:%d err:%d\n", queue_only, stcb->asoc.peers_rwnd, un_sent, stcb->asoc.total_flight, stcb->asoc.chunks_on_out_queue, stcb->asoc.total_output_queue_size, error); out: out_unlocked: if (local_soresv && stcb) { atomic_subtract_int(&stcb->asoc.sb_send_resv, sndlen); local_soresv = 0; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } if ((stcb) && hold_tcblock) { SCTP_TCB_UNLOCK(stcb); } if (stcb && free_cnt_applied) { atomic_add_int(&stcb->asoc.refcnt, -1); } #ifdef INVARIANTS if (stcb) { if (mtx_owned(&stcb->tcb_mtx)) { panic("Leaving with tcb mtx owned?"); } if (mtx_owned(&stcb->tcb_send_mtx)) { panic("Leaving with tcb send mtx owned?"); } } #endif #ifdef INVARIANTS if (inp) { sctp_validate_no_locks(inp); } else { printf("Warning - inp is NULL so cant validate locks\n"); } #endif if (top) { sctp_m_freem(top); } if (control) { sctp_m_freem(control); } return (error); } /* * generate an AUTHentication chunk, if required */ struct mbuf * sctp_add_auth_chunk(struct mbuf *m, struct mbuf **m_end, struct sctp_auth_chunk **auth_ret, uint32_t * offset, struct sctp_tcb *stcb, uint8_t chunk) { struct mbuf *m_auth; struct sctp_auth_chunk *auth; int chunk_len; if ((m_end == NULL) || (auth_ret == NULL) || (offset == NULL) || (stcb == NULL)) return (m); /* sysctl disabled auth? */ if (SCTP_BASE_SYSCTL(sctp_auth_disable)) return (m); /* peer doesn't do auth... */ if (!stcb->asoc.peer_supports_auth) { return (m); } /* does the requested chunk require auth? */ if (!sctp_auth_is_required_chunk(chunk, stcb->asoc.peer_auth_chunks)) { return (m); } m_auth = sctp_get_mbuf_for_msg(sizeof(*auth), 0, M_DONTWAIT, 1, MT_HEADER); if (m_auth == NULL) { /* no mbuf's */ return (m); } /* reserve some space if this will be the first mbuf */ if (m == NULL) SCTP_BUF_RESV_UF(m_auth, SCTP_MIN_OVERHEAD); /* fill in the AUTH chunk details */ auth = mtod(m_auth, struct sctp_auth_chunk *); bzero(auth, sizeof(*auth)); auth->ch.chunk_type = SCTP_AUTHENTICATION; auth->ch.chunk_flags = 0; chunk_len = sizeof(*auth) + sctp_get_hmac_digest_len(stcb->asoc.peer_hmac_id); auth->ch.chunk_length = htons(chunk_len); auth->hmac_id = htons(stcb->asoc.peer_hmac_id); /* key id and hmac digest will be computed and filled in upon send */ /* save the offset where the auth was inserted into the chain */ if (m != NULL) { struct mbuf *cn; *offset = 0; cn = m; while (cn) { *offset += SCTP_BUF_LEN(cn); cn = SCTP_BUF_NEXT(cn); } } else *offset = 0; /* update length and return pointer to the auth chunk */ SCTP_BUF_LEN(m_auth) = chunk_len; m = sctp_copy_mbufchain(m_auth, m, m_end, 1, chunk_len, 0); if (auth_ret != NULL) *auth_ret = auth; return (m); } #ifdef INET6 int sctp_v6src_match_nexthop(struct sockaddr_in6 *src6, sctp_route_t * ro) { struct nd_prefix *pfx = NULL; struct nd_pfxrouter *pfxrtr = NULL; struct sockaddr_in6 gw6; if (ro == NULL || ro->ro_rt == NULL || src6->sin6_family != AF_INET6) return (0); /* get prefix entry of address */ LIST_FOREACH(pfx, &MODULE_GLOBAL(nd_prefix), ndpr_entry) { if (pfx->ndpr_stateflags & NDPRF_DETACHED) continue; if (IN6_ARE_MASKED_ADDR_EQUAL(&pfx->ndpr_prefix.sin6_addr, &src6->sin6_addr, &pfx->ndpr_mask)) break; } /* no prefix entry in the prefix list */ if (pfx == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "No prefix entry for "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); return (0); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "v6src_match_nexthop(), Prefix entry is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); /* search installed gateway from prefix entry */ for (pfxrtr = pfx->ndpr_advrtrs.lh_first; pfxrtr; pfxrtr = pfxrtr->pfr_next) { memset(&gw6, 0, sizeof(struct sockaddr_in6)); gw6.sin6_family = AF_INET6; gw6.sin6_len = sizeof(struct sockaddr_in6); memcpy(&gw6.sin6_addr, &pfxrtr->router->rtaddr, sizeof(struct in6_addr)); SCTPDBG(SCTP_DEBUG_OUTPUT2, "prefix router is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&gw6); SCTPDBG(SCTP_DEBUG_OUTPUT2, "installed router is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); if (sctp_cmpaddr((struct sockaddr *)&gw6, ro->ro_rt->rt_gateway)) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is installed\n"); return (1); } } SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is not installed\n"); return (0); } #endif int sctp_v4src_match_nexthop(struct sctp_ifa *sifa, sctp_route_t * ro) { struct sockaddr_in *sin, *mask; struct ifaddr *ifa; struct in_addr srcnetaddr, gwnetaddr; if (ro == NULL || ro->ro_rt == NULL || sifa->address.sa.sa_family != AF_INET) { return (0); } ifa = (struct ifaddr *)sifa->ifa; mask = (struct sockaddr_in *)(ifa->ifa_netmask); sin = (struct sockaddr_in *)&sifa->address.sin; srcnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: src address is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", srcnetaddr.s_addr); sin = (struct sockaddr_in *)ro->ro_rt->rt_gateway; gwnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: nexthop is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", gwnetaddr.s_addr); if (srcnetaddr.s_addr == gwnetaddr.s_addr) { return (1); } return (0); } diff --git a/sys/netinet/sctp_pcb.c b/sys/netinet/sctp_pcb.c index 246ff0de0c5f..cdbc939d5294 100644 --- a/sys/netinet/sctp_pcb.c +++ b/sys/netinet/sctp_pcb.c @@ -1,6680 +1,6676 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include VNET_DEFINE(struct sctp_base_info, system_base_info); /* FIX: we don't handle multiple link local scopes */ /* "scopeless" replacement IN6_ARE_ADDR_EQUAL */ #ifdef INET6 int SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b) { struct sockaddr_in6 tmp_a, tmp_b; memcpy(&tmp_a, a, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) { return 0; } memcpy(&tmp_b, b, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) { return 0; } return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr)); } #endif void sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb) { /* * We really don't need to lock this, but I will just because it * does not hurt. */ SCTP_INP_INFO_RLOCK(); spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep); spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc); spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr); spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr); spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk); spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq); spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq); spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks); SCTP_INP_INFO_RUNLOCK(); } /* * Addresses are added to VRF's (Virtual Router's). For BSD we * have only the default VRF 0. We maintain a hash list of * VRF's. Each VRF has its own list of sctp_ifn's. Each of * these has a list of addresses. When we add a new address * to a VRF we lookup the ifn/ifn_index, if the ifn does * not exist we create it and add it to the list of IFN's * within the VRF. Once we have the sctp_ifn, we add the * address to the list. So we look something like: * * hash-vrf-table * vrf-> ifn-> ifn -> ifn * vrf | * ... +--ifa-> ifa -> ifa * vrf * * We keep these separate lists since the SCTP subsystem will * point to these from its source address selection nets structure. * When an address is deleted it does not happen right away on * the SCTP side, it gets scheduled. What we do when a * delete happens is immediately remove the address from * the master list and decrement the refcount. As our * addip iterator works through and frees the src address * selection pointing to the sctp_ifa, eventually the refcount * will reach 0 and we will delete it. Note that it is assumed * that any locking on system level ifn/ifa is done at the * caller of these functions and these routines will only * lock the SCTP structures as they add or delete things. * * Other notes on VRF concepts. * - An endpoint can be in multiple VRF's * - An association lives within a VRF and only one VRF. * - Any incoming packet we can deduce the VRF for by * looking at the mbuf/pak inbound (for BSD its VRF=0 :D) * - Any downward send call or connect call must supply the * VRF via ancillary data or via some sort of set default * VRF socket option call (again for BSD no brainer since * the VRF is always 0). * - An endpoint may add multiple VRF's to it. * - Listening sockets can accept associations in any * of the VRF's they are in but the assoc will end up * in only one VRF (gotten from the packet or connect/send). * */ struct sctp_vrf * sctp_allocate_vrf(int vrf_id) { struct sctp_vrf *vrf = NULL; struct sctp_vrflist *bucket; /* First allocate the VRF structure */ vrf = sctp_find_vrf(vrf_id); if (vrf) { /* Already allocated */ return (vrf); } SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf), SCTP_M_VRF); if (vrf == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif return (NULL); } /* setup the VRF */ memset(vrf, 0, sizeof(struct sctp_vrf)); vrf->vrf_id = vrf_id; LIST_INIT(&vrf->ifnlist); vrf->total_ifa_count = 0; vrf->refcount = 0; /* now also setup table ids */ SCTP_INIT_VRF_TABLEID(vrf); /* Init the HASH of addresses */ vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE, &vrf->vrf_addr_hashmark); if (vrf->vrf_addr_hash == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif SCTP_FREE(vrf, SCTP_M_VRF); return (NULL); } /* Add it to the hash table */ bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_INSERT_HEAD(bucket, vrf, next_vrf); atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); return (vrf); } struct sctp_ifn * sctp_find_ifn(void *ifn, uint32_t ifn_index) { struct sctp_ifn *sctp_ifnp; struct sctp_ifnlist *hash_ifn_head; /* * We assume the lock is held for the addresses if that's wrong * problems could occur :-) */ hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) { if (sctp_ifnp->ifn_index == ifn_index) { return (sctp_ifnp); } if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) { return (sctp_ifnp); } } return (NULL); } struct sctp_vrf * sctp_find_vrf(uint32_t vrf_id) { struct sctp_vrflist *bucket; struct sctp_vrf *liste; bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_FOREACH(liste, bucket, next_vrf) { if (vrf_id == liste->vrf_id) { return (liste); } } return (NULL); } void sctp_free_vrf(struct sctp_vrf *vrf) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) { if (vrf->vrf_addr_hash) { SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); vrf->vrf_addr_hash = NULL; } /* We zero'd the count */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); } } void sctp_free_ifn(struct sctp_ifn *sctp_ifnp) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) { /* We zero'd the count */ if (sctp_ifnp->vrf) { sctp_free_vrf(sctp_ifnp->vrf); } SCTP_FREE(sctp_ifnp, SCTP_M_IFN); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); } } void sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu) { struct sctp_ifn *sctp_ifnp; sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index); if (sctp_ifnp != NULL) { sctp_ifnp->ifn_mtu = mtu; } } void sctp_free_ifa(struct sctp_ifa *sctp_ifap) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) { /* We zero'd the count */ if (sctp_ifap->ifn_p) { sctp_free_ifn(sctp_ifap->ifn_p); } SCTP_FREE(sctp_ifap, SCTP_M_IFA); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); } } static void sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock) { struct sctp_ifn *found; found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index); if (found == NULL) { /* Not in the list.. sorry */ return; } if (hold_addr_lock == 0) SCTP_IPI_ADDR_WLOCK(); LIST_REMOVE(sctp_ifnp, next_bucket); LIST_REMOVE(sctp_ifnp, next_ifn); SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index, sctp_ifnp->registered_af); if (hold_addr_lock == 0) SCTP_IPI_ADDR_WUNLOCK(); /* Take away the reference, and possibly free it */ sctp_free_ifn(sctp_ifnp); } void sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { int len1, len2; len1 = strlen(if_name); len2 = strlen(sctp_ifap->ifn_p->ifn_name); if (len1 != len2) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different length %d vs %d - ignored\n", len1, len2); goto out; } if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID); sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; out: SCTP_IPI_ADDR_RUNLOCK(); } void sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { int len1, len2; len1 = strlen(if_name); len2 = strlen(sctp_ifap->ifn_p->ifn_name); if (len1 != len2) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different length %d vs %d - ignored\n", len1, len2); goto out; } if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE); sctp_ifap->localifa_flags |= SCTP_ADDR_VALID; out: SCTP_IPI_ADDR_RUNLOCK(); } /*- * Add an ifa to an ifn. * Register the interface as necessary. * NOTE: ADDR write lock MUST be held. */ static void sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap) { int ifa_af; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); /* update address counts */ sctp_ifnp->ifa_count++; ifa_af = sctp_ifap->address.sa.sa_family; if (ifa_af == AF_INET) sctp_ifnp->num_v4++; else sctp_ifnp->num_v6++; if (sctp_ifnp->ifa_count == 1) { /* register the new interface */ SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af); sctp_ifnp->registered_af = ifa_af; } } /*- * Remove an ifa from its ifn. * If no more addresses exist, remove the ifn too. Otherwise, re-register * the interface based on the remaining address families left. * NOTE: ADDR write lock MUST be held. */ static void sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap) { uint32_t ifn_index; LIST_REMOVE(sctp_ifap, next_ifa); if (sctp_ifap->ifn_p) { /* update address counts */ sctp_ifap->ifn_p->ifa_count--; if (sctp_ifap->address.sa.sa_family == AF_INET6) sctp_ifap->ifn_p->num_v6--; else if (sctp_ifap->address.sa.sa_family == AF_INET) sctp_ifap->ifn_p->num_v4--; ifn_index = sctp_ifap->ifn_p->ifn_index; if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) { /* remove the ifn, possibly freeing it */ sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED); } else { /* re-register address family type, if needed */ if ((sctp_ifap->ifn_p->num_v6 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET6)) { SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET6); SCTP_REGISTER_INTERFACE(ifn_index, AF_INET); sctp_ifap->ifn_p->registered_af = AF_INET; } else if ((sctp_ifap->ifn_p->num_v4 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET)) { SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET); SCTP_REGISTER_INTERFACE(ifn_index, AF_INET6); sctp_ifap->ifn_p->registered_af = AF_INET6; } /* free the ifn refcount */ sctp_free_ifn(sctp_ifap->ifn_p); } sctp_ifap->ifn_p = NULL; } } struct sctp_ifa * sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index, uint32_t ifn_type, const char *if_name, void *ifa, struct sockaddr *addr, uint32_t ifa_flags, int dynamic_add) { struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifnp = NULL; struct sctp_ifa *sctp_ifap = NULL; struct sctp_ifalist *hash_addr_head; struct sctp_ifnlist *hash_ifn_head; uint32_t hash_of_addr; int new_ifn_af = 0; #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif SCTP_IPI_ADDR_WLOCK(); sctp_ifnp = sctp_find_ifn(ifn, ifn_index); if (sctp_ifnp) { vrf = sctp_ifnp->vrf; } else { vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { vrf = sctp_allocate_vrf(vrf_id); if (vrf == NULL) { SCTP_IPI_ADDR_WUNLOCK(); return (NULL); } } } if (sctp_ifnp == NULL) { /* * build one and add it, can't hold lock until after malloc * done though. */ SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, sizeof(struct sctp_ifn), SCTP_M_IFN); if (sctp_ifnp == NULL) { #ifdef INVARIANTS panic("No memory for IFN"); #endif return (NULL); } memset(sctp_ifnp, 0, sizeof(struct sctp_ifn)); sctp_ifnp->ifn_index = ifn_index; sctp_ifnp->ifn_p = ifn; sctp_ifnp->ifn_type = ifn_type; sctp_ifnp->refcount = 0; sctp_ifnp->vrf = vrf; atomic_add_int(&vrf->refcount, 1); sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family); if (if_name != NULL) { memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ); } else { memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ)); } hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_INIT(&sctp_ifnp->ifalist); SCTP_IPI_ADDR_WLOCK(); LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket); LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn); atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); new_ifn_af = 1; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Hmm, it already exists? */ if ((sctp_ifap->ifn_p) && (sctp_ifap->ifn_p->ifn_index == ifn_index)) { SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n", sctp_ifap->ifn_p->ifn_name, ifn_index, sctp_ifap); if (new_ifn_af) { /* Remove the created one that we don't want */ sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED); } if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) { /* easy to solve, just switch back to active */ SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n"); sctp_ifap->localifa_flags = SCTP_ADDR_VALID; sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); } exit_stage_left: SCTP_IPI_ADDR_WUNLOCK(); return (sctp_ifap); } else { if (sctp_ifap->ifn_p) { /* * The last IFN gets the address, remove the * old one */ SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n", sctp_ifap, sctp_ifap->ifn_p->ifn_name, sctp_ifap->ifn_p->ifn_index, if_name, ifn_index); /* remove the address from the old ifn */ sctp_remove_ifa_from_ifn(sctp_ifap); /* move the address over to the new ifn */ sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); goto exit_stage_left; } else { /* repair ifnp which was NULL ? */ sctp_ifap->localifa_flags = SCTP_ADDR_VALID; SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n", sctp_ifnp, sctp_ifap); sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); } goto exit_stage_left; } } SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA); if (sctp_ifap == NULL) { #ifdef INVARIANTS panic("No memory for IFA"); #endif return (NULL); } memset(sctp_ifap, 0, sizeof(struct sctp_ifa)); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifnp->refcount, 1); sctp_ifap->vrf_id = vrf_id; sctp_ifap->ifa = ifa; memcpy(&sctp_ifap->address, addr, addr->sa_len); sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE; sctp_ifap->flags = ifa_flags; /* Set scope */ switch (sctp_ifap->address.sa.sa_family) { case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifap->address.sin; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_loop = 1; } if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v4++; if (new_ifn_af) new_ifn_af = AF_INET; break; } #ifdef INET6 case AF_INET6: { /* ok to use deprecated addresses? */ struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) { sctp_ifap->src_is_loop = 1; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v6++; if (new_ifn_af) new_ifn_af = AF_INET6; break; } #endif default: new_ifn_af = 0; break; } hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa); if ((sctp_ifap->src_is_priv == 0) && (sctp_ifap->src_is_loop == 0)) { sctp_ifap->src_is_glob = 1; } SCTP_IPI_ADDR_WLOCK(); hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket); sctp_ifap->refcount = 1; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifnp->ifa_count++; vrf->total_ifa_count++; atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); if (new_ifn_af) { SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af); sctp_ifnp->registered_af = new_ifn_af; } SCTP_IPI_ADDR_WUNLOCK(); if (dynamic_add) { /* * Bump up the refcount so that when the timer completes it * will drop back down. */ struct sctp_laddr *wi; atomic_add_int(&sctp_ifap->refcount, 1); wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Opps, must decrement the count */ sctp_del_addr_from_vrf(vrf_id, addr, ifn_index, if_name); return (NULL); } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_ADD_IP_ADDRESS; SCTP_WQ_ADDR_LOCK(); LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_WQ_ADDR_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } else { /* it's ready for use */ sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE; } return (sctp_ifap); } void sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr, uint32_t ifn_index, const char *if_name) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_WLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out_now; } #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Validate the delete */ if (sctp_ifap->ifn_p) { int valid = 0; /*- * The name has priority over the ifn_index * if its given. We do this especially for * panda who might recycle indexes fast. */ if (if_name) { int len1, len2; len1 = min(SCTP_IFNAMSIZ, strlen(if_name)); len2 = min(SCTP_IFNAMSIZ, strlen(sctp_ifap->ifn_p->ifn_name)); if (len1 && len2 && (len1 == len2)) { /* we can compare them */ if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) == 0) { /* * They match its a correct * delete */ valid = 1; } } } if (!valid) { /* last ditch check ifn_index */ if (ifn_index == sctp_ifap->ifn_p->ifn_index) { valid = 1; } } if (!valid) { SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n", ifn_index, ((if_name == NULL) ? "NULL" : if_name)); SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n", sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name); SCTP_IPI_ADDR_WUNLOCK(); return; } } SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap); sctp_ifap->localifa_flags &= SCTP_ADDR_VALID; sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; vrf->total_ifa_count--; LIST_REMOVE(sctp_ifap, next_bucket); sctp_remove_ifa_from_ifn(sctp_ifap); } #ifdef SCTP_DEBUG else { SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:", ifn_index); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif out_now: SCTP_IPI_ADDR_WUNLOCK(); if (sctp_ifap) { struct sctp_laddr *wi; wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Oops, must decrement the count */ sctp_free_ifa(sctp_ifap); return; } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_DEL_IP_ADDRESS; SCTP_WQ_ADDR_LOCK(); /* * Should this really be a tailq? As it is we will process * the newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_WQ_ADDR_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } return; } static struct sctp_tcb * sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from, struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id) { /**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */ /* * If we support the TCP model, then we must now dig through to see * if we can find our endpoint in the list of tcp ep's. */ uint16_t lport, rport; struct sctppcbhead *ephead; struct sctp_inpcb *inp; struct sctp_laddr *laddr; struct sctp_tcb *stcb; struct sctp_nets *net; if ((to == NULL) || (from == NULL)) { return (NULL); } if (to->sa_family == AF_INET && from->sa_family == AF_INET) { lport = ((struct sockaddr_in *)to)->sin_port; rport = ((struct sockaddr_in *)from)->sin_port; } else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) { lport = ((struct sockaddr_in6 *)to)->sin6_port; rport = ((struct sockaddr_in6 *)from)->sin6_port; } else { return NULL; } ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; /* * Ok now for each of the guys in this bucket we must look and see: * - Does the remote port match. - Does there single association's * addresses match this address (to). If so we update p_ep to point * to this ep and return the tcb from it. */ LIST_FOREACH(inp, ephead, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if (lport != inp->sctp_lport) { SCTP_INP_RUNLOCK(inp); continue; } if (inp->def_vrf_id != vrf_id) { SCTP_INP_RUNLOCK(inp); continue; } /* check to see if the ep has one of the addresses */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* We are NOT bound all, so look further */ int match = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == to->sa_family) { /* see if it matches */ struct sockaddr_in *intf_addr, *sin; intf_addr = &laddr->ifa->address.sin; sin = (struct sockaddr_in *)to; if (from->sa_family == AF_INET) { if (sin->sin_addr.s_addr == intf_addr->sin_addr.s_addr) { match = 1; break; } } #ifdef INET6 if (from->sa_family == AF_INET6) { struct sockaddr_in6 *intf_addr6; struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *) to; intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { match = 1; break; } } #endif } } if (match == 0) { /* This endpoint does not have this address */ SCTP_INP_RUNLOCK(inp); continue; } } /* * Ok if we hit here the ep has the address, does it hold * the tcb? */ stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); continue; } SCTP_TCB_LOCK(stcb); if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } /* Does this TCB have a matching address? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._l_addr.sa.sa_family != from->sa_family) { /* not the same family, can't be a match */ continue; } switch (from->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *)&net->ro._l_addr; rsin = (struct sockaddr_in *)from; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)from; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); } return (NULL); } static int sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to) { int loopback_scope, ipv4_local_scope, local_scope, site_scope; int ipv4_addr_legal, ipv6_addr_legal; struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; loopback_scope = stcb->asoc.loopback_scope; ipv4_local_scope = stcb->asoc.ipv4_local_scope; local_scope = stcb->asoc.local_scope; site_scope = stcb->asoc.site_scope; ipv4_addr_legal = ipv6_addr_legal = 0; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ipv6_addr_legal = 1; if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) { ipv4_addr_legal = 1; } } else { ipv4_addr_legal = 1; } SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(stcb->asoc.vrf_id); if (vrf == NULL) { /* no vrf, no addresses */ SCTP_IPI_ADDR_RUNLOCK(); return (0); } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (sctp_is_addr_restricted(stcb, sctp_ifa)) continue; switch (sctp_ifa->address.sa.sa_family) { #ifdef INET case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin, *rsin; sin = &sctp_ifa->address.sin; rsin = (struct sockaddr_in *)to; if ((ipv4_local_scope == 0) && IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { continue; } if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } } break; #endif #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6, *rsin6; sin6 = &sctp_ifa->address.sin6; rsin6 = (struct sockaddr_in6 *)to; if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } } break; #endif default: /* TSNH */ break; } } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { if (sctp_is_addr_restricted(stcb, laddr->ifa)) { continue; } if (laddr->ifa->address.sa.sa_family != to->sa_family) { continue; } switch (to->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *)&laddr->ifa->address.sin; rsin = (struct sockaddr_in *)to; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6; rsin6 = (struct sockaddr_in6 *)to; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } break; } #endif default: /* TSNH */ break; } } } SCTP_IPI_ADDR_RUNLOCK(); return (0); } /* * rules for use * * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an * stcb, both will be locked (locked_tcb and stcb) but decrement will be done * (if locked == NULL). 3) Decrement happens on return ONLY if locked == * NULL. */ struct sctp_tcb * sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote, struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb) { struct sctpasochead *head; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; struct sctp_nets *net; uint16_t rport; inp = *inp_p; if (remote->sa_family == AF_INET) { rport = (((struct sockaddr_in *)remote)->sin_port); } else if (remote->sa_family == AF_INET6) { rport = (((struct sockaddr_in6 *)remote)->sin6_port); } else { return (NULL); } if (locked_tcb) { /* * UN-lock so we can do proper locking here this occurs when * called from load_addresses_from_init. */ atomic_add_int(&locked_tcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(locked_tcb); } SCTP_INP_INFO_RLOCK(); if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { /*- * Now either this guy is our listener or it's the * connector. If it is the one that issued the connect, then * it's only chance is to be the first TCB in the list. If * it is the acceptor, then do the special_lookup to hash * and find the real inp. */ if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) { /* to is peer addr, from is my addr */ stcb = sctp_tcb_special_locate(inp_p, remote, local, netp, inp->def_vrf_id); if ((stcb != NULL) && (locked_tcb == NULL)) { /* we have a locked tcb, lower refcount */ SCTP_INP_DECR_REF(inp); } if ((locked_tcb != NULL) && (locked_tcb != stcb)) { SCTP_INP_RLOCK(locked_tcb->sctp_ep); SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); } SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { goto null_return; } SCTP_TCB_LOCK(stcb); if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); goto null_return; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); goto null_return; } if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { SCTP_TCB_UNLOCK(stcb); goto null_return; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport, inp->sctp_hashmark)]; if (head == NULL) { goto null_return; } LIST_FOREACH(stcb, head, sctp_tcbhash) { if (stcb->rport != rport) { /* remote port does not match */ continue; } SCTP_TCB_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { SCTP_TCB_UNLOCK(stcb); continue; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *) &net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } null_return: /* clean up for returning null */ if (locked_tcb) { SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); /* not found */ return (NULL); } /* * Find an association for a specific endpoint using the association id given * out in the COMM_UP notification */ struct sctp_tcb * sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) { /* * Use my the assoc_id to find a endpoint */ struct sctpasochead *head; struct sctp_tcb *stcb; uint32_t id; if (inp == NULL) { SCTP_PRINTF("TSNH ep_associd\n"); return (NULL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_PRINTF("TSNH ep_associd0\n"); return (NULL); } id = (uint32_t) asoc_id; head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; if (head == NULL) { /* invalid id TSNH */ SCTP_PRINTF("TSNH ep_associd1\n"); return (NULL); } LIST_FOREACH(stcb, head, sctp_tcbasocidhash) { if (stcb->asoc.assoc_id == id) { if (inp != stcb->sctp_ep) { /* * some other guy has the same id active (id * collision ??). */ SCTP_PRINTF("TSNH ep_associd2\n"); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { continue; } if (want_lock) { SCTP_TCB_LOCK(stcb); } return (stcb); } } return (NULL); } struct sctp_tcb * sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) { struct sctp_tcb *stcb; SCTP_INP_RLOCK(inp); stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock); SCTP_INP_RUNLOCK(inp); return (stcb); } static struct sctp_inpcb * sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head, uint16_t lport, uint32_t vrf_id) { struct sctp_inpcb *inp; struct sockaddr_in *sin; #ifdef INET6 struct sockaddr_in6 *sin6; #endif struct sctp_laddr *laddr; #ifdef INET6 struct sockaddr_in6 *intf_addr6; #endif int fnd; /* * Endpoint probe expects that the INP_INFO is locked. */ sin = NULL; #ifdef INET6 sin6 = NULL; #endif switch (nam->sa_family) { case AF_INET: sin = (struct sockaddr_in *)nam; break; #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)nam; break; #endif default: /* unsupported family */ return (NULL); } if (head == NULL) return (NULL); LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) && (inp->sctp_lport == lport)) { /* got it */ if ((nam->sa_family == AF_INET) && (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* IPv4 on a IPv6 socket with ONLY IPv6 set */ SCTP_INP_RUNLOCK(inp); continue; } /* A V6 address and the endpoint is NOT bound V6 */ if (nam->sa_family == AF_INET6 && (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { SCTP_INP_RUNLOCK(inp); continue; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; SCTP_INP_RUNLOCK(inp); if (!fnd) continue; return (inp); } SCTP_INP_RUNLOCK(inp); } if ((nam->sa_family == AF_INET) && (sin->sin_addr.s_addr == INADDR_ANY)) { /* Can't hunt for one that has no address specified */ return (NULL); } #ifdef INET6 if ((nam->sa_family == AF_INET6) && (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) { /* Can't hunt for one that has no address specified */ return (NULL); } #endif /* * ok, not bound to all so see if we can find a EP bound to this * address. */ LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) { SCTP_INP_RUNLOCK(inp); continue; } /* * Ok this could be a likely candidate, look at all of its * addresses */ if (inp->sctp_lport != lport) { SCTP_INP_RUNLOCK(inp); continue; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; if (!fnd) { SCTP_INP_RUNLOCK(inp); continue; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ", laddr->ifa); if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == nam->sa_family) { /* possible, see if it matches */ struct sockaddr_in *intf_addr; intf_addr = &laddr->ifa->address.sin; switch (nam->sa_family) { case AF_INET: if (sin->sin_addr.s_addr == intf_addr->sin_addr.s_addr) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #ifdef INET6 case AF_INET6: intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #endif } } } SCTP_INP_RUNLOCK(inp); } return (NULL); } static struct sctp_inpcb * sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) { struct sctppcbhead *head; struct sctp_inpcb *t_inp; int fnd; head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; LIST_FOREACH(t_inp, head, sctp_hash) { if (t_inp->sctp_lport != lport) { continue; } /* is it in the VRF in question */ fnd = 0; if (t_inp->def_vrf_id == vrf_id) fnd = 1; if (!fnd) continue; /* This one is in use. */ /* check the v6/v4 binding issue */ if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(t_inp)) { if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { /* collision in V6 space */ return (t_inp); } else { /* inp is BOUND_V4 no conflict */ continue; } } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { /* t_inp is bound v4 and v6, conflict always */ return (t_inp); } else { /* t_inp is bound only V4 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* no conflict */ continue; } /* else fall through to conflict */ } return (t_inp); } return (NULL); } int sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp) { /* For 1-2-1 with port reuse */ struct sctppcbhead *head; struct sctp_inpcb *tinp; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) { /* only works with port reuse on */ return (-1); } if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) { return (0); } SCTP_INP_RUNLOCK(inp); head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; /* Kick out all non-listeners to the TCP hash */ LIST_FOREACH(tinp, head, sctp_hash) { if (tinp->sctp_lport != inp->sctp_lport) { continue; } if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { continue; } if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { continue; } if (tinp->sctp_socket->so_qlimit) { continue; } SCTP_INP_WLOCK(tinp); LIST_REMOVE(tinp, sctp_hash); head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))]; tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; LIST_INSERT_HEAD(head, tinp, sctp_hash); SCTP_INP_WUNLOCK(tinp); } SCTP_INP_WLOCK(inp); /* Pull from where he was */ LIST_REMOVE(inp, sctp_hash); inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL; head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; LIST_INSERT_HEAD(head, inp, sctp_hash); SCTP_INP_WUNLOCK(inp); SCTP_INP_RLOCK(inp); return (0); } struct sctp_inpcb * sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, uint32_t vrf_id) { /* * First we check the hash table to see if someone has this port * bound with just the port. */ struct sctp_inpcb *inp; struct sctppcbhead *head; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; int lport; unsigned int i; if (nam->sa_family == AF_INET) { sin = (struct sockaddr_in *)nam; lport = ((struct sockaddr_in *)nam)->sin_port; } else if (nam->sa_family == AF_INET6) { sin6 = (struct sockaddr_in6 *)nam; lport = ((struct sockaddr_in6 *)nam)->sin6_port; } else { /* unsupported family */ return (NULL); } /* * I could cheat here and just cast to one of the types but we will * do it right. It also provides the check against an Unsupported * type too. */ /* Find the head of the ALLADDR chain */ if (have_lock == 0) { SCTP_INP_INFO_RLOCK(); } head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; inp = sctp_endpoint_probe(nam, head, lport, vrf_id); /* * If the TCP model exists it could be that the main listening * endpoint is gone but there still exists a connected socket for * this guy. If so we can return the first one that we find. This * may NOT be the correct one so the caller should be wary on the * returned INP. Currently the only caller that sets find_tcp_pool * is in bindx where we are verifying that a user CAN bind the * address. He either has bound it already, or someone else has, or * its open to bind, so this is good enough. */ if (inp == NULL && find_tcp_pool) { for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) { head = &SCTP_BASE_INFO(sctp_tcpephash)[i]; inp = sctp_endpoint_probe(nam, head, lport, vrf_id); if (inp) { break; } } } if (inp) { SCTP_INP_INCR_REF(inp); } if (have_lock == 0) { SCTP_INP_INFO_RUNLOCK(); } return (inp); } /* * Find an association for an endpoint with the pointer to whom you want to * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may * need to change the *to to some other struct like a mbuf... */ struct sctp_tcb * sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from, struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, uint32_t vrf_id) { struct sctp_inpcb *inp = NULL; struct sctp_tcb *retval; SCTP_INP_INFO_RLOCK(); if (find_tcp_pool) { if (inp_p != NULL) { retval = sctp_tcb_special_locate(inp_p, from, to, netp, vrf_id); } else { retval = sctp_tcb_special_locate(&inp, from, to, netp, vrf_id); } if (retval != NULL) { SCTP_INP_INFO_RUNLOCK(); return (retval); } } inp = sctp_pcb_findep(to, 0, 1, vrf_id); if (inp_p != NULL) { *inp_p = inp; } SCTP_INP_INFO_RUNLOCK(); if (inp == NULL) { return (NULL); } /* * ok, we have an endpoint, now lets find the assoc for it (if any) * we now place the source address or from in the to of the find * endpoint call. Since in reality this chain is used from the * inbound packet side. */ if (inp_p != NULL) { retval = sctp_findassociation_ep_addr(inp_p, from, netp, to, NULL); } else { retval = sctp_findassociation_ep_addr(&inp, from, netp, to, NULL); } return retval; } /* * This routine will grub through the mbuf that is a INIT or INIT-ACK and * find all addresses that the sender has specified in any address list. Each * address will be used to lookup the TCB and see if one exits. */ static struct sctp_tcb * sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, struct sockaddr *dest) { struct sockaddr_in sin4; struct sockaddr_in6 sin6; struct sctp_paramhdr *phdr, parm_buf; struct sctp_tcb *retval; uint32_t ptype, plen; memset(&sin4, 0, sizeof(sin4)); memset(&sin6, 0, sizeof(sin6)); sin4.sin_len = sizeof(sin4); sin4.sin_family = AF_INET; sin4.sin_port = sh->src_port; sin6.sin6_len = sizeof(sin6); sin6.sin6_family = AF_INET6; sin6.sin6_port = sh->src_port; retval = NULL; offset += sizeof(struct sctp_init_chunk); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr != NULL) { /* now we must see if we want the parameter */ ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (plen == 0) { break; } if (ptype == SCTP_IPV4_ADDRESS && plen == sizeof(struct sctp_ipv4addr_param)) { /* Get the rest of the address */ struct sctp_ipv4addr_param ip4_parm, *p4; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm))); if (phdr == NULL) { return (NULL); } p4 = (struct sctp_ipv4addr_param *)phdr; memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr)); /* look it up */ retval = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin4, netp, dest, NULL); if (retval != NULL) { return (retval); } } else if (ptype == SCTP_IPV6_ADDRESS && plen == sizeof(struct sctp_ipv6addr_param)) { /* Get the rest of the address */ struct sctp_ipv6addr_param ip6_parm, *p6; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm))); if (phdr == NULL) { return (NULL); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr)); /* look it up */ retval = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin6, netp, dest, NULL); if (retval != NULL) { return (retval); } } offset += SCTP_SIZE32(plen); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } return (NULL); } static struct sctp_tcb * sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport, uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag) { /* * Use my vtag to hash. If we find it we then verify the source addr * is in the assoc. If all goes well we save a bit on rec of a * packet. */ struct sctpasochead *head; struct sctp_nets *net; struct sctp_tcb *stcb; *netp = NULL; *inp_p = NULL; SCTP_INP_INFO_RLOCK(); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag, SCTP_BASE_INFO(hashasocmark))]; if (head == NULL) { /* invalid vtag */ SCTP_INP_INFO_RUNLOCK(); return (NULL); } LIST_FOREACH(stcb, head, sctp_asocs) { SCTP_INP_RLOCK(stcb->sctp_ep); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } SCTP_TCB_LOCK(stcb); SCTP_INP_RUNLOCK(stcb->sctp_ep); if (stcb->asoc.my_vtag == vtag) { /* candidate */ if (stcb->rport != rport) { SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->sctp_ep->sctp_lport != lport) { SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } /* RRS:Need toaddr check here */ if (sctp_does_stcb_own_this_addr(stcb, to) == 0) { /* Endpoint does not own this address */ SCTP_TCB_UNLOCK(stcb); continue; } if (remote_tag) { /* * If we have both vtags that's all we match * on */ if (stcb->asoc.peer_vtag == remote_tag) { /* * If both tags match we consider it * conclusive and check NO * source/destination addresses */ goto conclusive; } } if (skip_src_check) { conclusive: if (from) { net = sctp_findnet(stcb, from); } else { *netp = NULL; /* unknown */ } if (inp_p) *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } net = sctp_findnet(stcb, from); if (net) { /* yep its him. */ *netp = net; SCTP_STAT_INCR(sctps_vtagexpress); *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { /* * not him, this should only happen in rare * cases so I peg it. */ SCTP_STAT_INCR(sctps_vtagbogus); } } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_INFO_RUNLOCK(); return (NULL); } /* * Find an association with the pointer to the inbound IP packet. This can be * a IPv4 or IPv6 packet. */ struct sctp_tcb * sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) { int find_tcp_pool; struct ip *iph; struct sctp_tcb *retval; struct sockaddr_storage to_store, from_store; struct sockaddr *to = (struct sockaddr *)&to_store; struct sockaddr *from = (struct sockaddr *)&from_store; struct sctp_inpcb *inp; iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *from4; from4 = (struct sockaddr_in *)&from_store; bzero(from4, sizeof(*from4)); from4->sin_family = AF_INET; from4->sin_len = sizeof(struct sockaddr_in); from4->sin_addr.s_addr = iph->ip_src.s_addr; from4->sin_port = sh->src_port; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *from6; ip6 = mtod(m, struct ip6_hdr *); from6 = (struct sockaddr_in6 *)&from_store; bzero(from6, sizeof(*from6)); from6->sin6_family = AF_INET6; from6->sin6_len = sizeof(struct sockaddr_in6); from6->sin6_addr = ip6->ip6_src; from6->sin6_port = sh->src_port; /* Get the scopes in properly to the sin6 addr's */ /* we probably don't need these operations */ (void)sa6_recoverscope(from6); sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone)); break; } #endif default: /* Currently not supported. */ return (NULL); } switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *to4; to4 = (struct sockaddr_in *)&to_store; bzero(to4, sizeof(*to4)); to4->sin_family = AF_INET; to4->sin_len = sizeof(struct sockaddr_in); to4->sin_addr.s_addr = iph->ip_dst.s_addr; to4->sin_port = sh->dest_port; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *to6; ip6 = mtod(m, struct ip6_hdr *); to6 = (struct sockaddr_in6 *)&to_store; bzero(to6, sizeof(*to6)); to6->sin6_family = AF_INET6; to6->sin6_len = sizeof(struct sockaddr_in6); to6->sin6_addr = ip6->ip6_dst; to6->sin6_port = sh->dest_port; /* Get the scopes in properly to the sin6 addr's */ /* we probably don't need these operations */ (void)sa6_recoverscope(to6); sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone)); break; } #endif default: /* TSNH */ break; } if (sh->v_tag) { /* we only go down this path if vtag is non-zero */ retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0); if (retval) { return (retval); } } find_tcp_pool = 0; if ((ch->chunk_type != SCTP_INITIATION) && (ch->chunk_type != SCTP_INITIATION_ACK) && (ch->chunk_type != SCTP_COOKIE_ACK) && (ch->chunk_type != SCTP_COOKIE_ECHO)) { /* Other chunk types go to the tcp pool. */ find_tcp_pool = 1; } if (inp_p) { retval = sctp_findassociation_addr_sa(to, from, inp_p, netp, find_tcp_pool, vrf_id); inp = *inp_p; } else { retval = sctp_findassociation_addr_sa(to, from, &inp, netp, find_tcp_pool, vrf_id); } SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp); if (retval == NULL && inp) { /* Found a EP but not this address */ if ((ch->chunk_type == SCTP_INITIATION) || (ch->chunk_type == SCTP_INITIATION_ACK)) { /*- * special hook, we do NOT return linp or an * association that is linked to an existing * association that is under the TCP pool (i.e. no * listener exists). The endpoint finding routine * will always find a listener before examining the * TCP pool. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { if (inp_p) { *inp_p = NULL; } return (NULL); } retval = sctp_findassociation_special_addr(m, iphlen, offset, sh, &inp, netp, to); if (inp_p != NULL) { *inp_p = inp; } } } SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval); return (retval); } /* * lookup an association by an ASCONF lookup address. * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup */ struct sctp_tcb * sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) { struct sctp_tcb *stcb; struct sockaddr_in *sin; #ifdef INET6 struct sockaddr_in6 *sin6; #endif struct sockaddr_storage local_store, remote_store; struct sockaddr *to; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6; #endif struct sctp_paramhdr parm_buf, *phdr; int ptype; int zero_address = 0; memset(&local_store, 0, sizeof(local_store)); memset(&remote_store, 0, sizeof(remote_store)); to = (struct sockaddr *)&local_store; /* First get the destination address setup too. */ iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: /* its IPv4 */ sin = (struct sockaddr_in *)&local_store; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = sh->dest_port; sin->sin_addr.s_addr = iph->ip_dst.s_addr; break; #ifdef INET6 case IPV6_VERSION >> 4: /* its IPv6 */ ip6 = mtod(m, struct ip6_hdr *); sin6 = (struct sockaddr_in6 *)&local_store; sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = sh->dest_port; sin6->sin6_addr = ip6->ip6_dst; break; #endif default: return NULL; } phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &parm_buf, sizeof(struct sctp_paramhdr)); if (phdr == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n", __FUNCTION__); return NULL; } ptype = (int)((uint32_t) ntohs(phdr->param_type)); /* get the correlation address */ switch (ptype) { #ifdef INET6 case SCTP_IPV6_ADDRESS: { /* ipv6 address param */ struct sctp_ipv6addr_param *p6, p6_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) { return NULL; } p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p6_buf.ph, sizeof(*p6)); if (p6 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n", __FUNCTION__); return (NULL); } sin6 = (struct sockaddr_in6 *)&remote_store; sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = sh->src_port; memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr)); if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) zero_address = 1; break; } #endif case SCTP_IPV4_ADDRESS: { /* ipv4 address param */ struct sctp_ipv4addr_param *p4, p4_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) { return NULL; } p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p4_buf.ph, sizeof(*p4)); if (p4 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n", __FUNCTION__); return (NULL); } sin = (struct sockaddr_in *)&remote_store; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = sh->src_port; memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr)); if (sin->sin_addr.s_addr == INADDR_ANY) zero_address = 1; break; } default: /* invalid address param type */ return NULL; } if (zero_address) { stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 1, vrf_id, 0); /* * printf("findassociation_ep_asconf: zero lookup address * finds stcb 0x%x\n", (uint32_t)stcb); */ } else { stcb = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&remote_store, netp, to, NULL); } return (stcb); } /* * allocate a sctp_inpcb and setup a temporary binding to a port/all * addresses. This way if we don't get a bind we by default pick a ephemeral * port with all addresses bound. */ int sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id) { /* * we get called when a new endpoint starts up. We need to allocate * the sctp_inpcb structure from the zone and init it. Mark it as * unbound and find a port that we can use as an ephemeral with * INADDR_ANY. If the user binds later no problem we can then add in * the specific addresses. And setup the default parameters for the * EP. */ int i, error; struct sctp_inpcb *inp; struct sctp_pcb *m; struct timeval time; sctp_sharedkey_t *null_key; error = 0; SCTP_INP_INFO_WLOCK(); inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb); if (inp == NULL) { SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n"); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); return (ENOBUFS); } /* zap it */ bzero(inp, sizeof(*inp)); /* bump generations */ /* setup socket pointers */ inp->sctp_socket = so; inp->ip_inp.inp.inp_socket = so; inp->sctp_associd_counter = 1; inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT; inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off); /* init the small hash table we use to track asocid <-> tcb */ inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark); if (inp->sctp_asocidhash == NULL) { SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_INP_INFO_WUNLOCK(); return (ENOBUFS); } #ifdef IPSEC { struct inpcbpolicy *pcb_sp = NULL; error = ipsec_init_policy(so, &pcb_sp); /* Arrange to share the policy */ inp->ip_inp.inp.inp_sp = pcb_sp; ((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp; } if (error != 0) { SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_INP_INFO_WUNLOCK(); return error; } #endif /* IPSEC */ SCTP_INCR_EP_COUNT(); inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl); SCTP_INP_INFO_WUNLOCK(); so->so_pcb = (caddr_t)inp; if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) || (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) { /* UDP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure it is NON-BLOCKING IO for UDP */ /* SCTP_SET_SO_NBIO(so); */ } else if (SCTP_SO_TYPE(so) == SOCK_STREAM) { /* TCP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure we have blocking IO by default */ SCTP_CLEAR_SO_NBIO(so); } else { /* * unsupported socket type (RAW, etc)- in case we missed it * in protosw */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP); so->so_pcb = NULL; SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (EOPNOTSUPP); } if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize), &inp->sctp_hashmark); if (inp->sctp_tcbhash == NULL) { SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); so->so_pcb = NULL; SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (ENOBUFS); } inp->def_vrf_id = vrf_id; SCTP_INP_INFO_WLOCK(); SCTP_INP_LOCK_INIT(inp); INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp"); SCTP_INP_READ_INIT(inp); SCTP_ASOC_CREATE_LOCK_INIT(inp); /* lock the new ep */ SCTP_INP_WLOCK(inp); /* add it to the info area */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list); SCTP_INP_INFO_WUNLOCK(); TAILQ_INIT(&inp->read_queue); LIST_INIT(&inp->sctp_addr_list); LIST_INIT(&inp->sctp_asoc_list); #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ LIST_INIT(&inp->sctp_asoc_free_list); #endif /* Init the timer structure for signature change */ SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer); inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE; /* now init the actual endpoint default data */ m = &inp->sctp_ep; /* setup the base timeout information */ m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default)); m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default)); m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default)); m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default)); m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default)); /* all max/min max are in ms */ m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default); m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default); m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default); m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default); m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default); m->max_open_streams_intome = MAX_SCTP_STREAMS; m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default); m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default); m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default); m->sctp_sws_sender = SCTP_SWS_SENDER_DEF; m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF; m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default); m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module); m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module); /* number of streams to pre-open on a association */ m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default); /* Add adaptation cookie */ m->adaptation_layer_indicator = 0x504C5253; /* seed random number generator */ m->random_counter = 1; m->store_at = SCTP_SIGNATURE_SIZE; SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers)); sctp_fill_random_store(m); /* Minimum cookie size */ m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) + sizeof(struct sctp_state_cookie); m->size_of_a_cookie += SCTP_SIGNATURE_SIZE; /* Setup the initial secret */ (void)SCTP_GETTIME_TIMEVAL(&time); m->time_of_secret_change = time.tv_sec; for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { m->secret_key[0][i] = sctp_select_initial_TSN(m); } sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); /* How long is a cookie good for ? */ m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default)); /* * Initialize authentication parameters */ m->local_hmacs = sctp_default_supported_hmaclist(); m->local_auth_chunks = sctp_alloc_chunklist(); sctp_auth_set_default_chunks(m->local_auth_chunks); LIST_INIT(&m->shared_keys); /* add default NULL key as key id 0 */ null_key = sctp_alloc_sharedkey(); sctp_insert_sharedkey(&m->shared_keys, null_key); SCTP_INP_WUNLOCK(inp); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 12); #endif return (error); } void sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, struct sctp_tcb *stcb) { struct sctp_nets *net; uint16_t lport, rport; struct sctppcbhead *head; struct sctp_laddr *laddr, *oladdr; atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(old_inp); SCTP_INP_WLOCK(new_inp); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); new_inp->sctp_ep.time_of_secret_change = old_inp->sctp_ep.time_of_secret_change; memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key, sizeof(old_inp->sctp_ep.secret_key)); new_inp->sctp_ep.current_secret_number = old_inp->sctp_ep.current_secret_number; new_inp->sctp_ep.last_secret_number = old_inp->sctp_ep.last_secret_number; new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie; /* make it so new data pours into the new socket */ stcb->sctp_socket = new_inp->sctp_socket; stcb->sctp_ep = new_inp; /* Copy the port across */ lport = new_inp->sctp_lport = old_inp->sctp_lport; rport = stcb->rport; /* Pull the tcb from the old association */ LIST_REMOVE(stcb, sctp_tcbhash); LIST_REMOVE(stcb, sctp_tcblist); if (stcb->asoc.in_asocid_hash) { LIST_REMOVE(stcb, sctp_tcbasocidhash); } /* Now insert the new_inp into the TCP connected hash */ head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; LIST_INSERT_HEAD(head, new_inp, sctp_hash); /* Its safe to access */ new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; /* Now move the tcb into the endpoint list */ LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist); /* * Question, do we even need to worry about the ep-hash since we * only have one connection? Probably not :> so lets get rid of it * and not suck up any kernel memory in that. */ if (stcb->asoc.in_asocid_hash) { struct sctpasochead *lhd; lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id, new_inp->hashasocidmark)]; LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash); } /* Ok. Let's restart timer. */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp, stcb, net); } SCTP_INP_INFO_WUNLOCK(); if (new_inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark); new_inp->sctp_tcbhash = NULL; } if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* Subset bound, so copy in the laddr list from the old_inp */ LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) { laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* * Gak, what can we do? This assoc is really * HOSED. We probably should send an abort * here. */ SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n"); continue; } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = oladdr->ifa; atomic_add_int(&laddr->ifa->refcount, 1); LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr, sctp_nxt_addr); new_inp->laddr_count++; } } /* * Now any running timers need to be adjusted since we really don't * care if they are running or not just blast in the new_inp into * all of them. */ stcb->asoc.hb_timer.ep = (void *)new_inp; stcb->asoc.dack_timer.ep = (void *)new_inp; stcb->asoc.asconf_timer.ep = (void *)new_inp; stcb->asoc.strreset_timer.ep = (void *)new_inp; stcb->asoc.shut_guard_timer.ep = (void *)new_inp; stcb->asoc.autoclose_timer.ep = (void *)new_inp; stcb->asoc.delayed_event_timer.ep = (void *)new_inp; stcb->asoc.delete_prim_timer.ep = (void *)new_inp; /* now what about the nets? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { net->pmtu_timer.ep = (void *)new_inp; net->rxt_timer.ep = (void *)new_inp; net->fr_timer.ep = (void *)new_inp; } SCTP_INP_WUNLOCK(new_inp); SCTP_INP_WUNLOCK(old_inp); } /* sctp_ifap is used to bypass normal local address validation checks */ int sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, struct sctp_ifa *sctp_ifap, struct thread *p) { /* bind a ep to a socket address */ struct sctppcbhead *head; struct sctp_inpcb *inp, *inp_tmp; struct inpcb *ip_inp; int port_reuse_active = 0; int bindall; uint16_t lport; int error; uint32_t vrf_id; lport = 0; error = 0; bindall = 1; inp = (struct sctp_inpcb *)so->so_pcb; ip_inp = (struct inpcb *)so->so_pcb; #ifdef SCTP_DEBUG if (addr) { SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n", ntohs(((struct sockaddr_in *)addr)->sin_port)); SCTPDBG(SCTP_DEBUG_PCB1, "Addr :"); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { /* already did a bind, subsequent binds NOT allowed ! */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } #ifdef INVARIANTS if (p == NULL) panic("null proc/thread"); #endif if (addr != NULL) { switch (addr->sa_family) { case AF_INET: { struct sockaddr_in *sin; /* IPV6_V6ONLY socket? */ if (SCTP_IPV6_V6ONLY(ip_inp)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } if (addr->sa_len != sizeof(*sin)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } sin = (struct sockaddr_in *)addr; lport = sin->sin_port; /* * For LOOPBACK the prison_local_ip4() call * will transmute the ip address to the * proper value. */ if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } if (sin->sin_addr.s_addr != INADDR_ANY) { bindall = 0; } break; } #ifdef INET6 case AF_INET6: { /* * Only for pure IPv6 Address. (No IPv4 * Mapped!) */ struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (addr->sa_len != sizeof(*sin6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } lport = sin6->sin6_port; /* * For LOOPBACK the prison_local_ip6() call * will transmute the ipv6 address to the * proper value. */ if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr, (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { bindall = 0; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } /* this must be cleared for ifa_ifwithaddr() */ sin6->sin6_scope_id = 0; break; } #endif default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT); return (EAFNOSUPPORT); } } SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); /* Setup a vrf_id to be the default for the non-bind-all case. */ vrf_id = inp->def_vrf_id; /* increase our count due to the unlock we do */ SCTP_INP_INCR_REF(inp); if (lport) { /* * Did the caller specify a port? if so we must see if a ep * already has this one bound. */ /* got to be root to get at low ports */ if (ntohs(lport) < IPPORT_RESERVED) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } } if (p == NULL) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } SCTP_INP_WUNLOCK(inp); if (bindall) { vrf_id = inp->def_vrf_id; inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); /* unlock info */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { /* * Ok, must be one-2-one and * allowing port re-use */ port_reuse_active = 1; goto continue_anyway; } SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } else { inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); /* unlock info */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { /* * Ok, must be one-2-one and * allowing port re-use */ port_reuse_active = 1; goto continue_anyway; } SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } continue_anyway: SCTP_INP_WLOCK(inp); if (bindall) { /* verify that no lport is not used by a singleton */ if ((port_reuse_active == 0) && (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id)) ) { /* Sorry someone already has this one bound */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { port_reuse_active = 1; } else { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } } } else { uint16_t first, last, candidate; uint16_t count; int done; if (ip_inp->inp_flags & INP_HIGHPORT) { first = MODULE_GLOBAL(ipport_hifirstauto); last = MODULE_GLOBAL(ipport_hilastauto); } else if (ip_inp->inp_flags & INP_LOWPORT) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } first = MODULE_GLOBAL(ipport_lowfirstauto); last = MODULE_GLOBAL(ipport_lowlastauto); } else { first = MODULE_GLOBAL(ipport_firstauto); last = MODULE_GLOBAL(ipport_lastauto); } if (first > last) { uint16_t temp; temp = first; first = last; last = temp; } count = last - first + 1; /* number of candidates */ candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count); done = 0; while (!done) { if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) { done = 1; } if (!done) { if (--count == 0) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } if (candidate == last) candidate = first; else candidate = candidate + 1; } } lport = htons(candidate); } SCTP_INP_DECR_REF(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* * this really should not happen. The guy did a non-blocking * bind and then did a close at the same time. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } /* ok we look clear to give out this port, so lets setup the binding */ if (bindall) { /* binding to all addresses, so just set in the proper flags */ inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL; /* set the automatic addr changes from kernel flag */ if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } /* * set the automatic mobility_base from kernel flag (by * micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } /* * set the automatic mobility_fasthandoff from kernel flag * (by micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } } else { /* * bind specific, make sure flags is off and add a new * address structure to the sctp_addr_list inside the ep * structure. * * We will need to allocate one and insert it at the head. The * socketopt call can just insert new addresses in there as * well. It will also have to do the embed scope kame hack * too (before adding). */ struct sctp_ifa *ifa; struct sockaddr_storage store_sa; memset(&store_sa, 0, sizeof(store_sa)); if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&store_sa; memcpy(sin, addr, sizeof(struct sockaddr_in)); sin->sin_port = 0; } else if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&store_sa; memcpy(sin6, addr, sizeof(struct sockaddr_in6)); sin6->sin6_port = 0; } /* * first find the interface with the bound address need to * zero out the port to find the address! yuck! can't do * this earlier since need port for sctp_pcb_findep() */ if (sctp_ifap != NULL) ifa = sctp_ifap; else { /* * Note for BSD we hit here always other O/S's will * pass things in via the sctp_ifap argument * (Panda). */ ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa, vrf_id, SCTP_ADDR_NOT_LOCKED); } if (ifa == NULL) { /* Can't find an interface with that address */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL); return (EADDRNOTAVAIL); } if (addr->sa_family == AF_INET6) { /* GAK, more FIXME IFA lock? */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } /* we're not bound all */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL; /* allow bindx() to send ASCONF's for binding changes */ sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); /* clear automatic addr changes from kernel flag */ sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); /* add this address to the endpoint list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0); if (error != 0) { SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } inp->laddr_count++; } /* find the bucket */ if (port_reuse_active) { /* Put it into tcp 1-2-1 hash */ head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))]; inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; } else { head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; } /* put it in the bucket */ LIST_INSERT_HEAD(head, inp, sctp_hash); SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n", head, ntohs(lport), port_reuse_active); /* set in the port */ inp->sctp_lport = lport; /* turn off just the unbound flag */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (0); } static void sctp_iterator_inp_being_freed(struct sctp_inpcb *inp) { struct sctp_iterator *it, *nit; /* * We enter with the only the ITERATOR_LOCK in place and a write * lock on the inp_info stuff. */ it = sctp_it_ctl.cur_it; if (it && (it->vn != curvnet)) { /* Its not looking at our VNET */ return; } if (it && (it->inp == inp)) { /* * This is tricky and we hold the iterator lock, but when it * returns and gets the lock (when we release it) the * iterator will try to operate on inp. We need to stop that * from happening. But of course the iterator has a * reference on the stcb and inp. We can mark it and it will * stop. * * If its a single iterator situation, we set the end iterator * flag. Otherwise we set the iterator to go to the next * inp. * */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; } else { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP; } } /* * Now go through and remove any single reference to our inp that * may be still pending on the list */ SCTP_IPI_ITERATOR_WQ_LOCK(); TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { if (it->vn != curvnet) { continue; } if (it->inp == inp) { /* This one points to me is it inp specific? */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { /* Remove and free this one */ TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); } else { it->inp = LIST_NEXT(it->inp, sctp_list); if (it->inp) { SCTP_INP_INCR_REF(it->inp); } } /* * When its put in the refcnt is incremented so decr * it */ SCTP_INP_DECR_REF(inp); } } SCTP_IPI_ITERATOR_WQ_UNLOCK(); } /* release sctp_inpcb unbind the port */ void sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from) { /* * Here we free a endpoint. We must find it (if it is in the Hash * table) and remove it from there. Then we must also find it in the * overall list and remove it from there. After all removals are * complete then any timer has to be stopped. Then start the actual * freeing. a) Any local lists. b) Any associations. c) The hash of * all associations. d) finally the ep itself. */ struct sctp_pcb *m; struct sctp_tcb *asoc, *nasoc; struct sctp_laddr *laddr, *nladdr; struct inpcb *ip_pcb; struct socket *so; int being_refed = 0; struct sctp_queued_to_read *sq, *nsq; int cnt; sctp_sharedkey_t *shared_key, *nshared_key; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 0); #endif SCTP_ITERATOR_LOCK(); /* mark any iterators on the list or being processed */ sctp_iterator_inp_being_freed(inp); SCTP_ITERATOR_UNLOCK(); so = inp->sctp_socket; if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* been here before.. eeks.. get out of here */ SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 1); #endif return; } SCTP_ASOC_CREATE_LOCK(inp); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP; /* socket is gone, so no more wakeups allowed */ inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; } /* First time through we have the socket lock, after that no more. */ sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL, SCTP_FROM_SCTP_PCB + SCTP_LOC_1); if (inp->control) { sctp_m_freem(inp->control); inp->control = NULL; } if (inp->pkt) { sctp_m_freem(inp->pkt); inp->pkt = NULL; } m = &inp->sctp_ep; ip_pcb = &inp->ip_inp.inp; /* we could just cast the main pointer * here but I will be nice :> (i.e. * ip_pcb = ep;) */ if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { int cnt_in_sd; cnt_in_sd = 0; LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { SCTP_TCB_LOCK(asoc); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* Skip guys being freed */ cnt_in_sd++; if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { /* * Special case - we did not start a * kill timer on the asoc due to it * was not closed. So go ahead and * start it now. */ asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); } SCTP_TCB_UNLOCK(asoc); continue; } if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) && (asoc->asoc.total_output_queue_size == 0)) { /* * If we have data in queue, we don't want * to just free since the app may have done, * send()/close or connect/send/close. And * it wants the data to get across first. */ /* Just abandon things in the front states */ if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) { cnt_in_sd++; } continue; } /* Disconnect the socket please */ asoc->sctp_socket = NULL; asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET; if ((asoc->asoc.size_on_reasm_queue > 0) || (asoc->asoc.control_pdapi) || (asoc->asoc.size_on_all_streams > 0) || (so && (so->so_rcv.sb_cc > 0)) ) { /* Left with Data unread */ struct mbuf *op_err; op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* Fill in the user initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) { cnt_in_sd++; } continue; } else if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.stream_queue_cnt == 0) ) { if (asoc->asoc.locked_on_sending) { goto abort_anyway; } if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* * there is nothing queued to send, * so I send shutdown */ sctp_send_shutdown(asoc, asoc->asoc.primary_destination); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED); } } else { /* mark into shutdown pending */ struct sctp_stream_queue_pending *sp; asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); if (asoc->asoc.locked_on_sending) { sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue), sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n", asoc->asoc.locked_on_sending, asoc->asoc.locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* * Fill in the user * initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) { cnt_in_sd++; } continue; } else { sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } cnt_in_sd++; SCTP_TCB_UNLOCK(asoc); } /* now is there some left in our SHUTDOWN state? */ if (cnt_in_sd) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 2); #endif inp->sctp_socket = NULL; SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } } inp->sctp_socket = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != SCTP_PCB_FLAGS_UNBOUND) { /* * ok, this guy has been bound. It's port is somewhere in * the SCTP_BASE_INFO(hash table). Remove it! */ LIST_REMOVE(inp, sctp_hash); inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND; } /* * If there is a timer running to kill us, forget it, since it may * have a contest on the INP lock.. which would cause us to die ... */ cnt = 0; LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { SCTP_TCB_LOCK(asoc); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); } cnt++; SCTP_TCB_UNLOCK(asoc); continue; } /* Free associations that are NOT killing us */ if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) && ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) { struct mbuf *op_err; uint32_t *ippp; op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* Fill in the user initiated abort */ struct sctp_paramhdr *ph; SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); } else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { cnt++; SCTP_TCB_UNLOCK(asoc); continue; } if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) { cnt++; } } if (cnt) { /* Ok we have someone out there that will kill us */ (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 3); #endif SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } if (SCTP_INP_LOCK_CONTENDED(inp)) being_refed++; if (SCTP_INP_READ_CONTENDED(inp)) being_refed++; if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp)) being_refed++; if ((inp->refcount) || (being_refed) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) { (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 4); #endif sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } inp->sctp_ep.signature_change.type = 0; inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE; /* * Remove it from the list .. last thing we need a lock for. */ LIST_REMOVE(inp, sctp_list); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); /* * Now we release all locks. Since this INP cannot be found anymore * except possibly by the kill timer that might be running. We call * the drain function here. It should hit the case were it sees the * ACTIVE flag cleared and exit out freeing us to proceed and * destroy everything. */ if (from != SCTP_CALLED_FROM_INPKILL_TIMER) { (void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer); } else { /* Probably un-needed */ (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 5); #endif if ((inp->sctp_asocidhash) != NULL) { SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark); inp->sctp_asocidhash = NULL; } /* sa_ignore FREED_MEMORY */ TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) { /* Its only abandoned if it had data left */ if (sq->length) SCTP_STAT_INCR(sctps_left_abandon); TAILQ_REMOVE(&inp->read_queue, sq, next); sctp_free_remote_addr(sq->whoFrom); if (so) so->so_rcv.sb_cc -= sq->length; if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } /* * no need to free the net count, since at this point all * assoc's are gone. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); } /* Now the sctp_pcb things */ /* * free each asoc if it is not already closed/free. we can't use the * macro here since le_next will get freed as part of the * sctp_free_assoc() call. */ cnt = 0; if (so) { #ifdef IPSEC ipsec_delete_pcbpolicy(ip_pcb); #endif /* IPSEC */ /* Unlocks not needed since the socket is gone now */ } if (ip_pcb->inp_options) { (void)sctp_m_free(ip_pcb->inp_options); ip_pcb->inp_options = 0; } if (ip_pcb->inp_moptions) { inp_freemoptions(ip_pcb->inp_moptions); ip_pcb->inp_moptions = 0; } #ifdef INET6 if (ip_pcb->inp_vflag & INP_IPV6) { struct in6pcb *in6p; in6p = (struct in6pcb *)inp; ip6_freepcbopts(in6p->in6p_outputopts); } #endif /* INET6 */ ip_pcb->inp_vflag = 0; /* free up authentication fields */ if (inp->sctp_ep.local_auth_chunks != NULL) sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); if (inp->sctp_ep.local_hmacs != NULL) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ } /* * if we have an address list the following will free the list of * ifaddr's that are set into this ep. Again macro limitations here, * since the LIST_FOREACH could be a bad idea. */ LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { sctp_remove_laddr(laddr); } #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) { LIST_REMOVE(asoc, sctp_tcblist); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc); SCTP_DECR_ASOC_COUNT(); } /* *** END TEMP CODE *** */ #endif /* Now lets see about freeing the EP hash table. */ if (inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); inp->sctp_tcbhash = NULL; } /* Now we must put the ep memory back into the zone pool */ INP_LOCK_DESTROY(&inp->ip_inp.inp); SCTP_INP_LOCK_DESTROY(inp); SCTP_INP_READ_DESTROY(inp); SCTP_ASOC_CREATE_LOCK_DESTROY(inp); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_DECR_EP_COUNT(); } struct sctp_nets * sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr) { struct sctp_nets *net; /* locate the address */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr)) return (net); } return (NULL); } int sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id) { struct sctp_ifa *sctp_ifa; sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED); if (sctp_ifa) { return (1); } else { return (0); } } /* * add's a remote endpoint address, done with the INIT/INIT-ACK as well as * when a ASCONF arrives that adds it. It will also initialize all the cwnd * stats of stuff. */ int sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr, int set_scope, int from) { /* * The following is redundant to the same lines in the * sctp_aloc_assoc() but is needed since others call the add address * function */ struct sctp_nets *net, *netfirst; int addr_inscope; SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ", from); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr); netfirst = sctp_findnet(stcb, newaddr); if (netfirst) { /* * Lie and return ok, we don't want to make the association * go away for this behavior. It will happen in the TCP * model in a connected socket. It does not reach the hash * table until after the association is built so it can't be * found. Mark as reachable, since the initial creation will * have been cleared and the NOT_IN_ASSOC flag will have * been added... and we don't want to end up removing it * back out. */ if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) { netfirst->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED); } else { netfirst->dest_state = SCTP_ADDR_REACHABLE; } return (0); } addr_inscope = 1; if (newaddr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)newaddr; if (sin->sin_addr.s_addr == 0) { /* Invalid address */ return (-1); } /* zero out the bzero area */ memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); /* assure len is set */ sin->sin_len = sizeof(struct sockaddr_in); if (set_scope) { #ifdef SCTP_DONT_DO_PRIVADDR_SCOPE stcb->ipv4_local_scope = 1; #else if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { stcb->asoc.ipv4_local_scope = 1; } #endif /* SCTP_DONT_DO_PRIVADDR_SCOPE */ } else { /* Validate the address is in scope */ if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) && (stcb->asoc.ipv4_local_scope == 0)) { addr_inscope = 0; } } #ifdef INET6 } else if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)newaddr; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* Invalid address */ return (-1); } /* assure len is set */ sin6->sin6_len = sizeof(struct sockaddr_in6); if (set_scope) { if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) { stcb->asoc.loopback_scope = 1; stcb->asoc.local_scope = 0; stcb->asoc.ipv4_local_scope = 1; stcb->asoc.site_scope = 1; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { /* * If the new destination is a LINK_LOCAL we * must have common site scope. Don't set * the local scope since we may not share * all links, only loopback can do this. * Links on the local network would also be * on our private network for v4 too. */ stcb->asoc.ipv4_local_scope = 1; stcb->asoc.site_scope = 1; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { /* * If the new destination is SITE_LOCAL then * we must have site scope in common. */ stcb->asoc.site_scope = 1; } } else { /* Validate the address is in scope */ if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) && (stcb->asoc.loopback_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && (stcb->asoc.local_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && (stcb->asoc.site_scope == 0)) { addr_inscope = 0; } } #endif } else { /* not supported family type */ return (-1); } net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets); if (net == NULL) { return (-1); } SCTP_INCR_RADDR_COUNT(); bzero(net, sizeof(*net)); (void)SCTP_GETTIME_TIMEVAL(&net->start_time); memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len); if (newaddr->sa_family == AF_INET) { ((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport; } else if (newaddr->sa_family == AF_INET6) { ((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport; } net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id); if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) { stcb->asoc.loopback_scope = 1; stcb->asoc.ipv4_local_scope = 1; stcb->asoc.local_scope = 0; stcb->asoc.site_scope = 1; addr_inscope = 1; } net->failure_threshold = stcb->asoc.def_net_failure; if (addr_inscope == 0) { net->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_OUT_OF_SCOPE); } else { if (from == SCTP_ADDR_IS_CONFIRMED) /* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */ net->dest_state = SCTP_ADDR_REACHABLE; else net->dest_state = SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED; } /* * We set this to 0, the timer code knows that this means its an * initial value */ net->RTO = 0; net->RTO_measured = 0; stcb->asoc.numnets++; *(&net->ref_count) = 1; net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1; net->tos_flowlabel = 0; if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) { net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); } else { net->port = 0; } #ifdef INET if (newaddr->sa_family == AF_INET) net->tos_flowlabel = stcb->asoc.default_tos; #endif #ifdef INET6 if (newaddr->sa_family == AF_INET6) net->tos_flowlabel = stcb->asoc.default_flowlabel; #endif /* Init the timer structure */ SCTP_OS_TIMER_INIT(&net->rxt_timer.timer); SCTP_OS_TIMER_INIT(&net->fr_timer.timer); SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer); /* Now generate a route for this guy */ #ifdef INET6 /* KAME hack: embed scopeid */ if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); sin6->sin6_scope_id = 0; } #endif SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id); if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) { /* Get source address */ net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep, stcb, (sctp_route_t *) & net->ro, net, 0, stcb->asoc.vrf_id); /* Now get the interface MTU */ if (net->ro._s_addr && net->ro._s_addr->ifn_p) { net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p); } else { net->mtu = 0; } if (net->mtu == 0) { /* Huh ?? */ net->mtu = SCTP_DEFAULT_MTU; } else { uint32_t rmtu; rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt); if (rmtu == 0) { /* * Start things off to match mtu of * interface please. */ SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa, net->ro.ro_rt, net->mtu); } else { /* * we take the route mtu over the interface, * since the route may be leading out the * loopback, or a different interface. */ net->mtu = rmtu; } } if (from == SCTP_ALLOC_ASOC) { stcb->asoc.smallest_mtu = net->mtu; } } else { net->mtu = stcb->asoc.smallest_mtu; } #ifdef INET6 if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_recoverscope(sin6); } #endif if (net->port) { net->mtu -= sizeof(struct udphdr); } if (stcb->asoc.smallest_mtu > net->mtu) { stcb->asoc.smallest_mtu = net->mtu; } /* JRS - Use the congestion control given in the CC module */ stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); /* * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning * of assoc (2005/06/27, iyengar@cis.udel.edu) */ net->find_pseudo_cumack = 1; net->find_rtx_pseudo_cumack = 1; net->src_addr_selected = 0; netfirst = TAILQ_FIRST(&stcb->asoc.nets); if (net->ro.ro_rt == NULL) { /* Since we have no route put it at the back */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); } else if (netfirst == NULL) { /* We are the first one in the pool. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (netfirst->ro.ro_rt == NULL) { /* * First one has NO route. Place this one ahead of the first * one. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) { /* * This one has a different interface than the one at the * top of the list. Place it ahead. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else { /* * Ok we have the same interface as the first one. Move * forward until we find either a) one with a NULL route... * insert ahead of that b) one with a different ifp.. insert * after that. c) end of the list.. insert at the tail. */ struct sctp_nets *netlook; do { netlook = TAILQ_NEXT(netfirst, sctp_next); if (netlook == NULL) { /* End of the list */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); break; } else if (netlook->ro.ro_rt == NULL) { /* next one has NO route */ TAILQ_INSERT_BEFORE(netfirst, net, sctp_next); break; } else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) { TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook, net, sctp_next); break; } /* Shift forward */ netfirst = netlook; } while (netlook != NULL); } /* got to have a primary set */ if (stcb->asoc.primary_destination == 0) { stcb->asoc.primary_destination = net; } else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) && (net->ro.ro_rt) && ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { /* No route to current primary adopt new primary */ stcb->asoc.primary_destination = net; } sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb, net); /* Validate primary is first */ net = TAILQ_FIRST(&stcb->asoc.nets); if ((net != stcb->asoc.primary_destination) && (stcb->asoc.primary_destination)) { /* * first one on the list is NOT the primary sctp_cmpaddr() * is much more efficient if the primary is the first on the * list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } static uint32_t sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { uint32_t id; struct sctpasochead *head; struct sctp_tcb *lstcb; SCTP_INP_WLOCK(inp); try_again: if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* TSNH */ SCTP_INP_WUNLOCK(inp); return (0); } /* * We don't allow assoc id to be 0, this is needed otherwise if the * id were to wrap we would have issues with some socket options. */ if (inp->sctp_associd_counter == 0) { inp->sctp_associd_counter++; } id = inp->sctp_associd_counter; inp->sctp_associd_counter++; lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0); if (lstcb) { goto try_again; } head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash); stcb->asoc.in_asocid_hash = 1; SCTP_INP_WUNLOCK(inp); return id; } /* * allocate an association and add it to the endpoint. The caller must be * careful to add all additional addresses once they are know right away or * else the assoc will be may experience a blackout scenario. */ struct sctp_tcb * sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr, int *error, uint32_t override_tag, uint32_t vrf_id, struct thread *p ) { /* note the p argument is only valid in unbound sockets */ struct sctp_tcb *stcb; struct sctp_association *asoc; struct sctpasochead *head; uint16_t rport; int err; /* * Assumption made here: Caller has done a * sctp_findassociation_ep_addr(ep, addr's); to make sure the * address does not exist already. */ if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) { /* Hit max assoc, sorry no more */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } if (firstaddr == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RLOCK(inp); if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) || (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { /* * If its in the TCP pool, its NOT allowed to create an * association. The parent listener needs to call * sctp_aloc_assoc.. or the one-2-many socket. If a peeled * off, or connected one does this.. its an error. */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:"); #ifdef SCTP_DEBUG if (firstaddr) { SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr); SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", ntohs(((struct sockaddr_in *)firstaddr)->sin_port)); } else { SCTPDBG(SCTP_DEBUG_PCB3, "None\n"); } #endif /* SCTP_DEBUG */ if (firstaddr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)firstaddr; if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin->sin_port; } else if (firstaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)firstaddr; if ((sin6->sin6_port == 0) || (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin6->sin6_port; } else { /* not supported family type */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RUNLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { /* * If you have not performed a bind, then we need to do the * ephemeral bind for you. */ if ((err = sctp_inpcb_bind(inp->sctp_socket, (struct sockaddr *)NULL, (struct sctp_ifa *)NULL, p ))) { /* bind error, probably perm */ *error = err; return (NULL); } } stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb); if (stcb == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); *error = ENOMEM; return (NULL); } SCTP_INCR_ASOC_COUNT(); bzero(stcb, sizeof(*stcb)); asoc = &stcb->asoc; asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb); SCTP_TCB_LOCK_INIT(stcb); SCTP_TCB_SEND_LOCK_INIT(stcb); stcb->rport = rport; /* setup back pointer's */ stcb->sctp_ep = inp; stcb->sctp_socket = inp->sctp_socket; if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) { /* failed */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); *error = err; return (NULL); } /* and the port */ SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* inpcb freed while alloc going on */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_DECR_ASOC_COUNT(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_TCB_LOCK(stcb); /* now that my_vtag is set, add it to the hash */ head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* put it in the bucket in the vtag hash of assoc's for the system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); SCTP_INP_INFO_WUNLOCK(); if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) { /* failure.. memory error? */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } if (asoc->nr_mapping_array) { SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); asoc->nr_mapping_array = NULL; } SCTP_DECR_ASOC_COUNT(); SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } /* Init all the timers */ SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer); SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer); SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer); SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer); SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer); SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer); LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist); /* now file the port under the hash as well */ if (inp->sctp_tcbhash != NULL) { head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport, inp->sctp_hashmark)]; LIST_INSERT_HEAD(head, stcb, sctp_tcbhash); } SCTP_INP_WUNLOCK(inp); SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb); return (stcb); } void sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; asoc = &stcb->asoc; asoc->numnets--; TAILQ_REMOVE(&asoc->nets, net, sctp_next); if (net == asoc->primary_destination) { /* Reset primary */ struct sctp_nets *lnet; lnet = TAILQ_FIRST(&asoc->nets); /* * Mobility adaptation Ideally, if deleted destination is * the primary, it becomes a fast retransmission trigger by * the subsequent SET PRIMARY. (by micchie) */ if (sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE) || sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_FASTHANDOFF)) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n"); if (asoc->deleted_primary != NULL) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n"); goto out; } asoc->deleted_primary = net; atomic_add_int(&net->ref_count, 1); memset(&net->lastsa, 0, sizeof(net->lastsa)); memset(&net->lastsv, 0, sizeof(net->lastsv)); sctp_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_PRIM_DELETED); sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED, stcb->sctp_ep, stcb, NULL); } out: /* Try to find a confirmed primary */ asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0); } if (net == asoc->last_data_chunk_from) { /* Reset primary */ asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets); } if (net == asoc->last_control_chunk_from) { /* Clear net */ asoc->last_control_chunk_from = NULL; } sctp_free_remote_addr(net); } /* * remove a remote endpoint address from an association, it will fail if the * address does not exist. */ int sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr) { /* * Here we need to remove a remote address. This is quite simple, we * first find it in the list of address for the association * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE * on that item. Note we do not allow it to be removed if there are * no other addresses. */ struct sctp_association *asoc; struct sctp_nets *net, *nnet; asoc = &stcb->asoc; /* locate the address */ TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) { continue; } if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr, remaddr)) { /* we found the guy */ if (asoc->numnets < 2) { /* Must have at LEAST two remote addresses */ return (-1); } else { sctp_remove_net(stcb, net); return (0); } } } /* not found. */ return (-2); } void sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; if (!LIST_EMPTY(chain)) { LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; found = 1; break; } } if (found) break; } } } int sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; SCTP_INP_INFO_WLOCK(); chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; if (!LIST_EMPTY(chain)) { LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { found = 1; break; } } if (found) break; } } SCTP_INP_INFO_WUNLOCK(); return (found); } void sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct timeval now; int set, i; if (time == 0) { /* Its disabled */ return; } (void)SCTP_GETTIME_TIMEVAL(&now); chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; set = 0; if (!LIST_EMPTY(chain)) { /* Block(s) present, lets find space, and expire on the fly */ LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == 0) && !set) { twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[i].v_tag = tag; twait_block->vtag_block[i].lport = lport; twait_block->vtag_block[i].rport = rport; set = 1; } else if ((twait_block->vtag_block[i].v_tag) && ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; if (set == 0) { /* Reuse it for my new tag */ twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[i].v_tag = tag; twait_block->vtag_block[i].lport = lport; twait_block->vtag_block[i].rport = rport; set = 1; } } } if (set) { /* * We only do up to the block where we can * place our tag for audits */ break; } } } /* Need to add a new block to chain */ if (!set) { SCTP_MALLOC(twait_block, struct sctp_tagblock *, sizeof(struct sctp_tagblock), SCTP_M_TIMW); if (twait_block == NULL) { #ifdef INVARIANTS panic("Can not alloc tagblock"); #endif return; } memset(twait_block, 0, sizeof(struct sctp_tagblock)); LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock); twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[0].v_tag = tag; twait_block->vtag_block[0].lport = lport; twait_block->vtag_block[0].rport = rport; } } /*- * Free the association after un-hashing the remote port. This * function ALWAYS returns holding NO LOCK on the stcb. It DOES * expect that the input to this function IS a locked TCB. * It will return 0, if it did NOT destroy the association (instead * it unlocks it. It will return NON-zero if it either destroyed the * association OR the association is already destroyed. */ int sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location) { int i; struct sctp_association *asoc; struct sctp_nets *net, *nnet; struct sctp_laddr *laddr, *naddr; struct sctp_tmit_chunk *chk, *nchk; struct sctp_asconf_addr *aparam, *naparam; struct sctp_asconf_ack *aack, *naack; struct sctp_stream_reset_list *strrst, *nstrrst; struct sctp_queued_to_read *sq, *nsq; struct sctp_stream_queue_pending *sp, *nsp; sctp_sharedkey_t *shared_key, *nshared_key; struct socket *so; /* first, lets purge the entry from the hash table. */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 6); #endif if (stcb->asoc.state == 0) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 7); #endif /* there is no asoc, really TSNH :-0 */ return (1); } /* TEMP CODE */ if (stcb->freed_from_where == 0) { /* Only record the first place free happened from */ stcb->freed_from_where = from_location; } /* TEMP CODE */ asoc = &stcb->asoc; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; else so = inp->sctp_socket; /* * We used timer based freeing if a reader or writer is in the way. * So we first check if we are actually being called from a timer, * if so we abort early if a reader or writer is still in the way. */ if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && (from_inpcbfree == SCTP_NORMAL_PROC)) { /* * is it the timer driving us? if so are the reader/writers * gone? */ if (stcb->asoc.refcnt) { /* nope, reader or writer in the way */ sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); /* no asoc destroyed */ SCTP_TCB_UNLOCK(stcb); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 8); #endif return (0); } } /* now clean up any other timers */ (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); asoc->hb_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); asoc->dack_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); /*- * For stream reset we don't blast this unless * it is a str-reset timer, it might be the * free-asoc timer which we DON'T want to * disturb. */ if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET) asoc->strreset_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); asoc->asconf_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); asoc->autoclose_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); asoc->shut_guard_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); asoc->delayed_event_timer.self = NULL; /* Mobility adaptation */ (void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer); asoc->delete_prim_timer.self = NULL; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); net->fr_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); net->rxt_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); net->pmtu_timer.self = NULL; } /* Now the read queue needs to be cleaned up (only once) */ if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) { stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED; SCTP_INP_READ_LOCK(inp); TAILQ_FOREACH(sq, &inp->read_queue, next) { if (sq->stcb == stcb) { sq->do_not_ref_stcb = 1; sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn; /* * If there is no end, there never will be * now. */ if (sq->end_added == 0) { /* Held for PD-API clear that. */ sq->pdapi_aborted = 1; sq->held_length = 0; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) { /* * Need to add a PD-API * aborted indication. * Setting the control_pdapi * assures that it will be * added right after this * msg. */ uint32_t strseq; stcb->asoc.control_pdapi = sq; strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn; sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, stcb, SCTP_PARTIAL_DELIVERY_ABORTED, (void *)&strseq, SCTP_SO_LOCKED); stcb->asoc.control_pdapi = NULL; } } /* Add an end to wake them */ sq->end_added = 1; } } SCTP_INP_READ_UNLOCK(inp); if (stcb->block_entry) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET); stcb->block_entry->error = ECONNRESET; stcb->block_entry = NULL; } } if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) { /* * Someone holds a reference OR the socket is unaccepted * yet. */ if ((stcb->asoc.refcnt) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); } SCTP_TCB_UNLOCK(stcb); if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if (so) { /* Wake any reader/writers */ sctp_sorwakeup(inp, so); sctp_sowwakeup(inp, so); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 9); #endif /* no asoc destroyed */ return (0); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 10); #endif /* * When I reach here, no others want to kill the assoc yet.. and I * own the lock. Now its possible an abort comes in when I do the * lock exchange below to grab all the locks to do the final take * out. to prevent this we increment the count, which will start a * timer and blow out above thus assuring us that we hold exclusive * killing of the asoc. Note that after getting back the TCB lock we * will go ahead and increment the counter back up and stop any * timer a passing stranger may have started :-S */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); SCTP_TCB_LOCK(stcb); } /* Double check the GONE flag */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * For TCP type we need special handling when we are * connected. We also include the peel'ed off ones to. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED; inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED; if (so) { SOCK_LOCK(so); if (so->so_rcv.sb_cc == 0) { so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); } socantrcvmore_locked(so); sctp_sowwakeup(inp, so); sctp_sorwakeup(inp, so); SCTP_SOWAKEUP(so); } } } /* * Make it invalid too, that way if its about to run it will abort * and return. */ /* re-increment the lock */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, -1); } if (stcb->asoc.refcnt) { stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INFO_WUNLOCK(); SCTP_INP_WUNLOCK(inp); } SCTP_TCB_UNLOCK(stcb); return (0); } asoc->state = 0; if (inp->sctp_tcbhash) { LIST_REMOVE(stcb, sctp_tcbhash); } if (stcb->asoc.in_asocid_hash) { LIST_REMOVE(stcb, sctp_tcbasocidhash); } /* Now lets remove it from the list of ALL associations in the EP */ LIST_REMOVE(stcb, sctp_tcblist); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); } /* pull from vtag hash */ LIST_REMOVE(stcb, sctp_asocs); sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait), inp->sctp_lport, stcb->rport); /* * Now restop the timers to be sure this is paranoia at is finest! */ (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); } asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE; /* * The chunk lists and such SHOULD be empty but we check them just * in case. */ /* anything on the wheel needs to be removed */ for (i = 0; i < asoc->streamoutcnt; i++) { struct sctp_stream_out *outs; outs = &asoc->strmout[i]; /* now clean up any chunks here */ TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { TAILQ_REMOVE(&outs->outqueue, sp, next); if (sp->data) { if (so) { /* Still an open socket - report */ sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, (void *)sp, SCTP_SO_LOCKED); } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; sp->tail_mbuf = NULL; } } if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } sctp_free_spbufspace(stcb, asoc, sp); if (sp->holds_key_ref) sctp_auth_key_release(stcb, sp->auth_keyid); /* Free the zone stuff */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp); SCTP_DECR_STRMOQ_COUNT(); /* sa_ignore FREED_MEMORY */ } } /* sa_ignore FREED_MEMORY */ TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) { TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp); SCTP_FREE(strrst, SCTP_M_STRESET); } TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) { TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } sctp_free_remote_addr(sq->whoFrom); sq->whoFrom = NULL; sq->stcb = NULL; /* Free the ctl entry */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); /* sa_ignore FREED_MEMORY */ } TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) { TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); asoc->free_chunk_cnt--; /* sa_ignore FREED_MEMORY */ } /* pending send queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); if (chk->data) { if (so) { /* Still a socket? */ sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, chk, SCTP_SO_LOCKED); } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); if (chk->whoTo) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = NULL; } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } /* sent queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); if (chk->data) { if (so) { /* Still a socket? */ sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_SENT, chk, SCTP_SO_LOCKED); } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } /* control queue MAY not be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } /* ASCONF queue MAY not be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } if (asoc->nr_mapping_array) { SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); asoc->nr_mapping_array = NULL; } /* the stream outs */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } asoc->strm_realoutsize = asoc->streamoutcnt = 0; if (asoc->strmin) { struct sctp_queued_to_read *ctl, *nctl; for (i = 0; i < asoc->streamincnt; i++) { TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) { TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next); sctp_free_remote_addr(ctl->whoFrom); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } /* * We don't free the address here since all * the net's were freed above. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); } } SCTP_FREE(asoc->strmin, SCTP_M_STRMI); asoc->strmin = NULL; } asoc->streamincnt = 0; TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { #ifdef INVARIANTS if (SCTP_BASE_INFO(ipi_count_raddr) == 0) { panic("no net's left alloc'ed, or list points to itself"); } #endif TAILQ_REMOVE(&asoc->nets, net, sctp_next); sctp_free_remote_addr(net); } LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) { /* sa_ignore FREED_MEMORY */ sctp_remove_laddr(laddr); } /* pending asconf (address) parameters */ TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) { /* sa_ignore FREED_MEMORY */ TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); SCTP_FREE(aparam, SCTP_M_ASC_ADDR); } TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) { /* sa_ignore FREED_MEMORY */ TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); if (aack->data != NULL) { sctp_m_freem(aack->data); } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); } /* clean up auth stuff */ if (asoc->local_hmacs) sctp_free_hmaclist(asoc->local_hmacs); if (asoc->peer_hmacs) sctp_free_hmaclist(asoc->peer_hmacs); if (asoc->local_auth_chunks) sctp_free_chunklist(asoc->local_auth_chunks); if (asoc->peer_auth_chunks) sctp_free_chunklist(asoc->peer_auth_chunks); sctp_free_authinfo(&asoc->authinfo); LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ } /* Insert new items here :> */ /* Get rid of LOCK */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INFO_WUNLOCK(); SCTP_INP_RLOCK(inp); } #ifdef SCTP_TRACK_FREED_ASOCS if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* now clean up the tasoc itself */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); } else { LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist); } #else SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); #endif if (from_inpcbfree == SCTP_NORMAL_PROC) { if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* * If its NOT the inp_free calling us AND sctp_close * as been called, we call back... */ SCTP_INP_RUNLOCK(inp); /* * This will start the kill timer (if we are the * last one) since we hold an increment yet. But * this is the only safe way to do this since * otherwise if the socket closes at the same time * we are here we might collide in the cleanup. */ sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, SCTP_CALLED_DIRECTLY_NOCMPSET); SCTP_INP_DECR_REF(inp); goto out_of; } else { /* The socket is still open. */ SCTP_INP_DECR_REF(inp); } } if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_RUNLOCK(inp); } out_of: /* destroyed the asoc */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 11); #endif return (1); } /* * determine if a destination is "reachable" based upon the addresses bound * to the current endpoint (e.g. only v4 or v6 currently bound) */ /* * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use * assoc level v4/v6 flags, as the assoc *may* not have the same address * types bound as its endpoint */ int sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr) { struct sctp_inpcb *inp; int answer; /* * No locks here, the TCB, in all cases is already locked and an * assoc is up. There is either a INP lock by the caller applied (in * asconf case when deleting an address) or NOT in the HB case, * however if HB then the INP increment is up and the INP will not * be removed (on top of the fact that we have a TCB lock). So we * only want to read the sctp_flags, which is either bound-all or * not.. no protection needed since once an assoc is up you can't be * changing your binding. */ inp = stcb->sctp_ep; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* if bound all, destination is not restricted */ /* * RRS: Question during lock work: Is this correct? If you * are bound-all you still might need to obey the V4--V6 * flags??? IMO this bound-all stuff needs to be removed! */ return (1); } /* NOTE: all "scope" checks are done when local addresses are added */ if (destaddr->sa_family == AF_INET6) { answer = inp->ip_inp.inp.inp_vflag & INP_IPV6; } else if (destaddr->sa_family == AF_INET) { answer = inp->ip_inp.inp.inp_vflag & INP_IPV4; } else { /* invalid family, so it's unreachable */ answer = 0; } return (answer); } /* * update the inp_vflags on an endpoint */ static void sctp_update_ep_vflag(struct sctp_inpcb *inp) { struct sctp_laddr *laddr; /* first clear the flag */ inp->ip_inp.inp.inp_vflag = 0; /* set the flag based on addresses on the ep list */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { continue; } if (laddr->ifa->address.sa.sa_family == AF_INET6) { inp->ip_inp.inp.inp_vflag |= INP_IPV6; } else if (laddr->ifa->address.sa.sa_family == AF_INET) { inp->ip_inp.inp.inp_vflag |= INP_IPV4; } } } /* * Add the address to the endpoint local address list There is nothing to be * done if we are bound to all addresses */ void sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action) { struct sctp_laddr *laddr; int fnd, error = 0; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-useable addr. */ return; } } /* first, is it already present? */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd == 0) { /* Not in the ep list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action); if (error != 0) return; inp->laddr_count++; /* update inp_vflag flags */ if (ifa->address.sa.sa_family == AF_INET6) { inp->ip_inp.inp.inp_vflag |= INP_IPV6; } else if (ifa->address.sa.sa_family == AF_INET) { inp->ip_inp.inp.inp_vflag |= INP_IPV4; } } return; } /* * select a new (hopefully reachable) destination net (should only be used * when we deleted an ep addr that is the only usable source address to reach * the destination net) */ static void sctp_select_primary_destination(struct sctp_tcb *stcb) { struct sctp_nets *net; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* for now, we'll just pick the first reachable one we find */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) continue; if (sctp_destination_is_reachable(stcb, (struct sockaddr *)&net->ro._l_addr)) { /* found a reachable destination */ stcb->asoc.primary_destination = net; } } /* I can't there from here! ...we're gonna die shortly... */ } /* * Delete the address from the endpoint local address list There is nothing * to be done if we are bound to all addresses */ void sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; int fnd; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd && (inp->laddr_count < 2)) { /* can't delete unless there are at LEAST 2 addresses */ return; } if (fnd) { /* * clean up any use of this address go through our * associations and clear any last_used_address that match * this one for each assoc, see if a new primary_destination * is needed */ struct sctp_tcb *stcb; /* clean up "next_addr_touse" */ if (inp->next_addr_touse == laddr) /* delete this address */ inp->next_addr_touse = NULL; /* clean up "last_used_address" */ LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { struct sctp_nets *net; SCTP_TCB_LOCK(stcb); if (stcb->asoc.last_used_address == laddr) /* delete this address */ stcb->asoc.last_used_address = NULL; /* * Now spin through all the nets and purge any ref * to laddr */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._s_addr && (net->ro._s_addr->ifa == laddr->ifa)) { /* Yep, purge src address selected */ sctp_rtentry_t *rt; /* delete this address if cached */ rt = net->ro.ro_rt; if (rt != NULL) { RTFREE(rt); net->ro.ro_rt = NULL; } sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } } SCTP_TCB_UNLOCK(stcb); } /* for each tcb */ /* remove it from the ep list */ sctp_remove_laddr(laddr); inp->laddr_count--; /* update inp_vflag flags */ sctp_update_ep_vflag(inp); } return; } /* * Add the address to the TCB local address restricted list. * This is a "pending" address list (eg. addresses waiting for an * ASCONF-ACK response) and cannot be used as a valid source address. */ void sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; struct sctpladdr *list; /* * Assumes TCB is locked.. and possibly the INP. May need to * confirm/fix that if we need it and is not the case. */ list = &stcb->asoc.sctp_restricted_addrs; inp = stcb->sctp_ep; if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ return; } } /* does the address already exist? */ LIST_FOREACH(laddr, list, sctp_nxt_addr) { if (laddr->ifa == ifa) { return; } } /* add to the list */ (void)sctp_insert_laddr(list, ifa, 0); return; } /* * insert an laddr entry with the given ifa for the desired list */ int sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act) { struct sctp_laddr *laddr; laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = ifa; laddr->action = act; atomic_add_int(&ifa->refcount, 1); /* insert it */ LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr); return (0); } /* * Remove an laddr entry from the local address list (on an assoc) */ void sctp_remove_laddr(struct sctp_laddr *laddr) { /* remove from the list */ LIST_REMOVE(laddr, sctp_nxt_addr); sctp_free_ifa(laddr->ifa); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr); SCTP_DECR_LADDR_COUNT(); } /* * Remove a local address from the TCB local address restricted list */ void sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; /* * This is called by asconf work. It is assumed that a) The TCB is * locked and b) The INP is locked. This is true in as much as I can * trace through the entry asconf code where I did these locks. * Again, the ASCONF code is a bit different in that it does lock * the INP during its work often times. This must be since we don't * want other proc's looking up things while what they are looking * up is changing :-D */ inp = stcb->sctp_ep; /* if subset bound and don't allow ASCONF's, can't delete last */ if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) && sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { if (stcb->sctp_ep->laddr_count < 2) { /* can't delete last address */ return; } } LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { /* remove the address if it exists */ if (laddr->ifa == NULL) continue; if (laddr->ifa == ifa) { sctp_remove_laddr(laddr); return; } } /* address not found! */ return; } /* * Temporarily remove for __APPLE__ until we use the Tiger equivalents */ /* sysctl */ static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC; static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR; void sctp_pcb_init() { /* * SCTP initialization for the PCB structures should be called by * the sctp_init() funciton. */ int i; struct timeval tv; if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) { /* error I was called twice */ return; } SCTP_BASE_VAR(sctp_pcb_initialized) = 1; #if defined(SCTP_LOCAL_TRACE_BUF) bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log)); #endif (void)SCTP_GETTIME_TIMEVAL(&tv); #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec; SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec; #else SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec; SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec; #endif /* init the empty list of (All) Endpoints */ LIST_INIT(&SCTP_BASE_INFO(listhead)); /* init the hash table of endpoints */ TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale)); SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31), &SCTP_BASE_INFO(hashasocmark)); SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashmark)); SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashtcpmark)); SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize); SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH, &SCTP_BASE_INFO(hashvrfmark)); SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE, &SCTP_BASE_INFO(vrf_ifn_hashmark)); /* init the zones */ /* * FIX ME: Should check for NULL returns, but if it does fail we are * doomed to panic anyways... add later maybe. */ SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep", sizeof(struct sctp_inpcb), maxsockets); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc", sizeof(struct sctp_tcb), sctp_max_number_of_assoc); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr", sizeof(struct sctp_laddr), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr", sizeof(struct sctp_nets), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk", sizeof(struct sctp_tmit_chunk), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq", sizeof(struct sctp_queued_to_read), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out", sizeof(struct sctp_stream_queue_pending), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf", sizeof(struct sctp_asconf), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack", sizeof(struct sctp_asconf_ack), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); /* Master Lock INIT for info structure */ SCTP_INP_INFO_LOCK_INIT(); SCTP_STATLOG_INIT_LOCK(); SCTP_IPI_COUNT_INIT(); SCTP_IPI_ADDR_INIT(); #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_INIT(); #endif LIST_INIT(&SCTP_BASE_INFO(addr_wq)); SCTP_WQ_ADDR_INIT(); /* not sure if we need all the counts */ SCTP_BASE_INFO(ipi_count_ep) = 0; /* assoc/tcb zone info */ SCTP_BASE_INFO(ipi_count_asoc) = 0; /* local addrlist zone info */ SCTP_BASE_INFO(ipi_count_laddr) = 0; /* remote addrlist zone info */ SCTP_BASE_INFO(ipi_count_raddr) = 0; /* chunk info */ SCTP_BASE_INFO(ipi_count_chunk) = 0; /* socket queue zone info */ SCTP_BASE_INFO(ipi_count_readq) = 0; /* stream out queue cont */ SCTP_BASE_INFO(ipi_count_strmoq) = 0; SCTP_BASE_INFO(ipi_free_strmoq) = 0; SCTP_BASE_INFO(ipi_free_chunks) = 0; SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer)); /* Init the TIMEWAIT list */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]); } sctp_startup_iterator(); /* * INIT the default VRF which for BSD is the only one, other O/S's * may have more. But initially they must start with one and then * add the VRF's as addresses are added. */ sctp_init_vrf_list(SCTP_DEFAULT_VRF); } /* * Assumes that the SCTP_BASE_INFO() lock is NOT held. */ void sctp_pcb_finish(void) { struct sctp_vrflist *vrf_bucket; struct sctp_vrf *vrf, *nvrf; struct sctp_ifn *ifn, *nifn; struct sctp_ifa *ifa, *nifa; struct sctpvtaghead *chain; struct sctp_tagblock *twait_block, *prev_twait_block; struct sctp_laddr *wi, *nwi; int i; /* * Free BSD the it thread never exits but we do clean up. The only * way freebsd reaches here if we have VRF's but we still add the * ifdef to make it compile on old versions. */ { struct sctp_iterator *it, *nit; SCTP_IPI_ITERATOR_WQ_LOCK(); TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { if (it->vn != curvnet) { continue; } TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); SCTP_ITERATOR_LOCK(); if ((sctp_it_ctl.cur_it) && (sctp_it_ctl.cur_it->vn == curvnet)) { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; } SCTP_ITERATOR_UNLOCK(); } SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer)); SCTP_WQ_ADDR_LOCK(); LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { LIST_REMOVE(wi, sctp_nxt_addr); SCTP_DECR_LADDR_COUNT(); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi); } SCTP_WQ_ADDR_UNLOCK(); /* * free the vrf/ifn/ifa lists and hashes (be sure address monitor is * destroyed first). */ vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))]; LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) { LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) { LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) { /* free the ifa */ LIST_REMOVE(ifa, next_bucket); LIST_REMOVE(ifa, next_ifa); SCTP_FREE(ifa, SCTP_M_IFA); } /* free the ifn */ LIST_REMOVE(ifn, next_bucket); LIST_REMOVE(ifn, next_ifn); SCTP_FREE(ifn, SCTP_M_IFN); } SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); /* free the vrf */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); } /* free the vrf hashes */ SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark)); SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark)); /* * free the TIMEWAIT list elements malloc'd in the function * sctp_add_vtag_to_timewait()... */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { chain = &SCTP_BASE_INFO(vtag_timewait)[i]; if (!LIST_EMPTY(chain)) { prev_twait_block = NULL; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { if (prev_twait_block) { SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } prev_twait_block = twait_block; } SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } } /* free the locks and mutexes */ #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_DESTROY(); #endif SCTP_IPI_ADDR_DESTROY(); SCTP_STATLOG_DESTROY(); SCTP_INP_INFO_LOCK_DESTROY(); SCTP_WQ_ADDR_DESTROY(); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack)); /* Get rid of other stuff to */ if (SCTP_BASE_INFO(sctp_asochash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark)); if (SCTP_BASE_INFO(sctp_ephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark)); if (SCTP_BASE_INFO(sctp_tcpephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark)); } int sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m, int iphlen, int offset, int limit, struct sctphdr *sh, struct sockaddr *altsa) { /* * grub through the INIT pulling addresses and loading them to the * nets structure in the asoc. The from address in the mbuf should * also be loaded (if it is not already). This routine can be called * with either INIT or INIT-ACK's as long as the m points to the IP * packet and the offset points to the beginning of the parameters. */ struct sctp_inpcb *inp, *l_inp; struct sctp_nets *net, *nnet, *net_tmp; struct ip *iph; struct sctp_paramhdr *phdr, parm_buf; struct sctp_tcb *stcb_tmp; uint16_t ptype, plen; struct sockaddr *sa; struct sockaddr_storage dest_store; struct sockaddr *local_sa = (struct sockaddr *)&dest_store; struct sockaddr_in sin; struct sockaddr_in6 sin6; uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_random *p_random = NULL; uint16_t random_len = 0; uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_hmac_algo *hmacs = NULL; uint16_t hmacs_len = 0; uint8_t saw_asconf = 0; uint8_t saw_asconf_ack = 0; uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_chunk_list *chunks = NULL; uint16_t num_chunks = 0; sctp_key_t *new_key; uint32_t keylen; int got_random = 0, got_hmacs = 0, got_chklist = 0; /* First get the destination address setup too. */ memset(&sin, 0, sizeof(sin)); memset(&sin6, 0, sizeof(sin6)); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = stcb->rport; sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_port = stcb->rport; if (altsa == NULL) { iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *sin_2; sin_2 = (struct sockaddr_in *)(local_sa); memset(sin_2, 0, sizeof(sin)); sin_2->sin_family = AF_INET; sin_2->sin_len = sizeof(sin); sin_2->sin_port = sh->dest_port; sin_2->sin_addr.s_addr = iph->ip_dst.s_addr; sin.sin_addr = iph->ip_src; sa = (struct sockaddr *)&sin; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *sin6_2; ip6 = mtod(m, struct ip6_hdr *); sin6_2 = (struct sockaddr_in6 *)(local_sa); memset(sin6_2, 0, sizeof(sin6)); sin6_2->sin6_family = AF_INET6; sin6_2->sin6_len = sizeof(struct sockaddr_in6); sin6_2->sin6_port = sh->dest_port; sin6.sin6_addr = ip6->ip6_src; sa = (struct sockaddr *)&sin6; break; } #endif default: return (-1); break; } } else { /* * For cookies we use the src address NOT from the packet * but from the original INIT */ sa = altsa; } /* Turn off ECN until we get through all params */ stcb->asoc.ecn_allowed = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* mark all addresses that we have currently on the list */ net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC; } /* does the source address already exist? if so skip it */ l_inp = inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* no scope set here since we have a tcb already. */ if ((sa->sa_family == AF_INET) && (stcb->asoc.ipv4_addr_legal)) { if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) { return (-1); } } else if ((sa->sa_family == AF_INET6) && (stcb->asoc.ipv6_addr_legal)) { if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { return (-2); } } } else { if (net_tmp != NULL && stcb_tmp == stcb) { net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } else if (stcb_tmp != stcb) { /* It belongs to another association? */ if (stcb_tmp) SCTP_TCB_UNLOCK(stcb_tmp); return (-3); } } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-4); } /* * peer must explicitly turn this on. This may have been initialized * to be "on" in order to allow local addr changes while INIT's are * in flight. */ stcb->asoc.peer_supports_asconf = 0; /* now we must go through each of the params. */ phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); /* * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype, * (int)plen); */ if (offset + plen > limit) { break; } if (plen == 0) { break; } if (ptype == SCTP_IPV4_ADDRESS) { if (stcb->asoc.ipv4_addr_legal) { struct sctp_ipv4addr_param *p4, p4_buf; /* ok get the v4 address and check/add */ phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); if (plen != sizeof(struct sctp_ipv4addr_param) || phdr == NULL) { return (-5); } p4 = (struct sctp_ipv4addr_param *)phdr; sin.sin_addr.s_addr = p4->addr; if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { /* Skip multi-cast addresses */ goto next_param; } if ((sin.sin_addr.s_addr == INADDR_BROADCAST) || (sin.sin_addr.s_addr == INADDR_ANY)) { goto next_param; } sa = (struct sockaddr *)&sin; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* * no scope set since we have a tcb * already */ /* * we must validate the state again * here */ add_it_now: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-7); } if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) { return (-8); } } else if (stcb_tmp == stcb) { if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-10); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ if (stcb_tmp) { if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { /* * in setup state we * abort this guy */ sctp_abort_an_association(stcb_tmp->sctp_ep, stcb_tmp, 1, NULL, 0); goto add_it_now; } SCTP_TCB_UNLOCK(stcb_tmp); } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-12); } return (-13); } } } else if (ptype == SCTP_IPV6_ADDRESS) { if (stcb->asoc.ipv6_addr_legal) { /* ok get the v6 address and check/add */ struct sctp_ipv6addr_param *p6, p6_buf; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); if (plen != sizeof(struct sctp_ipv6addr_param) || phdr == NULL) { return (-14); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy((caddr_t)&sin6.sin6_addr, p6->addr, sizeof(p6->addr)); if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { /* Skip multi-cast addresses */ goto next_param; } if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { /* * Link local make no sense without * scope */ goto next_param; } sa = (struct sockaddr *)&sin6; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if (stcb_tmp == NULL && (inp == stcb->sctp_ep || inp == NULL)) { /* * we must validate the state again * here */ add_it_now6: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-16); } /* * we must add the address, no scope * set */ if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) { return (-17); } } else if (stcb_tmp == stcb) { /* * we must validate the state again * here */ if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-19); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ if (stcb_tmp) if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { /* * in setup state we * abort this guy */ sctp_abort_an_association(stcb_tmp->sctp_ep, stcb_tmp, 1, NULL, 0); goto add_it_now6; } SCTP_TCB_UNLOCK(stcb_tmp); if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-21); } return (-22); } } } else if (ptype == SCTP_ECN_CAPABLE) { stcb->asoc.ecn_allowed = 1; } else if (ptype == SCTP_ULP_ADAPTATION) { if (stcb->asoc.state != SCTP_STATE_OPEN) { struct sctp_adaptation_layer_indication ai, *aip; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ai, sizeof(ai)); aip = (struct sctp_adaptation_layer_indication *)phdr; if (aip) { stcb->asoc.peers_adaptation = ntohl(aip->indication); stcb->asoc.adaptation_needed = 1; } } } else if (ptype == SCTP_SET_PRIM_ADDR) { struct sctp_asconf_addr_param lstore, *fee; struct sctp_asconf_addrv4_param *fii; int lptype; struct sockaddr *lsa = NULL; stcb->asoc.peer_supports_asconf = 1; if (plen > sizeof(lstore)) { return (-23); } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore))); if (phdr == NULL) { return (-24); } fee = (struct sctp_asconf_addr_param *)phdr; lptype = ntohs(fee->addrp.ph.param_type); if (lptype == SCTP_IPV4_ADDRESS) { if (plen != sizeof(struct sctp_asconf_addrv4_param)) { SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addrv4_param), plen); } else { fii = (struct sctp_asconf_addrv4_param *)fee; sin.sin_addr.s_addr = fii->addrp.addr; lsa = (struct sockaddr *)&sin; } } else if (lptype == SCTP_IPV6_ADDRESS) { if (plen != sizeof(struct sctp_asconf_addr_param)) { SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addr_param), plen); } else { memcpy(sin6.sin6_addr.s6_addr, fee->addrp.addr, sizeof(fee->addrp.addr)); lsa = (struct sockaddr *)&sin6; } } if (lsa) { (void)sctp_set_primary_addr(stcb, sa, NULL); } } else if (ptype == SCTP_HAS_NAT_SUPPORT) { stcb->asoc.peer_supports_nat = 1; } else if (ptype == SCTP_PRSCTP_SUPPORTED) { /* Peer supports pr-sctp */ stcb->asoc.peer_supports_prsctp = 1; } else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { /* A supported extension chunk */ struct sctp_supported_chunk_types_param *pr_supported; uint8_t local_store[SCTP_PARAM_BUFFER_SIZE]; int num_ent, i; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen)); if (phdr == NULL) { return (-25); } stcb->asoc.peer_supports_asconf = 0; stcb->asoc.peer_supports_prsctp = 0; stcb->asoc.peer_supports_pktdrop = 0; stcb->asoc.peer_supports_strreset = 0; stcb->asoc.peer_supports_nr_sack = 0; stcb->asoc.peer_supports_auth = 0; pr_supported = (struct sctp_supported_chunk_types_param *)phdr; num_ent = plen - sizeof(struct sctp_paramhdr); for (i = 0; i < num_ent; i++) { switch (pr_supported->chunk_types[i]) { case SCTP_ASCONF: case SCTP_ASCONF_ACK: stcb->asoc.peer_supports_asconf = 1; break; case SCTP_FORWARD_CUM_TSN: stcb->asoc.peer_supports_prsctp = 1; break; case SCTP_PACKET_DROPPED: stcb->asoc.peer_supports_pktdrop = 1; break; case SCTP_NR_SELECTIVE_ACK: stcb->asoc.peer_supports_nr_sack = 1; break; case SCTP_STREAM_RESET: stcb->asoc.peer_supports_strreset = 1; break; case SCTP_AUTHENTICATION: stcb->asoc.peer_supports_auth = 1; break; default: /* one I have not learned yet */ break; } } - } else if (ptype == SCTP_ECN_NONCE_SUPPORTED) { - /* Peer supports ECN-nonce */ - stcb->asoc.peer_supports_ecn_nonce = 1; - stcb->asoc.ecn_nonce_allowed = 1; } else if (ptype == SCTP_RANDOM) { if (plen > sizeof(random_store)) break; if (got_random) { /* already processed a RANDOM */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)random_store, min(sizeof(random_store), plen)); if (phdr == NULL) return (-26); p_random = (struct sctp_auth_random *)phdr; random_len = plen - sizeof(*p_random); /* enforce the random length */ if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) { SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n"); return (-27); } got_random = 1; } else if (ptype == SCTP_HMAC_LIST) { int num_hmacs; int i; if (plen > sizeof(hmacs_store)) break; if (got_hmacs) { /* already processed a HMAC list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)hmacs_store, min(plen, sizeof(hmacs_store))); if (phdr == NULL) return (-28); hmacs = (struct sctp_auth_hmac_algo *)phdr; hmacs_len = plen - sizeof(*hmacs); num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); /* validate the hmac list */ if (sctp_verify_hmac_param(hmacs, num_hmacs)) { return (-29); } if (stcb->asoc.peer_hmacs != NULL) sctp_free_hmaclist(stcb->asoc.peer_hmacs); stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs); if (stcb->asoc.peer_hmacs != NULL) { for (i = 0; i < num_hmacs; i++) { (void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs, ntohs(hmacs->hmac_ids[i])); } } got_hmacs = 1; } else if (ptype == SCTP_CHUNK_LIST) { int i; if (plen > sizeof(chunks_store)) break; if (got_chklist) { /* already processed a Chunks list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)chunks_store, min(plen, sizeof(chunks_store))); if (phdr == NULL) return (-30); chunks = (struct sctp_auth_chunk_list *)phdr; num_chunks = plen - sizeof(*chunks); if (stcb->asoc.peer_auth_chunks != NULL) sctp_clear_chunklist(stcb->asoc.peer_auth_chunks); else stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist(); for (i = 0; i < num_chunks; i++) { (void)sctp_auth_add_chunk(chunks->chunk_types[i], stcb->asoc.peer_auth_chunks); /* record asconf/asconf-ack if listed */ if (chunks->chunk_types[i] == SCTP_ASCONF) saw_asconf = 1; if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) saw_asconf_ack = 1; } got_chklist = 1; } else if ((ptype == SCTP_HEARTBEAT_INFO) || (ptype == SCTP_STATE_COOKIE) || (ptype == SCTP_UNRECOG_PARAM) || (ptype == SCTP_COOKIE_PRESERVE) || (ptype == SCTP_SUPPORTED_ADDRTYPE) || (ptype == SCTP_ADD_IP_ADDRESS) || (ptype == SCTP_DEL_IP_ADDRESS) || (ptype == SCTP_ERROR_CAUSE_IND) || (ptype == SCTP_SUCCESS_REPORT)) { /* don't care */ ; } else { if ((ptype & 0x8000) == 0x0000) { /* * must stop processing the rest of the * param's. Any report bits were handled * with the call to * sctp_arethere_unrecognized_parameters() * when the INIT or INIT-ACK was first seen. */ break; } } next_param: offset += SCTP_SIZE32(plen); if (offset >= limit) { break; } phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } /* Now check to see if we need to purge any addresses */ TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) { if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) == SCTP_ADDR_NOT_IN_ASSOC) { /* This address has been removed from the asoc */ /* remove and free it */ stcb->asoc.numnets--; TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next); sctp_free_remote_addr(net); if (net == stcb->asoc.primary_destination) { stcb->asoc.primary_destination = NULL; sctp_select_primary_destination(stcb); } } } /* validate authentication required parameters */ if (got_random && got_hmacs) { stcb->asoc.peer_supports_auth = 1; } else { stcb->asoc.peer_supports_auth = 0; } if (!stcb->asoc.peer_supports_auth && got_chklist) { /* peer does not support auth but sent a chunks list? */ return (-31); } if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf && !stcb->asoc.peer_supports_auth) { /* peer supports asconf but not auth? */ return (-32); } else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) && ((saw_asconf == 0) || (saw_asconf_ack == 0))) { return (-33); } /* concatenate the full random key */ keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len; if (chunks != NULL) { keylen += sizeof(*chunks) + num_chunks; } new_key = sctp_alloc_key(keylen); if (new_key != NULL) { /* copy in the RANDOM */ if (p_random != NULL) { keylen = sizeof(*p_random) + random_len; bcopy(p_random, new_key->key, keylen); } /* append in the AUTH chunks */ if (chunks != NULL) { bcopy(chunks, new_key->key + keylen, sizeof(*chunks) + num_chunks); keylen += sizeof(*chunks) + num_chunks; } /* append in the HMACs */ if (hmacs != NULL) { bcopy(hmacs, new_key->key + keylen, sizeof(*hmacs) + hmacs_len); } } else { /* failed to get memory for the key */ return (-34); } if (stcb->asoc.authinfo.peer_random != NULL) sctp_free_key(stcb->asoc.authinfo.peer_random); stcb->asoc.authinfo.peer_random = new_key; sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); return (0); } int sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa, struct sctp_nets *net) { /* make sure the requested primary address exists in the assoc */ if (net == NULL && sa) net = sctp_findnet(stcb, sa); if (net == NULL) { /* didn't find the requested primary address! */ return (-1); } else { /* set the primary address */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* Must be confirmed, so queue to set */ net->dest_state |= SCTP_ADDR_REQ_PRIMARY; return (0); } stcb->asoc.primary_destination = net; net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY; net = TAILQ_FIRST(&stcb->asoc.nets); if (net != stcb->asoc.primary_destination) { /* * first one on the list is NOT the primary * sctp_cmpaddr() is much more efficient if the * primary is the first on the list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } } int sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait) { /* * This function serves two purposes. It will see if a TAG can be * re-used and return 1 for yes it is ok and 0 for don't use that * tag. A secondary function it will do is purge out old tags that * can be removed. */ struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct sctpasochead *head; struct sctp_tcb *stcb; int i; SCTP_INP_INFO_RLOCK(); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))]; if (head == NULL) { /* invalid vtag */ goto skip_vtag_check; } LIST_FOREACH(stcb, head, sctp_asocs) { /* * We choose not to lock anything here. TCB's can't be * removed since we have the read lock, so they can't be * freed on us, same thing for the INP. I may be wrong with * this assumption, but we will go with it for now :-) */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { continue; } if (stcb->asoc.my_vtag == tag) { /* candidate */ if (stcb->rport != rport) { continue; } if (stcb->sctp_ep->sctp_lport != lport) { continue; } /* Its a used tag set */ SCTP_INP_INFO_RUNLOCK(); return (0); } } skip_vtag_check: chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; /* Now what about timed wait ? */ if (!LIST_EMPTY(chain)) { /* * Block(s) are present, lets see if we have this tag in the * list */ LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if (twait_block->vtag_block[i].v_tag == 0) { /* not used */ continue; } else if ((long)twait_block->vtag_block[i].tv_sec_at_expire < now->tv_sec) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; } else if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { /* Bad tag, sorry :< */ SCTP_INP_INFO_RUNLOCK(); return (0); } } } } SCTP_INP_INFO_RUNLOCK(); return (1); } static sctp_assoc_t reneged_asoc_ids[256]; static uint8_t reneged_at = 0; static void sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { /* * We must hunt this association for MBUF's past the cumack (i.e. * out of order data that we can renege on). */ struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *nchk; uint32_t cumulative_tsn_p1; struct sctp_queued_to_read *ctl, *nctl; int cnt, strmat; uint32_t gap, i; int fnd = 0; /* We look for anything larger than the cum-ack + 1 */ asoc = &stcb->asoc; if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) { /* none we can reneg on. */ return; } SCTP_STAT_INCR(sctps_protocol_drains_done); cumulative_tsn_p1 = asoc->cumulative_tsn + 1; cnt = 0; /* First look in the re-assembly queue */ TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) { /* Yep it is above cum-ack */ cnt++; SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn); asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size); sctp_ucount_decr(asoc->cnt_on_reasm_queue); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); } } /* Ok that was fun, now we will drain all the inbound streams? */ for (strmat = 0; strmat < asoc->streamincnt; strmat++) { TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) { if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) { /* Yep it is above cum-ack */ cnt++; SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn); asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length); sctp_ucount_decr(asoc->cnt_on_all_streams); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } sctp_free_remote_addr(ctl->whoFrom); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); } } } if (cnt) { /* We must back down to see what the new highest is */ for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { asoc->highest_tsn_inside_map = i; fnd = 1; break; } } if (!fnd) { asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; } /* * Question, should we go through the delivery queue? The * only reason things are on here is the app not reading OR * a p-d-api up. An attacker COULD send enough in to * initiate the PD-API and then send a bunch of stuff to * other streams... these would wind up on the delivery * queue.. and then we would not get to them. But in order * to do this I then have to back-track and un-deliver * sequence numbers in streams.. el-yucko. I think for now * we will NOT look at the delivery queue and leave it to be * something to consider later. An alternative would be to * abort the P-D-API with a notification and then deliver * the data.... Or another method might be to keep track of * how many times the situation occurs and if we see a * possible attack underway just abort the association. */ #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt); #endif /* * Now do we need to find a new * asoc->highest_tsn_inside_map? */ asoc->last_revoke_count = cnt; (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); /* sa_ignore NO_NULL_CHK */ sctp_send_sack(stcb); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED); reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb); reneged_at++; } /* * Another issue, in un-setting the TSN's in the mapping array we * DID NOT adjust the highest_tsn marker. This will cause one of * two things to occur. It may cause us to do extra work in checking * for our mapping array movement. More importantly it may cause us * to SACK every datagram. This may not be a bad thing though since * we will recover once we get our cum-ack above and all this stuff * we dumped recovered. */ } void sctp_drain() { /* * We must walk the PCB lists for ALL associations here. The system * is LOW on MBUF's and needs help. This is where reneging will * occur. We really hope this does NOT happen! */ VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct sctp_inpcb *inp; struct sctp_tcb *stcb; SCTP_STAT_INCR(sctps_protocol_drain_calls); if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { #ifdef VIMAGE continue; #else return; #endif } SCTP_INP_INFO_RLOCK(); LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { /* For each endpoint */ SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { /* For each association */ SCTP_TCB_LOCK(stcb); sctp_drain_mbufs(inp, stcb); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } SCTP_INP_INFO_RUNLOCK(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * start a new iterator * iterates through all endpoints and associations based on the pcb_state * flags and asoc_state. "af" (mandatory) is executed for all matching * assocs and "ef" (optional) is executed when the iterator completes. * "inpf" (optional) is executed for each new endpoint as it is being * iterated through. inpe (optional) is called when the inp completes * its way through all the stcbs. */ int sctp_initiate_iterator(inp_func inpf, asoc_func af, inp_func inpe, uint32_t pcb_state, uint32_t pcb_features, uint32_t asoc_state, void *argp, uint32_t argi, end_func ef, struct sctp_inpcb *s_inp, uint8_t chunk_output_off) { struct sctp_iterator *it = NULL; if (af == NULL) { return (-1); } SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator), SCTP_M_ITER); if (it == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); return (ENOMEM); } memset(it, 0, sizeof(*it)); it->function_assoc = af; it->function_inp = inpf; if (inpf) it->done_current_ep = 0; else it->done_current_ep = 1; it->function_atend = ef; it->pointer = argp; it->val = argi; it->pcb_flags = pcb_state; it->pcb_features = pcb_features; it->asoc_state = asoc_state; it->function_inp_end = inpe; it->no_chunk_output = chunk_output_off; it->vn = curvnet; if (s_inp) { /* Assume lock is held here */ it->inp = s_inp; SCTP_INP_INCR_REF(it->inp); it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP; } else { SCTP_INP_INFO_RLOCK(); it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead)); if (it->inp) { SCTP_INP_INCR_REF(it->inp); } SCTP_INP_INFO_RUNLOCK(); it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP; } SCTP_IPI_ITERATOR_WQ_LOCK(); TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (sctp_it_ctl.iterator_running == 0) { sctp_wakeup_iterator(); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); /* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */ return (0); } diff --git a/sys/netinet/sctp_structs.h b/sys/netinet/sctp_structs.h index 4a8fa4694b8b..c17729f055f2 100644 --- a/sys/netinet/sctp_structs.h +++ b/sys/netinet/sctp_structs.h @@ -1,1169 +1,1156 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_structs.h,v 1.13 2005/03/06 16:04:18 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_structs_h__ #define __sctp_structs_h__ #include #include #include struct sctp_timer { sctp_os_timer_t timer; int type; /* * Depending on the timer type these will be setup and cast with the * appropriate entity. */ void *ep; void *tcb; void *net; void *vnet; /* for sanity checking */ void *self; uint32_t ticks; uint32_t stopped_from; }; struct sctp_foo_stuff { struct sctp_inpcb *inp; uint32_t lineno; uint32_t ticks; int updown; }; /* * This is the information we track on each interface that we know about from * the distant end. */ TAILQ_HEAD(sctpnetlisthead, sctp_nets); struct sctp_stream_reset_list { TAILQ_ENTRY(sctp_stream_reset_list) next_resp; uint32_t tsn; int number_entries; struct sctp_stream_reset_out_request req; }; TAILQ_HEAD(sctp_resethead, sctp_stream_reset_list); /* * Users of the iterator need to malloc a iterator with a call to * sctp_initiate_iterator(inp_func, assoc_func, inp_func, pcb_flags, pcb_features, * asoc_state, void-ptr-arg, uint32-arg, end_func, inp); * * Use the following two defines if you don't care what pcb flags are on the EP * and/or you don't care what state the association is in. * * Note that if you specify an INP as the last argument then ONLY each * association of that single INP will be executed upon. Note that the pcb * flags STILL apply so if the inp you specify has different pcb_flags then * what you put in pcb_flags nothing will happen. use SCTP_PCB_ANY_FLAGS to * assure the inp you specify gets treated. */ #define SCTP_PCB_ANY_FLAGS 0x00000000 #define SCTP_PCB_ANY_FEATURES 0x00000000 #define SCTP_ASOC_ANY_STATE 0x00000000 typedef void (*asoc_func) (struct sctp_inpcb *, struct sctp_tcb *, void *ptr, uint32_t val); typedef int (*inp_func) (struct sctp_inpcb *, void *ptr, uint32_t val); typedef void (*end_func) (void *ptr, uint32_t val); struct sctp_iterator { TAILQ_ENTRY(sctp_iterator) sctp_nxt_itr; struct vnet *vn; struct sctp_timer tmr; struct sctp_inpcb *inp; /* current endpoint */ struct sctp_tcb *stcb; /* current* assoc */ struct sctp_inpcb *next_inp; /* special hook to skip to */ asoc_func function_assoc; /* per assoc function */ inp_func function_inp; /* per endpoint function */ inp_func function_inp_end; /* end INP function */ end_func function_atend;/* iterator completion function */ void *pointer; /* pointer for apply func to use */ uint32_t val; /* value for apply func to use */ uint32_t pcb_flags; /* endpoint flags being checked */ uint32_t pcb_features; /* endpoint features being checked */ uint32_t asoc_state; /* assoc state being checked */ uint32_t iterator_flags; uint8_t no_chunk_output; uint8_t done_current_ep; }; /* iterator_flags values */ #define SCTP_ITERATOR_DO_ALL_INP 0x00000001 #define SCTP_ITERATOR_DO_SINGLE_INP 0x00000002 TAILQ_HEAD(sctpiterators, sctp_iterator); struct sctp_copy_all { struct sctp_inpcb *inp; /* ep */ struct mbuf *m; struct sctp_sndrcvinfo sndrcv; int sndlen; int cnt_sent; int cnt_failed; }; struct sctp_asconf_iterator { struct sctpladdr list_of_work; int cnt; }; struct iterator_control { struct mtx ipi_iterator_wq_mtx; struct mtx it_mtx; SCTP_PROCESS_STRUCT thread_proc; struct sctpiterators iteratorhead; struct sctp_iterator *cur_it; uint32_t iterator_running; uint32_t iterator_flags; }; #define SCTP_ITERATOR_MUST_EXIT 0x00000001 #define SCTP_ITERATOR_STOP_CUR_IT 0x00000002 #define SCTP_ITERATOR_STOP_CUR_INP 0x00000004 struct sctp_net_route { sctp_rtentry_t *ro_rt; void *ro_lle; union sctp_sockstore _l_addr; /* remote peer addr */ struct sctp_ifa *_s_addr; /* our selected src addr */ }; struct htcp { uint16_t alpha; /* Fixed point arith, << 7 */ uint8_t beta; /* Fixed point arith, << 7 */ uint8_t modeswitch; /* Delay modeswitch until we had at least one * congestion event */ uint32_t last_cong; /* Time since last congestion event end */ uint32_t undo_last_cong; uint16_t bytes_acked; uint32_t bytecount; uint32_t minRTT; uint32_t maxRTT; uint32_t undo_maxRTT; uint32_t undo_old_maxB; /* Bandwidth estimation */ uint32_t minB; uint32_t maxB; uint32_t old_maxB; uint32_t Bi; uint32_t lasttime; }; struct sctp_nets { TAILQ_ENTRY(sctp_nets) sctp_next; /* next link */ /* * Things on the top half may be able to be split into a common * structure shared by all. */ struct sctp_timer pmtu_timer; /* * The following two in combination equate to a route entry for v6 * or v4. */ struct sctp_net_route ro; /* mtu discovered so far */ uint32_t mtu; uint32_t ssthresh; /* not sure about this one for split */ uint32_t last_cwr_tsn; uint32_t cwr_window_tsn; uint32_t ecn_ce_pkt_cnt; uint32_t lost_cnt; /* smoothed average things for RTT and RTO itself */ int lastsa; int lastsv; int rtt; /* last measured rtt value in ms */ unsigned int RTO; /* This is used for SHUTDOWN/SHUTDOWN-ACK/SEND or INIT timers */ struct sctp_timer rxt_timer; struct sctp_timer fr_timer; /* for early fr */ /* last time in seconds I sent to it */ struct timeval last_sent_time; int ref_count; /* Congestion stats per destination */ /* * flight size variables and such, sorry Vern, I could not avoid * this if I wanted performance :> */ uint32_t flight_size; uint32_t cwnd; /* actual cwnd */ uint32_t prev_cwnd; /* cwnd before any processing */ uint32_t partial_bytes_acked; /* in CA tracks when to incr a MTU */ uint32_t prev_rtt; /* tracking variables to avoid the aloc/free in sack processing */ unsigned int net_ack; unsigned int net_ack2; /* * JRS - 5/8/07 - Variable to track last time a destination was * active for CMT PF */ uint32_t last_active; /* * CMT variables (iyengar@cis.udel.edu) */ uint32_t this_sack_highest_newack; /* tracks highest TSN newly * acked for a given dest in * the current SACK. Used in * SFR and HTNA algos */ uint32_t pseudo_cumack; /* CMT CUC algorithm. Maintains next expected * pseudo-cumack for this destination */ uint32_t rtx_pseudo_cumack; /* CMT CUC algorithm. Maintains next * expected pseudo-cumack for this * destination */ /* CMT fast recovery variables */ uint32_t fast_recovery_tsn; uint32_t heartbeat_random1; uint32_t heartbeat_random2; uint32_t tos_flowlabel; struct timeval start_time; /* time when this net was created */ struct timeval last_measured_rtt; uint32_t marked_retrans;/* number or DATA chunks marked for timer * based retransmissions */ uint32_t marked_fastretrans; /* if this guy is ok or not ... status */ uint16_t dest_state; /* number of transmit failures to down this guy */ uint16_t failure_threshold; /* error stats on destination */ uint16_t error_count; /* UDP port number in case of UDP tunneling */ uint16_t port; uint8_t fast_retran_loss_recovery; uint8_t will_exit_fast_recovery; /* Flags that probably can be combined into dest_state */ uint8_t fast_retran_ip; /* fast retransmit in progress */ uint8_t hb_responded; uint8_t saw_newack; /* CMT's SFR algorithm flag */ uint8_t src_addr_selected; /* if we split we move */ uint8_t indx_of_eligible_next_to_use; uint8_t addr_is_local; /* its a local address (if known) could move * in split */ /* * CMT variables (iyengar@cis.udel.edu) */ uint8_t find_pseudo_cumack; /* CMT CUC algorithm. Flag used to * find a new pseudocumack. This flag * is set after a new pseudo-cumack * has been received and indicates * that the sender should find the * next pseudo-cumack expected for * this destination */ uint8_t find_rtx_pseudo_cumack; /* CMT CUCv2 algorithm. Flag used to * find a new rtx-pseudocumack. This * flag is set after a new * rtx-pseudo-cumack has been received * and indicates that the sender * should find the next * rtx-pseudo-cumack expected for this * destination */ uint8_t new_pseudo_cumack; /* CMT CUC algorithm. Flag used to * indicate if a new pseudo-cumack or * rtx-pseudo-cumack has been received */ uint8_t window_probe; /* Doing a window probe? */ uint8_t RTO_measured; /* Have we done the first measure */ uint8_t last_hs_used; /* index into the last HS table entry we used */ /* JRS - struct used in HTCP algorithm */ struct htcp htcp_ca; }; struct sctp_data_chunkrec { uint32_t TSN_seq; /* the TSN of this transmit */ uint16_t stream_seq; /* the stream sequence number of this transmit */ uint16_t stream_number; /* the stream number of this guy */ uint32_t payloadtype; uint32_t context; /* from send */ - /* ECN Nonce: Nonce Value for this chunk */ - uint8_t ect_nonce; uint8_t fwd_tsn_cnt; /* * part of the Highest sacked algorithm to be able to stroke counts * on ones that are FR'd. */ uint32_t fast_retran_tsn; /* sending_seq at the time of FR */ struct timeval timetodrop; /* time we drop it from queue */ uint8_t doing_fast_retransmit; uint8_t rcv_flags; /* flags pulled from data chunk on inbound for * outbound holds sending flags for PR-SCTP. */ uint8_t state_flags; uint8_t chunk_was_revoked; }; TAILQ_HEAD(sctpchunk_listhead, sctp_tmit_chunk); /* The lower byte is used to enumerate PR_SCTP policies */ #define CHUNK_FLAGS_PR_SCTP_TTL SCTP_PR_SCTP_TTL #define CHUNK_FLAGS_PR_SCTP_BUF SCTP_PR_SCTP_BUF #define CHUNK_FLAGS_PR_SCTP_RTX SCTP_PR_SCTP_RTX /* The upper byte is used a a bit mask */ #define CHUNK_FLAGS_FRAGMENT_OK 0x0100 struct chk_id { uint16_t id; uint16_t can_take_data; }; struct sctp_tmit_chunk { union { struct sctp_data_chunkrec data; struct chk_id chunk_id; } rec; struct sctp_association *asoc; /* bp to asoc this belongs to */ struct timeval sent_rcv_time; /* filled in if RTT being calculated */ struct mbuf *data; /* pointer to mbuf chain of data */ struct mbuf *last_mbuf; /* pointer to last mbuf in chain */ struct sctp_nets *whoTo; TAILQ_ENTRY(sctp_tmit_chunk) sctp_next; /* next link */ int32_t sent; /* the send status */ uint16_t snd_count; /* number of times I sent */ uint16_t flags; /* flags, such as FRAGMENT_OK */ uint16_t send_size; uint16_t book_size; uint16_t mbcnt; uint16_t auth_keyid; uint8_t holds_key_ref; /* flag if auth keyid refcount is held */ uint8_t pad_inplace; uint8_t do_rtt; uint8_t book_size_scale; uint8_t no_fr_allowed; uint8_t pr_sctp_on; uint8_t copy_by_ref; uint8_t window_probe; }; /* * The first part of this structure MUST be the entire sinfo structure. Maybe * I should have made it a sub structure... we can circle back later and do * that if we want. */ struct sctp_queued_to_read { /* sinfo structure Pluse more */ uint16_t sinfo_stream; /* off the wire */ uint16_t sinfo_ssn; /* off the wire */ uint16_t sinfo_flags; /* SCTP_UNORDERED from wire use SCTP_EOF for * EOR */ uint32_t sinfo_ppid; /* off the wire */ uint32_t sinfo_context; /* pick this up from assoc def context? */ uint32_t sinfo_timetolive; /* not used by kernel */ uint32_t sinfo_tsn; /* Use this in reassembly as first TSN */ uint32_t sinfo_cumtsn; /* Use this in reassembly as last TSN */ sctp_assoc_t sinfo_assoc_id; /* our assoc id */ /* Non sinfo stuff */ uint32_t length; /* length of data */ uint32_t held_length; /* length held in sb */ struct sctp_nets *whoFrom; /* where it came from */ struct mbuf *data; /* front of the mbuf chain of data with * PKT_HDR */ struct mbuf *tail_mbuf; /* used for multi-part data */ struct mbuf *aux_data; /* used to hold/cache control if o/s does not * take it from us */ struct sctp_tcb *stcb; /* assoc, used for window update */ TAILQ_ENTRY(sctp_queued_to_read) next; uint16_t port_from; uint16_t spec_flags; /* Flags to hold the notification field */ uint8_t do_not_ref_stcb; uint8_t end_added; uint8_t pdapi_aborted; uint8_t some_taken; }; /* This data structure will be on the outbound * stream queues. Data will be pulled off from * the front of the mbuf data and chunk-ified * by the output routines. We will custom * fit every chunk we pull to the send/sent * queue to make up the next full packet * if we can. An entry cannot be removed * from the stream_out queue until * the msg_is_complete flag is set. This * means at times data/tail_mbuf MIGHT * be NULL.. If that occurs it happens * for one of two reasons. Either the user * is blocked on a send() call and has not * awoken to copy more data down... OR * the user is in the explict MSG_EOR mode * and wrote some data, but has not completed * sending. */ struct sctp_stream_queue_pending { struct mbuf *data; struct mbuf *tail_mbuf; struct timeval ts; struct sctp_nets *net; TAILQ_ENTRY(sctp_stream_queue_pending) next; uint32_t length; uint32_t timetolive; uint32_t ppid; uint32_t context; uint16_t sinfo_flags; uint16_t stream; uint16_t strseq; uint16_t act_flags; uint16_t auth_keyid; uint8_t holds_key_ref; uint8_t msg_is_complete; uint8_t some_taken; uint8_t pr_sctp_on; uint8_t sender_all_done; uint8_t put_last_out; uint8_t discard_rest; }; /* * this struct contains info that is used to track inbound stream data and * help with ordering. */ TAILQ_HEAD(sctpwheelunrel_listhead, sctp_stream_in); struct sctp_stream_in { struct sctp_readhead inqueue; uint16_t stream_no; uint16_t last_sequence_delivered; /* used for re-order */ uint8_t delivery_started; }; TAILQ_HEAD(sctpwheel_listhead, sctp_stream_out); TAILQ_HEAD(sctplist_listhead, sctp_stream_queue_pending); /* Round-robin schedulers */ struct ss_rr { /* next link in wheel */ TAILQ_ENTRY(sctp_stream_out) next_spoke; }; /* Priority scheduler */ struct ss_prio { /* next link in wheel */ TAILQ_ENTRY(sctp_stream_out) next_spoke; /* priority id */ uint16_t priority; }; /* Fair Bandwidth scheduler */ struct ss_fb { /* next link in wheel */ TAILQ_ENTRY(sctp_stream_out) next_spoke; /* stores message size */ int32_t rounds; }; /* * This union holds all data necessary for * different stream schedulers. */ union scheduling_data { struct sctpwheel_listhead out_wheel; struct sctplist_listhead out_list; }; /* * This union holds all parameters per stream * necessary for different stream schedulers. */ union scheduling_parameters { struct ss_rr rr; struct ss_prio prio; struct ss_fb fb; }; /* This struct is used to track the traffic on outbound streams */ struct sctp_stream_out { struct sctp_streamhead outqueue; union scheduling_parameters ss_params; uint16_t stream_no; uint16_t next_sequence_sent; /* next one I expect to send out */ uint8_t last_msg_incomplete; }; /* used to keep track of the addresses yet to try to add/delete */ TAILQ_HEAD(sctp_asconf_addrhead, sctp_asconf_addr); struct sctp_asconf_addr { TAILQ_ENTRY(sctp_asconf_addr) next; struct sctp_asconf_addr_param ap; struct sctp_ifa *ifa; /* save the ifa for add/del ip */ uint8_t sent; /* has this been sent yet? */ uint8_t special_del; /* not to be used in lookup */ }; struct sctp_scoping { uint8_t ipv4_addr_legal; uint8_t ipv6_addr_legal; uint8_t loopback_scope; uint8_t ipv4_local_scope; uint8_t local_scope; uint8_t site_scope; }; #define SCTP_TSN_LOG_SIZE 40 struct sctp_tsn_log { void *stcb; uint32_t tsn; uint16_t strm; uint16_t seq; uint16_t sz; uint16_t flgs; uint16_t in_pos; uint16_t in_out; }; #define SCTP_FS_SPEC_LOG_SIZE 200 struct sctp_fs_spec_log { uint32_t sent; uint32_t total_flight; uint32_t tsn; uint16_t book; uint8_t incr; uint8_t decr; }; /* This struct is here to cut out the compatiabilty * pad that bulks up both the inp and stcb. The non * pad portion MUST stay in complete sync with * sctp_sndrcvinfo... i.e. if sinfo_xxxx is added * this must be done here too. */ struct sctp_nonpad_sndrcvinfo { uint16_t sinfo_stream; uint16_t sinfo_ssn; uint16_t sinfo_flags; uint32_t sinfo_ppid; uint32_t sinfo_context; uint32_t sinfo_timetolive; uint32_t sinfo_tsn; uint32_t sinfo_cumtsn; sctp_assoc_t sinfo_assoc_id; }; /* * JRS - Structure to hold function pointers to the functions responsible * for congestion control. */ struct sctp_cc_functions { void (*sctp_set_initial_cc_param) (struct sctp_tcb *stcb, struct sctp_nets *net); void (*sctp_cwnd_update_after_sack) (struct sctp_tcb *stcb, struct sctp_association *asoc, int accum_moved, int reneged_all, int will_exit); void (*sctp_cwnd_update_after_fr) (struct sctp_tcb *stcb, struct sctp_association *asoc); void (*sctp_cwnd_update_after_timeout) (struct sctp_tcb *stcb, struct sctp_nets *net); void (*sctp_cwnd_update_after_ecn_echo) (struct sctp_tcb *stcb, - struct sctp_nets *net); + struct sctp_nets *net, int in_window, int num_pkt_lost); void (*sctp_cwnd_update_after_packet_dropped) (struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_pktdrop_chunk *cp, uint32_t * bottle_bw, uint32_t * on_queue); void (*sctp_cwnd_update_after_output) (struct sctp_tcb *stcb, struct sctp_nets *net, int burst_limit); void (*sctp_cwnd_update_after_fr_timer) (struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net); }; /* * RS - Structure to hold function pointers to the functions responsible * for stream scheduling. */ struct sctp_ss_functions { void (*sctp_ss_init) (struct sctp_tcb *stcb, struct sctp_association *asoc, int holds_lock); void (*sctp_ss_clear) (struct sctp_tcb *stcb, struct sctp_association *asoc, int clear_values, int holds_lock); void (*sctp_ss_init_stream) (struct sctp_stream_out *strq); void (*sctp_ss_add_to_stream) (struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp, int holds_lock); int (*sctp_ss_is_empty) (struct sctp_tcb *stcb, struct sctp_association *asoc); void (*sctp_ss_remove_from_stream) (struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp, int holds_lock); struct sctp_stream_out *(*sctp_ss_select_stream) (struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_association *asoc); void (*sctp_ss_scheduled) (struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_association *asoc, struct sctp_stream_out *strq, int moved_how_much); void (*sctp_ss_packet_done) (struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_association *asoc); int (*sctp_ss_get_value) (struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq, uint16_t * value); int (*sctp_ss_set_value) (struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq, uint16_t value); }; /* used to save ASCONF chunks for retransmission */ TAILQ_HEAD(sctp_asconf_head, sctp_asconf); struct sctp_asconf { TAILQ_ENTRY(sctp_asconf) next; uint32_t serial_number; uint16_t snd_count; struct mbuf *data; uint16_t len; }; /* used to save ASCONF-ACK chunks for retransmission */ TAILQ_HEAD(sctp_asconf_ackhead, sctp_asconf_ack); struct sctp_asconf_ack { TAILQ_ENTRY(sctp_asconf_ack) next; uint32_t serial_number; struct sctp_nets *last_sent_to; struct mbuf *data; uint16_t len; }; /* * Here we have information about each individual association that we track. * We probably in production would be more dynamic. But for ease of * implementation we will have a fixed array that we hunt for in a linear * fashion. */ struct sctp_association { /* association state */ int state; /* queue of pending addrs to add/delete */ struct sctp_asconf_addrhead asconf_queue; struct timeval time_entered; /* time we entered state */ struct timeval time_last_rcvd; struct timeval time_last_sent; struct timeval time_last_sat_advance; struct sctp_nonpad_sndrcvinfo def_send; /* timers and such */ struct sctp_timer hb_timer; /* hb timer */ struct sctp_timer dack_timer; /* Delayed ack timer */ struct sctp_timer asconf_timer; /* asconf */ struct sctp_timer strreset_timer; /* stream reset */ struct sctp_timer shut_guard_timer; /* shutdown guard */ struct sctp_timer autoclose_timer; /* automatic close timer */ struct sctp_timer delayed_event_timer; /* timer for delayed events */ struct sctp_timer delete_prim_timer; /* deleting primary dst */ /* list of restricted local addresses */ struct sctpladdr sctp_restricted_addrs; /* last local address pending deletion (waiting for an address add) */ struct sctp_ifa *asconf_addr_del_pending; /* Deleted primary destination (used to stop timer) */ struct sctp_nets *deleted_primary; struct sctpnetlisthead nets; /* remote address list */ /* Free chunk list */ struct sctpchunk_listhead free_chunks; /* Control chunk queue */ struct sctpchunk_listhead control_send_queue; /* ASCONF chunk queue */ struct sctpchunk_listhead asconf_send_queue; /* * Once a TSN hits the wire it is moved to the sent_queue. We * maintain two counts here (don't know if any but retran_cnt is * needed). The idea is that the sent_queue_retran_cnt reflects how * many chunks have been marked for retranmission by either T3-rxt * or FR. */ struct sctpchunk_listhead sent_queue; struct sctpchunk_listhead send_queue; /* re-assembly queue for fragmented chunks on the inbound path */ struct sctpchunk_listhead reasmqueue; /* * this queue is used when we reach a condition that we can NOT put * data into the socket buffer. We track the size of this queue and * set our rwnd to the space in the socket minus also the * size_on_delivery_queue. */ union scheduling_data ss_data; /* * This pointer will be set to NULL most of the time. But when we * have a fragmented message, where we could not get out all of the * message at the last send then this will point to the stream to go * get data from. */ struct sctp_stream_out *locked_on_sending; /* If an iterator is looking at me, this is it */ struct sctp_iterator *stcb_starting_point_for_iterator; /* ASCONF save the last ASCONF-ACK so we can resend it if necessary */ struct sctp_asconf_ackhead asconf_ack_sent; /* * pointer to last stream reset queued to control queue by us with * requests. */ struct sctp_tmit_chunk *str_reset; /* * if Source Address Selection happening, this will rotate through * the link list. */ struct sctp_laddr *last_used_address; /* stream arrays */ struct sctp_stream_in *strmin; struct sctp_stream_out *strmout; uint8_t *mapping_array; /* primary destination to use */ struct sctp_nets *primary_destination; /* For CMT */ struct sctp_nets *last_net_cmt_send_started; /* last place I got a data chunk from */ struct sctp_nets *last_data_chunk_from; /* last place I got a control from */ struct sctp_nets *last_control_chunk_from; /* circular looking for output selection */ struct sctp_stream_out *last_out_stream; /* * wait to the point the cum-ack passes req->send_reset_at_tsn for * any req on the list. */ struct sctp_resethead resetHead; /* queue of chunks waiting to be sent into the local stack */ struct sctp_readhead pending_reply_queue; /* JRS - the congestion control functions are in this struct */ struct sctp_cc_functions cc_functions; /* * JRS - value to store the currently loaded congestion control * module */ uint32_t congestion_control_module; /* RS - the stream scheduling functions are in this struct */ struct sctp_ss_functions ss_functions; /* RS - value to store the currently loaded stream scheduling module */ uint32_t stream_scheduling_module; uint32_t vrf_id; uint32_t cookie_preserve_req; /* ASCONF next seq I am sending out, inits at init-tsn */ uint32_t asconf_seq_out; uint32_t asconf_seq_out_acked; /* ASCONF last received ASCONF from peer, starts at peer's TSN-1 */ uint32_t asconf_seq_in; /* next seq I am sending in str reset messages */ uint32_t str_reset_seq_out; /* next seq I am expecting in str reset messages */ uint32_t str_reset_seq_in; /* various verification tag information */ uint32_t my_vtag; /* The tag to be used. if assoc is re-initited * by remote end, and I have unlocked this * will be regenerated to a new random value. */ uint32_t peer_vtag; /* The peers last tag */ uint32_t my_vtag_nonce; uint32_t peer_vtag_nonce; uint32_t assoc_id; /* This is the SCTP fragmentation threshold */ uint32_t smallest_mtu; /* * Special hook for Fast retransmit, allows us to track the highest * TSN that is NEW in this SACK if gap ack blocks are present. */ uint32_t this_sack_highest_gap; /* * The highest consecutive TSN that has been acked by peer on my * sends */ uint32_t last_acked_seq; /* The next TSN that I will use in sending. */ uint32_t sending_seq; /* Original seq number I used ??questionable to keep?? */ uint32_t init_seq_number; /* The Advanced Peer Ack Point, as required by the PR-SCTP */ /* (A1 in Section 4.2) */ uint32_t advanced_peer_ack_point; /* * The highest consequetive TSN at the bottom of the mapping array * (for his sends). */ uint32_t cumulative_tsn; /* * Used to track the mapping array and its offset bits. This MAY be * lower then cumulative_tsn. */ uint32_t mapping_array_base_tsn; /* * used to track highest TSN we have received and is listed in the * mapping array. */ uint32_t highest_tsn_inside_map; /* EY - new NR variables used for nr_sack based on mapping_array */ uint8_t *nr_mapping_array; uint32_t highest_tsn_inside_nr_map; - uint32_t last_echo_tsn; uint32_t fast_recovery_tsn; uint32_t sat_t3_recovery_tsn; uint32_t tsn_last_delivered; /* * For the pd-api we should re-write this a bit more efficent. We * could have multiple sctp_queued_to_read's that we are building at * once. Now we only do this when we get ready to deliver to the * socket buffer. Note that we depend on the fact that the struct is * "stuck" on the read queue until we finish all the pd-api. */ struct sctp_queued_to_read *control_pdapi; uint32_t tsn_of_pdapi_last_delivered; uint32_t pdapi_ppid; uint32_t context; uint32_t last_reset_action[SCTP_MAX_RESET_PARAMS]; uint32_t last_sending_seq[SCTP_MAX_RESET_PARAMS]; uint32_t last_base_tsnsent[SCTP_MAX_RESET_PARAMS]; #ifdef SCTP_ASOCLOG_OF_TSNS /* * special log - This adds considerable size to the asoc, but * provides a log that you can use to detect problems via kgdb. */ struct sctp_tsn_log in_tsnlog[SCTP_TSN_LOG_SIZE]; struct sctp_tsn_log out_tsnlog[SCTP_TSN_LOG_SIZE]; uint32_t cumack_log[SCTP_TSN_LOG_SIZE]; uint32_t cumack_logsnt[SCTP_TSN_LOG_SIZE]; uint16_t tsn_in_at; uint16_t tsn_out_at; uint16_t tsn_in_wrapped; uint16_t tsn_out_wrapped; uint16_t cumack_log_at; uint16_t cumack_log_atsnt; #endif /* SCTP_ASOCLOG_OF_TSNS */ #ifdef SCTP_FS_SPEC_LOG struct sctp_fs_spec_log fslog[SCTP_FS_SPEC_LOG_SIZE]; uint16_t fs_index; #endif /* * window state information and smallest MTU that I use to bound * segmentation */ uint32_t peers_rwnd; uint32_t my_rwnd; uint32_t my_last_reported_rwnd; uint32_t sctp_frag_point; uint32_t total_output_queue_size; uint32_t sb_cc; /* shadow of sb_cc */ uint32_t sb_send_resv; /* amount reserved on a send */ uint32_t my_rwnd_control_len; /* shadow of sb_mbcnt used for rwnd * control */ - /* 32 bit nonce stuff */ - uint32_t nonce_resync_tsn; - uint32_t nonce_wait_tsn; uint32_t default_flowlabel; uint32_t pr_sctp_cnt; - int ctrl_queue_cnt; /* could be removed REM */ + int ctrl_queue_cnt; /* could be removed REM - NO IT CAN'T!! RRS */ /* * All outbound datagrams queue into this list from the individual * stream queue. Here they get assigned a TSN and then await * sending. The stream seq comes when it is first put in the * individual str queue */ unsigned int stream_queue_cnt; unsigned int send_queue_cnt; unsigned int sent_queue_cnt; unsigned int sent_queue_cnt_removeable; /* * Number on sent queue that are marked for retran until this value * is 0 we only send one packet of retran'ed data. */ unsigned int sent_queue_retran_cnt; unsigned int size_on_reasm_queue; unsigned int cnt_on_reasm_queue; unsigned int fwd_tsn_cnt; /* amount of data (bytes) currently in flight (on all destinations) */ unsigned int total_flight; /* Total book size in flight */ unsigned int total_flight_count; /* count of chunks used with * book total */ /* count of destinaton nets and list of destination nets */ unsigned int numnets; /* Total error count on this association */ unsigned int overall_error_count; unsigned int cnt_msg_on_sb; /* All stream count of chunks for delivery */ unsigned int size_on_all_streams; unsigned int cnt_on_all_streams; /* Heart Beat delay in ticks */ unsigned int heart_beat_delay; /* autoclose */ unsigned int sctp_autoclose_ticks; /* how many preopen streams we have */ unsigned int pre_open_streams; /* How many streams I support coming into me */ unsigned int max_inbound_streams; /* the cookie life I award for any cookie, in seconds */ unsigned int cookie_life; /* time to delay acks for */ unsigned int delayed_ack; unsigned int old_delayed_ack; unsigned int sack_freq; unsigned int data_pkts_seen; unsigned int numduptsns; int dup_tsns[SCTP_MAX_DUP_TSNS]; unsigned int initial_init_rto_max; /* initial RTO for INIT's */ unsigned int initial_rto; /* initial send RTO */ unsigned int minrto; /* per assoc RTO-MIN */ unsigned int maxrto; /* per assoc RTO-MAX */ /* authentication fields */ sctp_auth_chklist_t *local_auth_chunks; sctp_auth_chklist_t *peer_auth_chunks; sctp_hmaclist_t *local_hmacs; /* local HMACs supported */ sctp_hmaclist_t *peer_hmacs; /* peer HMACs supported */ struct sctp_keyhead shared_keys; /* assoc's shared keys */ sctp_authinfo_t authinfo; /* randoms, cached keys */ /* * refcnt to block freeing when a sender or receiver is off coping * user data in. */ uint32_t refcnt; uint32_t chunks_on_out_queue; /* total chunks floating around, * locked by send socket buffer */ uint32_t peers_adaptation; uint16_t peer_hmac_id; /* peer HMAC id to send */ /* * Being that we have no bag to collect stale cookies, and that we * really would not want to anyway.. we will count them in this * counter. We of course feed them to the pigeons right away (I have * always thought of pigeons as flying rats). */ uint16_t stale_cookie_count; /* * For the partial delivery API, if up, invoked this is what last * TSN I delivered */ uint16_t str_of_pdapi; uint16_t ssn_of_pdapi; /* counts of actual built streams. Allocation may be more however */ /* could re-arrange to optimize space here. */ uint16_t streamincnt; uint16_t streamoutcnt; uint16_t strm_realoutsize; /* my maximum number of retrans of INIT and SEND */ /* copied from SCTP but should be individually setable */ uint16_t max_init_times; uint16_t max_send_times; uint16_t def_net_failure; /* * lock flag: 0 is ok to send, 1+ (duals as a retran count) is * awaiting ACK */ uint16_t mapping_array_size; uint16_t last_strm_seq_delivered; uint16_t last_strm_no_delivered; uint16_t last_revoke_count; int16_t num_send_timers_up; uint16_t stream_locked_on; uint16_t ecn_echo_cnt_onq; uint16_t free_chunk_cnt; uint8_t stream_locked; uint8_t authenticated; /* packet authenticated ok */ /* * This flag indicates that a SACK need to be sent. Initially this * is 1 to send the first sACK immediately. */ uint8_t send_sack; /* max burst after fast retransmit completes */ uint32_t max_burst; uint8_t sat_network; /* RTT is in range of sat net or greater */ uint8_t sat_network_lockout; /* lockout code */ uint8_t burst_limit_applied; /* Burst limit in effect at last send? */ /* flag goes on when we are doing a partial delivery api */ uint8_t hb_random_values[4]; uint8_t fragmented_delivery_inprogress; uint8_t fragment_flags; uint8_t last_flags_delivered; uint8_t hb_ect_randombit; uint8_t hb_random_idx; uint8_t hb_is_disabled; /* is the hb disabled? */ uint8_t default_tos; uint8_t asconf_del_pending; /* asconf delete last addr pending */ - /* ECN Nonce stuff */ - uint8_t receiver_nonce_sum; /* nonce I sum and put in my sack */ - uint8_t ecn_nonce_allowed; /* Tells us if ECN nonce is on */ - uint8_t nonce_sum_check;/* On off switch used during re-sync */ - uint8_t nonce_wait_for_ecne; /* flag when we expect a ECN */ - uint8_t peer_supports_ecn_nonce; - /* * This value, plus all other ack'd but above cum-ack is added * together to cross check against the bit that we have yet to * define (probably in the SACK). When the cum-ack is updated, this * sum is updated as well. */ - uint8_t nonce_sum_expect_base; + /* Flag to tell if ECN is allowed */ uint8_t ecn_allowed; /* flag to indicate if peer can do asconf */ uint8_t peer_supports_asconf; /* EY - flag to indicate if peer can do nr_sack */ uint8_t peer_supports_nr_sack; /* pr-sctp support flag */ uint8_t peer_supports_prsctp; /* peer authentication support flag */ uint8_t peer_supports_auth; /* stream resets are supported by the peer */ uint8_t peer_supports_strreset; uint8_t peer_supports_nat; /* * packet drop's are supported by the peer, we don't really care * about this but we bookkeep it anyway. */ uint8_t peer_supports_pktdrop; /* Do we allow V6/V4? */ uint8_t ipv4_addr_legal; uint8_t ipv6_addr_legal; /* Address scoping flags */ /* scope value for IPv4 */ uint8_t ipv4_local_scope; /* scope values for IPv6 */ uint8_t local_scope; uint8_t site_scope; /* loopback scope */ uint8_t loopback_scope; /* flags to handle send alternate net tracking */ uint8_t used_alt_onsack; uint8_t used_alt_asconfack; uint8_t fast_retran_loss_recovery; uint8_t sat_t3_loss_recovery; uint8_t dropped_special_cnt; uint8_t seen_a_sack_this_pkt; uint8_t stream_reset_outstanding; uint8_t stream_reset_out_is_outstanding; uint8_t delayed_connection; uint8_t ifp_had_enobuf; uint8_t saw_sack_with_frags; uint8_t saw_sack_with_nr_frags; uint8_t in_asocid_hash; uint8_t assoc_up_sent; uint8_t adaptation_needed; uint8_t adaptation_sent; /* CMT variables */ uint8_t cmt_dac_pkts_rcvd; uint8_t sctp_cmt_on_off; uint8_t iam_blocking; uint8_t cookie_how[8]; /* EY 05/05/08 - NR_SACK variable */ uint8_t sctp_nr_sack_on_off; /* JRS 5/21/07 - CMT PF variable */ uint8_t sctp_cmt_pf; /* * The mapping array is used to track out of order sequences above * last_acked_seq. 0 indicates packet missing 1 indicates packet * rec'd. We slide it up every time we raise last_acked_seq and 0 * trailing locactions out. If I get a TSN above the array * mappingArraySz, I discard the datagram and let retransmit happen. */ uint32_t marked_retrans; uint32_t timoinit; uint32_t timodata; uint32_t timosack; uint32_t timoshutdown; uint32_t timoheartbeat; uint32_t timocookie; uint32_t timoshutdownack; struct timeval start_time; struct timeval discontinuity_time; }; #endif diff --git a/sys/netinet/sctp_sysctl.c b/sys/netinet/sctp_sysctl.c index 255d95d80eda..4a75dfd780bf 100644 --- a/sys/netinet/sctp_sysctl.c +++ b/sys/netinet/sctp_sysctl.c @@ -1,1113 +1,1107 @@ /*- * Copyright (c) 2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include /* * sysctl tunable variables */ void sctp_init_sysctls() { SCTP_BASE_SYSCTL(sctp_sendspace) = SCTPCTL_MAXDGRAM_DEFAULT; SCTP_BASE_SYSCTL(sctp_recvspace) = SCTPCTL_RECVSPACE_DEFAULT; SCTP_BASE_SYSCTL(sctp_auto_asconf) = SCTPCTL_AUTOASCONF_DEFAULT; SCTP_BASE_SYSCTL(sctp_multiple_asconfs) = SCTPCTL_MULTIPLEASCONFS_DEFAULT; SCTP_BASE_SYSCTL(sctp_ecn_enable) = SCTPCTL_ECN_ENABLE_DEFAULT; - SCTP_BASE_SYSCTL(sctp_ecn_nonce) = SCTPCTL_ECN_NONCE_DEFAULT; SCTP_BASE_SYSCTL(sctp_strict_sacks) = SCTPCTL_STRICT_SACKS_DEFAULT; #if !defined(SCTP_WITH_NO_CSUM) SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) = SCTPCTL_LOOPBACK_NOCSUM_DEFAULT; #endif SCTP_BASE_SYSCTL(sctp_strict_init) = SCTPCTL_STRICT_INIT_DEFAULT; SCTP_BASE_SYSCTL(sctp_peer_chunk_oh) = SCTPCTL_PEER_CHKOH_DEFAULT; SCTP_BASE_SYSCTL(sctp_max_burst_default) = SCTPCTL_MAXBURST_DEFAULT; SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = SCTPCTL_MAXCHUNKS_DEFAULT; SCTP_BASE_SYSCTL(sctp_hashtblsize) = SCTPCTL_TCBHASHSIZE_DEFAULT; SCTP_BASE_SYSCTL(sctp_pcbtblsize) = SCTPCTL_PCBHASHSIZE_DEFAULT; SCTP_BASE_SYSCTL(sctp_min_split_point) = SCTPCTL_MIN_SPLIT_POINT_DEFAULT; SCTP_BASE_SYSCTL(sctp_chunkscale) = SCTPCTL_CHUNKSCALE_DEFAULT; SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default) = SCTPCTL_DELAYED_SACK_TIME_DEFAULT; SCTP_BASE_SYSCTL(sctp_sack_freq_default) = SCTPCTL_SACK_FREQ_DEFAULT; SCTP_BASE_SYSCTL(sctp_system_free_resc_limit) = SCTPCTL_SYS_RESOURCE_DEFAULT; SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit) = SCTPCTL_ASOC_RESOURCE_DEFAULT; SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default) = SCTPCTL_HEARTBEAT_INTERVAL_DEFAULT; SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default) = SCTPCTL_PMTU_RAISE_TIME_DEFAULT; SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default) = SCTPCTL_SHUTDOWN_GUARD_TIME_DEFAULT; SCTP_BASE_SYSCTL(sctp_secret_lifetime_default) = SCTPCTL_SECRET_LIFETIME_DEFAULT; SCTP_BASE_SYSCTL(sctp_rto_max_default) = SCTPCTL_RTO_MAX_DEFAULT; SCTP_BASE_SYSCTL(sctp_rto_min_default) = SCTPCTL_RTO_MIN_DEFAULT; SCTP_BASE_SYSCTL(sctp_rto_initial_default) = SCTPCTL_RTO_INITIAL_DEFAULT; SCTP_BASE_SYSCTL(sctp_init_rto_max_default) = SCTPCTL_INIT_RTO_MAX_DEFAULT; SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default) = SCTPCTL_VALID_COOKIE_LIFE_DEFAULT; SCTP_BASE_SYSCTL(sctp_init_rtx_max_default) = SCTPCTL_INIT_RTX_MAX_DEFAULT; SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default) = SCTPCTL_ASSOC_RTX_MAX_DEFAULT; SCTP_BASE_SYSCTL(sctp_path_rtx_max_default) = SCTPCTL_PATH_RTX_MAX_DEFAULT; SCTP_BASE_SYSCTL(sctp_add_more_threshold) = SCTPCTL_ADD_MORE_ON_OUTPUT_DEFAULT; SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default) = SCTPCTL_OUTGOING_STREAMS_DEFAULT; SCTP_BASE_SYSCTL(sctp_cmt_on_off) = SCTPCTL_CMT_ON_OFF_DEFAULT; /* EY */ SCTP_BASE_SYSCTL(sctp_nr_sack_on_off) = SCTPCTL_NR_SACK_ON_OFF_DEFAULT; SCTP_BASE_SYSCTL(sctp_cmt_use_dac) = SCTPCTL_CMT_USE_DAC_DEFAULT; SCTP_BASE_SYSCTL(sctp_cmt_pf) = SCTPCTL_CMT_PF_DEFAULT; SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) = SCTPCTL_CWND_MAXBURST_DEFAULT; SCTP_BASE_SYSCTL(sctp_early_fr) = SCTPCTL_EARLY_FAST_RETRAN_DEFAULT; SCTP_BASE_SYSCTL(sctp_early_fr_msec) = SCTPCTL_EARLY_FAST_RETRAN_MSEC_DEFAULT; SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) = SCTPCTL_ASCONF_AUTH_NOCHK_DEFAULT; SCTP_BASE_SYSCTL(sctp_auth_disable) = SCTPCTL_AUTH_DISABLE_DEFAULT; SCTP_BASE_SYSCTL(sctp_nat_friendly) = SCTPCTL_NAT_FRIENDLY_DEFAULT; SCTP_BASE_SYSCTL(sctp_L2_abc_variable) = SCTPCTL_ABC_L_VAR_DEFAULT; SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count) = SCTPCTL_MAX_CHAINED_MBUFS_DEFAULT; SCTP_BASE_SYSCTL(sctp_do_drain) = SCTPCTL_DO_SCTP_DRAIN_DEFAULT; SCTP_BASE_SYSCTL(sctp_hb_maxburst) = SCTPCTL_HB_MAX_BURST_DEFAULT; SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit) = SCTPCTL_ABORT_AT_LIMIT_DEFAULT; SCTP_BASE_SYSCTL(sctp_strict_data_order) = SCTPCTL_STRICT_DATA_ORDER_DEFAULT; SCTP_BASE_SYSCTL(sctp_min_residual) = SCTPCTL_MIN_RESIDUAL_DEFAULT; SCTP_BASE_SYSCTL(sctp_max_retran_chunk) = SCTPCTL_MAX_RETRAN_CHUNK_DEFAULT; SCTP_BASE_SYSCTL(sctp_logging_level) = SCTPCTL_LOGGING_LEVEL_DEFAULT; /* JRS - Variable for default congestion control module */ SCTP_BASE_SYSCTL(sctp_default_cc_module) = SCTPCTL_DEFAULT_CC_MODULE_DEFAULT; /* RS - Variable for default stream scheduling module */ SCTP_BASE_SYSCTL(sctp_default_ss_module) = SCTPCTL_DEFAULT_SS_MODULE_DEFAULT; SCTP_BASE_SYSCTL(sctp_default_frag_interleave) = SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DEFAULT; SCTP_BASE_SYSCTL(sctp_mobility_base) = SCTPCTL_MOBILITY_BASE_DEFAULT; SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) = SCTPCTL_MOBILITY_FASTHANDOFF_DEFAULT; SCTP_BASE_SYSCTL(sctp_vtag_time_wait) = SCTPCTL_TIME_WAIT_DEFAULT; SCTP_BASE_SYSCTL(sctp_buffer_splitting) = SCTPCTL_BUFFER_SPLITTING_DEFAULT; SCTP_BASE_SYSCTL(sctp_initial_cwnd) = SCTPCTL_INITIAL_CWND_DEFAULT; #if defined(SCTP_LOCAL_TRACE_BUF) memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); #endif SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable) = SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_DEFAULT; SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) = SCTPCTL_UDP_TUNNELING_PORT_DEFAULT; SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) = SCTPCTL_SACK_IMMEDIATELY_ENABLE_DEFAULT; SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly) = SCTPCTL_NAT_FRIENDLY_INITS_DEFAULT; #if defined(SCTP_DEBUG) SCTP_BASE_SYSCTL(sctp_debug_on) = SCTPCTL_DEBUG_DEFAULT; #endif #if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_BASE_SYSCTL(sctp_output_unlocked) = SCTPCTL_OUTPUT_UNLOCKED_DEFAULT; #endif } /* It returns an upper limit. No filtering is done here */ static unsigned int number_of_addresses(struct sctp_inpcb *inp) { int cnt; struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; struct sctp_laddr *laddr; cnt = 0; /* neither Mac OS X nor FreeBSD support mulitple routing functions */ if ((vrf = sctp_find_vrf(inp->def_vrf_id)) == NULL) { return (0); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->address.sa.sa_family == AF_INET) || (sctp_ifa->address.sa.sa_family == AF_INET6)) { cnt++; } } } } else { LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if ((laddr->ifa->address.sa.sa_family == AF_INET) || (laddr->ifa->address.sa.sa_family == AF_INET6)) { cnt++; } } } return (cnt); } static int copy_out_local_addresses(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sysctl_req *req) { struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; int loopback_scope, ipv4_local_scope, local_scope, site_scope; int ipv4_addr_legal, ipv6_addr_legal; struct sctp_vrf *vrf; struct xsctp_laddr xladdr; struct sctp_laddr *laddr; int error; /* Turn on all the appropriate scope */ if (stcb) { /* use association specific values */ loopback_scope = stcb->asoc.loopback_scope; ipv4_local_scope = stcb->asoc.ipv4_local_scope; local_scope = stcb->asoc.local_scope; site_scope = stcb->asoc.site_scope; } else { /* use generic values for endpoints */ loopback_scope = 1; ipv4_local_scope = 1; local_scope = 1; site_scope = 1; } /* use only address families of interest */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ipv6_addr_legal = 1; if (SCTP_IPV6_V6ONLY(inp)) { ipv4_addr_legal = 0; } else { ipv4_addr_legal = 1; } } else { ipv4_addr_legal = 1; ipv6_addr_legal = 0; } /* neither Mac OS X nor FreeBSD support mulitple routing functions */ if ((vrf = sctp_find_vrf(inp->def_vrf_id)) == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (-1); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) /* Skip loopback if loopback_scope not set */ continue; LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (stcb) { /* * ignore if blacklisted at * association level */ if (sctp_is_addr_restricted(stcb, sctp_ifa)) continue; } switch (sctp_ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifa->address.sa; if (sin->sin_addr.s_addr == 0) continue; if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) continue; } else { continue; } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifa->address.sa; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) continue; if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { /* * bad link * local * address */ if (sa6_recoverscope(sin6) != 0) continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) continue; } else { continue; } break; #endif default: continue; } memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); memcpy((void *)&xladdr.address, (const void *)&sctp_ifa->address, sizeof(union sctp_sockstore)); SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); if (error) { return (error); } else { SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); } } } } else { LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { /* ignore if blacklisted at association level */ if (stcb && sctp_is_addr_restricted(stcb, laddr->ifa)) continue; memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); memcpy((void *)&xladdr.address, (const void *)&laddr->ifa->address, sizeof(union sctp_sockstore)); xladdr.start_time.tv_sec = (uint32_t) laddr->start_time.tv_sec; xladdr.start_time.tv_usec = (uint32_t) laddr->start_time.tv_usec; SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); if (error) { return (error); } else { SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); } } } memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); xladdr.last = 1; SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); if (error) { return (error); } else { SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); return (0); } } /* * sysctl functions */ static int sctp_assoclist(SYSCTL_HANDLER_ARGS) { unsigned int number_of_endpoints; unsigned int number_of_local_addresses; unsigned int number_of_associations; unsigned int number_of_remote_addresses; unsigned int n; int error; struct sctp_inpcb *inp; struct sctp_tcb *stcb; struct sctp_nets *net; struct xsctp_inpcb xinpcb; struct xsctp_tcb xstcb; struct xsctp_raddr xraddr; struct socket *so; number_of_endpoints = 0; number_of_local_addresses = 0; number_of_associations = 0; number_of_remote_addresses = 0; SCTP_INP_INFO_RLOCK(); if (req->oldptr == USER_ADDR_NULL) { LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { SCTP_INP_RLOCK(inp); number_of_endpoints++; number_of_local_addresses += number_of_addresses(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { number_of_associations++; number_of_local_addresses += number_of_addresses(inp); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { number_of_remote_addresses++; } } SCTP_INP_RUNLOCK(inp); } SCTP_INP_INFO_RUNLOCK(); n = (number_of_endpoints + 1) * sizeof(struct xsctp_inpcb) + (number_of_local_addresses + number_of_endpoints + number_of_associations) * sizeof(struct xsctp_laddr) + (number_of_associations + number_of_endpoints) * sizeof(struct xsctp_tcb) + (number_of_remote_addresses + number_of_associations) * sizeof(struct xsctp_raddr); /* request some more memory than needed */ req->oldidx = (n + n / 8); return 0; } if (req->newptr != USER_ADDR_NULL) { SCTP_INP_INFO_RUNLOCK(); SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_SYSCTL, EPERM); return EPERM; } LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* if its allgone it is being freed - skip it */ goto skip; } xinpcb.last = 0; xinpcb.local_port = ntohs(inp->sctp_lport); xinpcb.flags = inp->sctp_flags; xinpcb.features = inp->sctp_features; xinpcb.total_sends = inp->total_sends; xinpcb.total_recvs = inp->total_recvs; xinpcb.total_nospaces = inp->total_nospaces; xinpcb.fragmentation_point = inp->sctp_frag_point; so = inp->sctp_socket; if ((so == NULL) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { xinpcb.qlen = 0; xinpcb.maxqlen = 0; } else { xinpcb.qlen = so->so_qlen; xinpcb.maxqlen = so->so_qlimit; } SCTP_INP_INCR_REF(inp); SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xinpcb, sizeof(struct xsctp_inpcb)); if (error) { SCTP_INP_DECR_REF(inp); return error; } SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); error = copy_out_local_addresses(inp, NULL, req); if (error) { SCTP_INP_DECR_REF(inp); return error; } LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); xstcb.last = 0; xstcb.local_port = ntohs(inp->sctp_lport); xstcb.remote_port = ntohs(stcb->rport); if (stcb->asoc.primary_destination != NULL) xstcb.primary_addr = stcb->asoc.primary_destination->ro._l_addr; xstcb.heartbeat_interval = stcb->asoc.heart_beat_delay; xstcb.state = SCTP_GET_STATE(&stcb->asoc); /* FIXME */ /* 7.0 does not support these */ xstcb.assoc_id = sctp_get_associd(stcb); xstcb.peers_rwnd = stcb->asoc.peers_rwnd; xstcb.in_streams = stcb->asoc.streamincnt; xstcb.out_streams = stcb->asoc.streamoutcnt; xstcb.max_nr_retrans = stcb->asoc.overall_error_count; xstcb.primary_process = 0; /* not really supported * yet */ xstcb.T1_expireries = stcb->asoc.timoinit + stcb->asoc.timocookie; xstcb.T2_expireries = stcb->asoc.timoshutdown + stcb->asoc.timoshutdownack; xstcb.retransmitted_tsns = stcb->asoc.marked_retrans; xstcb.start_time.tv_sec = (uint32_t) stcb->asoc.start_time.tv_sec; xstcb.start_time.tv_usec = (uint32_t) stcb->asoc.start_time.tv_usec; xstcb.discontinuity_time.tv_sec = (uint32_t) stcb->asoc.discontinuity_time.tv_sec; xstcb.discontinuity_time.tv_usec = (uint32_t) stcb->asoc.discontinuity_time.tv_usec; xstcb.total_sends = stcb->total_sends; xstcb.total_recvs = stcb->total_recvs; xstcb.local_tag = stcb->asoc.my_vtag; xstcb.remote_tag = stcb->asoc.peer_vtag; xstcb.initial_tsn = stcb->asoc.init_seq_number; xstcb.highest_tsn = stcb->asoc.sending_seq - 1; xstcb.cumulative_tsn = stcb->asoc.last_acked_seq; xstcb.cumulative_tsn_ack = stcb->asoc.cumulative_tsn; xstcb.mtu = stcb->asoc.smallest_mtu; xstcb.refcnt = stcb->asoc.refcnt; SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xstcb, sizeof(struct xsctp_tcb)); if (error) { SCTP_INP_DECR_REF(inp); atomic_subtract_int(&stcb->asoc.refcnt, 1); return error; } SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); error = copy_out_local_addresses(inp, stcb, req); if (error) { SCTP_INP_DECR_REF(inp); atomic_subtract_int(&stcb->asoc.refcnt, 1); return error; } TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { xraddr.last = 0; xraddr.address = net->ro._l_addr; xraddr.active = ((net->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE); xraddr.confirmed = ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0); xraddr.heartbeat_enabled = ((net->dest_state & SCTP_ADDR_NOHB) == 0); xraddr.rto = net->RTO; xraddr.max_path_rtx = net->failure_threshold; xraddr.rtx = net->marked_retrans; xraddr.error_counter = net->error_count; xraddr.cwnd = net->cwnd; xraddr.flight_size = net->flight_size; xraddr.mtu = net->mtu; xraddr.rtt = net->rtt; xraddr.start_time.tv_sec = (uint32_t) net->start_time.tv_sec; xraddr.start_time.tv_usec = (uint32_t) net->start_time.tv_usec; SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xraddr, sizeof(struct xsctp_raddr)); if (error) { SCTP_INP_DECR_REF(inp); atomic_subtract_int(&stcb->asoc.refcnt, 1); return error; } SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); } atomic_subtract_int(&stcb->asoc.refcnt, 1); memset((void *)&xraddr, 0, sizeof(struct xsctp_raddr)); xraddr.last = 1; SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); error = SYSCTL_OUT(req, &xraddr, sizeof(struct xsctp_raddr)); if (error) { SCTP_INP_DECR_REF(inp); return error; } SCTP_INP_INFO_RLOCK(); SCTP_INP_RLOCK(inp); } SCTP_INP_DECR_REF(inp); SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); memset((void *)&xstcb, 0, sizeof(struct xsctp_tcb)); xstcb.last = 1; error = SYSCTL_OUT(req, &xstcb, sizeof(struct xsctp_tcb)); if (error) { return error; } skip: SCTP_INP_INFO_RLOCK(); } SCTP_INP_INFO_RUNLOCK(); memset((void *)&xinpcb, 0, sizeof(struct xsctp_inpcb)); xinpcb.last = 1; error = SYSCTL_OUT(req, &xinpcb, sizeof(struct xsctp_inpcb)); return error; } #define RANGECHK(var, min, max) \ if ((var) < (min)) { (var) = (min); } \ else if ((var) > (max)) { (var) = (max); } static int sysctl_sctp_udp_tunneling_check(SYSCTL_HANDLER_ARGS) { int error; uint32_t old_sctp_udp_tunneling_port; SCTP_INP_INFO_RLOCK(); old_sctp_udp_tunneling_port = SCTP_BASE_SYSCTL(sctp_udp_tunneling_port); SCTP_INP_INFO_RUNLOCK(); error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); if (error == 0) { RANGECHK(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port), SCTPCTL_UDP_TUNNELING_PORT_MIN, SCTPCTL_UDP_TUNNELING_PORT_MAX); if (old_sctp_udp_tunneling_port == SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)) { error = 0; goto out; } SCTP_INP_INFO_WLOCK(); if (old_sctp_udp_tunneling_port) { sctp_over_udp_stop(); } if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)) { if (sctp_over_udp_start()) { SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) = 0; } } SCTP_INP_INFO_WUNLOCK(); } out: return (error); } static int sysctl_sctp_check(SYSCTL_HANDLER_ARGS) { int error; error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); if (error == 0) { RANGECHK(SCTP_BASE_SYSCTL(sctp_sendspace), SCTPCTL_MAXDGRAM_MIN, SCTPCTL_MAXDGRAM_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_recvspace), SCTPCTL_RECVSPACE_MIN, SCTPCTL_RECVSPACE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_auto_asconf), SCTPCTL_AUTOASCONF_MIN, SCTPCTL_AUTOASCONF_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_ecn_enable), SCTPCTL_ECN_ENABLE_MIN, SCTPCTL_ECN_ENABLE_MAX); - RANGECHK(SCTP_BASE_SYSCTL(sctp_ecn_nonce), SCTPCTL_ECN_NONCE_MIN, SCTPCTL_ECN_NONCE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_strict_sacks), SCTPCTL_STRICT_SACKS_MIN, SCTPCTL_STRICT_SACKS_MAX); #if !defined(SCTP_WITH_NO_CSUM) RANGECHK(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback), SCTPCTL_LOOPBACK_NOCSUM_MIN, SCTPCTL_LOOPBACK_NOCSUM_MAX); #endif RANGECHK(SCTP_BASE_SYSCTL(sctp_strict_init), SCTPCTL_STRICT_INIT_MIN, SCTPCTL_STRICT_INIT_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_peer_chunk_oh), SCTPCTL_PEER_CHKOH_MIN, SCTPCTL_PEER_CHKOH_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_max_burst_default), SCTPCTL_MAXBURST_MIN, SCTPCTL_MAXBURST_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue), SCTPCTL_MAXCHUNKS_MIN, SCTPCTL_MAXCHUNKS_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_hashtblsize), SCTPCTL_TCBHASHSIZE_MIN, SCTPCTL_TCBHASHSIZE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_pcbtblsize), SCTPCTL_PCBHASHSIZE_MIN, SCTPCTL_PCBHASHSIZE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_min_split_point), SCTPCTL_MIN_SPLIT_POINT_MIN, SCTPCTL_MIN_SPLIT_POINT_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_chunkscale), SCTPCTL_CHUNKSCALE_MIN, SCTPCTL_CHUNKSCALE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default), SCTPCTL_DELAYED_SACK_TIME_MIN, SCTPCTL_DELAYED_SACK_TIME_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_sack_freq_default), SCTPCTL_SACK_FREQ_MIN, SCTPCTL_SACK_FREQ_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_system_free_resc_limit), SCTPCTL_SYS_RESOURCE_MIN, SCTPCTL_SYS_RESOURCE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit), SCTPCTL_ASOC_RESOURCE_MIN, SCTPCTL_ASOC_RESOURCE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default), SCTPCTL_HEARTBEAT_INTERVAL_MIN, SCTPCTL_HEARTBEAT_INTERVAL_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default), SCTPCTL_PMTU_RAISE_TIME_MIN, SCTPCTL_PMTU_RAISE_TIME_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default), SCTPCTL_SHUTDOWN_GUARD_TIME_MIN, SCTPCTL_SHUTDOWN_GUARD_TIME_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default), SCTPCTL_SECRET_LIFETIME_MIN, SCTPCTL_SECRET_LIFETIME_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_max_default), SCTPCTL_RTO_MAX_MIN, SCTPCTL_RTO_MAX_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_min_default), SCTPCTL_RTO_MIN_MIN, SCTPCTL_RTO_MIN_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_initial_default), SCTPCTL_RTO_INITIAL_MIN, SCTPCTL_RTO_INITIAL_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_init_rto_max_default), SCTPCTL_INIT_RTO_MAX_MIN, SCTPCTL_INIT_RTO_MAX_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default), SCTPCTL_VALID_COOKIE_LIFE_MIN, SCTPCTL_VALID_COOKIE_LIFE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_init_rtx_max_default), SCTPCTL_INIT_RTX_MAX_MIN, SCTPCTL_INIT_RTX_MAX_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default), SCTPCTL_ASSOC_RTX_MAX_MIN, SCTPCTL_ASSOC_RTX_MAX_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_path_rtx_max_default), SCTPCTL_PATH_RTX_MAX_MIN, SCTPCTL_PATH_RTX_MAX_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_add_more_threshold), SCTPCTL_ADD_MORE_ON_OUTPUT_MIN, SCTPCTL_ADD_MORE_ON_OUTPUT_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default), SCTPCTL_OUTGOING_STREAMS_MIN, SCTPCTL_OUTGOING_STREAMS_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_cmt_on_off), SCTPCTL_CMT_ON_OFF_MIN, SCTPCTL_CMT_ON_OFF_MAX); /* EY */ RANGECHK(SCTP_BASE_SYSCTL(sctp_nr_sack_on_off), SCTPCTL_NR_SACK_ON_OFF_MIN, SCTPCTL_NR_SACK_ON_OFF_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_cmt_use_dac), SCTPCTL_CMT_USE_DAC_MIN, SCTPCTL_CMT_USE_DAC_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_cmt_pf), SCTPCTL_CMT_PF_MIN, SCTPCTL_CMT_PF_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst), SCTPCTL_CWND_MAXBURST_MIN, SCTPCTL_CWND_MAXBURST_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_early_fr), SCTPCTL_EARLY_FAST_RETRAN_MIN, SCTPCTL_EARLY_FAST_RETRAN_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_early_fr_msec), SCTPCTL_EARLY_FAST_RETRAN_MSEC_MIN, SCTPCTL_EARLY_FAST_RETRAN_MSEC_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk), SCTPCTL_ASCONF_AUTH_NOCHK_MIN, SCTPCTL_ASCONF_AUTH_NOCHK_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_auth_disable), SCTPCTL_AUTH_DISABLE_MIN, SCTPCTL_AUTH_DISABLE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_nat_friendly), SCTPCTL_NAT_FRIENDLY_MIN, SCTPCTL_NAT_FRIENDLY_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_L2_abc_variable), SCTPCTL_ABC_L_VAR_MIN, SCTPCTL_ABC_L_VAR_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count), SCTPCTL_MAX_CHAINED_MBUFS_MIN, SCTPCTL_MAX_CHAINED_MBUFS_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_do_drain), SCTPCTL_DO_SCTP_DRAIN_MIN, SCTPCTL_DO_SCTP_DRAIN_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_hb_maxburst), SCTPCTL_HB_MAX_BURST_MIN, SCTPCTL_HB_MAX_BURST_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit), SCTPCTL_ABORT_AT_LIMIT_MIN, SCTPCTL_ABORT_AT_LIMIT_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_strict_data_order), SCTPCTL_STRICT_DATA_ORDER_MIN, SCTPCTL_STRICT_DATA_ORDER_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_min_residual), SCTPCTL_MIN_RESIDUAL_MIN, SCTPCTL_MIN_RESIDUAL_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_max_retran_chunk), SCTPCTL_MAX_RETRAN_CHUNK_MIN, SCTPCTL_MAX_RETRAN_CHUNK_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_logging_level), SCTPCTL_LOGGING_LEVEL_MIN, SCTPCTL_LOGGING_LEVEL_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_default_cc_module), SCTPCTL_DEFAULT_CC_MODULE_MIN, SCTPCTL_DEFAULT_CC_MODULE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_default_ss_module), SCTPCTL_DEFAULT_SS_MODULE_MIN, SCTPCTL_DEFAULT_SS_MODULE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_default_frag_interleave), SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MIN, SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_vtag_time_wait), SCTPCTL_TIME_WAIT_MIN, SCTPCTL_TIME_WAIT_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_buffer_splitting), SCTPCTL_BUFFER_SPLITTING_MIN, SCTPCTL_BUFFER_SPLITTING_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_initial_cwnd), SCTPCTL_INITIAL_CWND_MIN, SCTPCTL_INITIAL_CWND_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_mobility_base), SCTPCTL_MOBILITY_BASE_MIN, SCTPCTL_MOBILITY_BASE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff), SCTPCTL_MOBILITY_FASTHANDOFF_MIN, SCTPCTL_MOBILITY_FASTHANDOFF_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable), SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_MIN, SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_enable_sack_immediately), SCTPCTL_SACK_IMMEDIATELY_ENABLE_MIN, SCTPCTL_SACK_IMMEDIATELY_ENABLE_MAX); RANGECHK(SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly), SCTPCTL_NAT_FRIENDLY_INITS_MIN, SCTPCTL_NAT_FRIENDLY_INITS_MAX); #ifdef SCTP_DEBUG RANGECHK(SCTP_BASE_SYSCTL(sctp_debug_on), SCTPCTL_DEBUG_MIN, SCTPCTL_DEBUG_MAX); #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) RANGECHK(SCTP_BASE_SYSCTL(sctp_output_unlocked), SCTPCTL_OUTPUT_UNLOCKED_MIN, SCTPCTL_OUTPUT_UNLOCKED_MAX); #endif } return (error); } #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) static int sysctl_stat_get(SYSCTL_HANDLER_ARGS) { int cpu, error; struct sctpstat sb, *sarry; memset(&sb, 0, sizeof(sb)); for (cpu = 0; cpu < mp_ncpus; cpu++) { sarry = &SCTP_BASE_STATS[cpu]; if (sarry->sctps_discontinuitytime.tv_sec > sb.sctps_discontinuitytime.tv_sec) { sb.sctps_discontinuitytime.tv_sec = sarry->sctps_discontinuitytime.tv_sec; sb.sctps_discontinuitytime.tv_usec = sarry->sctps_discontinuitytime.tv_usec; } sb.sctps_currestab += sarry->sctps_currestab; sb.sctps_activeestab += sarry->sctps_activeestab; sb.sctps_restartestab += sarry->sctps_restartestab; sb.sctps_collisionestab += sarry->sctps_collisionestab; sb.sctps_passiveestab += sarry->sctps_passiveestab; sb.sctps_aborted += sarry->sctps_aborted; sb.sctps_shutdown += sarry->sctps_shutdown; sb.sctps_outoftheblue += sarry->sctps_outoftheblue; sb.sctps_checksumerrors += sarry->sctps_checksumerrors; sb.sctps_outcontrolchunks += sarry->sctps_outcontrolchunks; sb.sctps_outorderchunks += sarry->sctps_outorderchunks; sb.sctps_outunorderchunks += sarry->sctps_outunorderchunks; sb.sctps_incontrolchunks += sarry->sctps_incontrolchunks; sb.sctps_inorderchunks += sarry->sctps_inorderchunks; sb.sctps_inunorderchunks += sarry->sctps_inunorderchunks; sb.sctps_fragusrmsgs += sarry->sctps_fragusrmsgs; sb.sctps_reasmusrmsgs += sarry->sctps_reasmusrmsgs; sb.sctps_outpackets += sarry->sctps_outpackets; sb.sctps_inpackets += sarry->sctps_inpackets; sb.sctps_recvpackets += sarry->sctps_recvpackets; sb.sctps_recvdatagrams += sarry->sctps_recvdatagrams; sb.sctps_recvpktwithdata += sarry->sctps_recvpktwithdata; sb.sctps_recvsacks += sarry->sctps_recvsacks; sb.sctps_recvdata += sarry->sctps_recvdata; sb.sctps_recvdupdata += sarry->sctps_recvdupdata; sb.sctps_recvheartbeat += sarry->sctps_recvheartbeat; sb.sctps_recvheartbeatack += sarry->sctps_recvheartbeatack; sb.sctps_recvecne += sarry->sctps_recvecne; sb.sctps_recvauth += sarry->sctps_recvauth; sb.sctps_recvauthmissing += sarry->sctps_recvauthmissing; sb.sctps_recvivalhmacid += sarry->sctps_recvivalhmacid; sb.sctps_recvivalkeyid += sarry->sctps_recvivalkeyid; sb.sctps_recvauthfailed += sarry->sctps_recvauthfailed; sb.sctps_recvexpress += sarry->sctps_recvexpress; sb.sctps_recvexpressm += sarry->sctps_recvexpressm; sb.sctps_recvnocrc += sarry->sctps_recvnocrc; sb.sctps_recvswcrc += sarry->sctps_recvswcrc; sb.sctps_recvhwcrc += sarry->sctps_recvhwcrc; sb.sctps_sendpackets += sarry->sctps_sendpackets; sb.sctps_sendsacks += sarry->sctps_sendsacks; sb.sctps_senddata += sarry->sctps_senddata; sb.sctps_sendretransdata += sarry->sctps_sendretransdata; sb.sctps_sendfastretrans += sarry->sctps_sendfastretrans; sb.sctps_sendmultfastretrans += sarry->sctps_sendmultfastretrans; sb.sctps_sendheartbeat += sarry->sctps_sendheartbeat; sb.sctps_sendecne += sarry->sctps_sendecne; sb.sctps_sendauth += sarry->sctps_sendauth; sb.sctps_senderrors += sarry->sctps_senderrors; sb.sctps_sendnocrc += sarry->sctps_sendnocrc; sb.sctps_sendswcrc += sarry->sctps_sendswcrc; sb.sctps_sendhwcrc += sarry->sctps_sendhwcrc; sb.sctps_pdrpfmbox += sarry->sctps_pdrpfmbox; sb.sctps_pdrpfehos += sarry->sctps_pdrpfehos; sb.sctps_pdrpmbda += sarry->sctps_pdrpmbda; sb.sctps_pdrpmbct += sarry->sctps_pdrpmbct; sb.sctps_pdrpbwrpt += sarry->sctps_pdrpbwrpt; sb.sctps_pdrpcrupt += sarry->sctps_pdrpcrupt; sb.sctps_pdrpnedat += sarry->sctps_pdrpnedat; sb.sctps_pdrppdbrk += sarry->sctps_pdrppdbrk; sb.sctps_pdrptsnnf += sarry->sctps_pdrptsnnf; sb.sctps_pdrpdnfnd += sarry->sctps_pdrpdnfnd; sb.sctps_pdrpdiwnp += sarry->sctps_pdrpdiwnp; sb.sctps_pdrpdizrw += sarry->sctps_pdrpdizrw; sb.sctps_pdrpbadd += sarry->sctps_pdrpbadd; sb.sctps_pdrpmark += sarry->sctps_pdrpmark; sb.sctps_timoiterator += sarry->sctps_timoiterator; sb.sctps_timodata += sarry->sctps_timodata; sb.sctps_timowindowprobe += sarry->sctps_timowindowprobe; sb.sctps_timoinit += sarry->sctps_timoinit; sb.sctps_timosack += sarry->sctps_timosack; sb.sctps_timoshutdown += sarry->sctps_timoshutdown; sb.sctps_timoheartbeat += sarry->sctps_timoheartbeat; sb.sctps_timocookie += sarry->sctps_timocookie; sb.sctps_timosecret += sarry->sctps_timosecret; sb.sctps_timopathmtu += sarry->sctps_timopathmtu; sb.sctps_timoshutdownack += sarry->sctps_timoshutdownack; sb.sctps_timoshutdownguard += sarry->sctps_timoshutdownguard; sb.sctps_timostrmrst += sarry->sctps_timostrmrst; sb.sctps_timoearlyfr += sarry->sctps_timoearlyfr; sb.sctps_timoasconf += sarry->sctps_timoasconf; sb.sctps_timodelprim += sarry->sctps_timodelprim; sb.sctps_timoautoclose += sarry->sctps_timoautoclose; sb.sctps_timoassockill += sarry->sctps_timoassockill; sb.sctps_timoinpkill += sarry->sctps_timoinpkill; sb.sctps_earlyfrstart += sarry->sctps_earlyfrstart; sb.sctps_earlyfrstop += sarry->sctps_earlyfrstop; sb.sctps_earlyfrmrkretrans += sarry->sctps_earlyfrmrkretrans; sb.sctps_earlyfrstpout += sarry->sctps_earlyfrstpout; sb.sctps_earlyfrstpidsck1 += sarry->sctps_earlyfrstpidsck1; sb.sctps_earlyfrstpidsck2 += sarry->sctps_earlyfrstpidsck2; sb.sctps_earlyfrstpidsck3 += sarry->sctps_earlyfrstpidsck3; sb.sctps_earlyfrstpidsck4 += sarry->sctps_earlyfrstpidsck4; sb.sctps_earlyfrstrid += sarry->sctps_earlyfrstrid; sb.sctps_earlyfrstrout += sarry->sctps_earlyfrstrout; sb.sctps_earlyfrstrtmr += sarry->sctps_earlyfrstrtmr; sb.sctps_hdrops += sarry->sctps_hdrops; sb.sctps_badsum += sarry->sctps_badsum; sb.sctps_noport += sarry->sctps_noport; sb.sctps_badvtag += sarry->sctps_badvtag; sb.sctps_badsid += sarry->sctps_badsid; sb.sctps_nomem += sarry->sctps_nomem; sb.sctps_fastretransinrtt += sarry->sctps_fastretransinrtt; sb.sctps_markedretrans += sarry->sctps_markedretrans; sb.sctps_naglesent += sarry->sctps_naglesent; sb.sctps_naglequeued += sarry->sctps_naglequeued; sb.sctps_maxburstqueued += sarry->sctps_maxburstqueued; sb.sctps_ifnomemqueued += sarry->sctps_ifnomemqueued; sb.sctps_windowprobed += sarry->sctps_windowprobed; sb.sctps_lowlevelerr += sarry->sctps_lowlevelerr; sb.sctps_lowlevelerrusr += sarry->sctps_lowlevelerrusr; sb.sctps_datadropchklmt += sarry->sctps_datadropchklmt; sb.sctps_datadroprwnd += sarry->sctps_datadroprwnd; sb.sctps_ecnereducedcwnd += sarry->sctps_ecnereducedcwnd; sb.sctps_vtagexpress += sarry->sctps_vtagexpress; sb.sctps_vtagbogus += sarry->sctps_vtagbogus; sb.sctps_primary_randry += sarry->sctps_primary_randry; sb.sctps_cmt_randry += sarry->sctps_cmt_randry; sb.sctps_slowpath_sack += sarry->sctps_slowpath_sack; sb.sctps_wu_sacks_sent += sarry->sctps_wu_sacks_sent; sb.sctps_sends_with_flags += sarry->sctps_sends_with_flags; sb.sctps_sends_with_unord += sarry->sctps_sends_with_unord; sb.sctps_sends_with_eof += sarry->sctps_sends_with_eof; sb.sctps_sends_with_abort += sarry->sctps_sends_with_abort; sb.sctps_protocol_drain_calls += sarry->sctps_protocol_drain_calls; sb.sctps_protocol_drains_done += sarry->sctps_protocol_drains_done; sb.sctps_read_peeks += sarry->sctps_read_peeks; sb.sctps_cached_chk += sarry->sctps_cached_chk; sb.sctps_cached_strmoq += sarry->sctps_cached_strmoq; sb.sctps_left_abandon += sarry->sctps_left_abandon; sb.sctps_send_burst_avoid += sarry->sctps_send_burst_avoid; sb.sctps_send_cwnd_avoid += sarry->sctps_send_cwnd_avoid; sb.sctps_fwdtsn_map_over += sarry->sctps_fwdtsn_map_over; } error = SYSCTL_OUT(req, &sb, sizeof(sb)); return (error); } #endif #if defined(SCTP_LOCAL_TRACE_BUF) static int sysctl_sctp_cleartrace(SYSCTL_HANDLER_ARGS) { int error = 0; memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); return (error); } #endif /* * sysctl definitions */ SYSCTL_PROC(_net_inet_sctp, OID_AUTO, sendspace, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_sendspace), 0, sysctl_sctp_check, "IU", SCTPCTL_MAXDGRAM_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, recvspace, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_recvspace), 0, sysctl_sctp_check, "IU", SCTPCTL_RECVSPACE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, auto_asconf, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_auto_asconf), 0, sysctl_sctp_check, "IU", SCTPCTL_AUTOASCONF_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, ecn_enable, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_ecn_enable), 0, sysctl_sctp_check, "IU", SCTPCTL_ECN_ENABLE_DESC); -SYSCTL_PROC(_net_inet_sctp, OID_AUTO, ecn_nonce, CTLTYPE_UINT | CTLFLAG_RW, - &SCTP_BASE_SYSCTL(sctp_ecn_nonce), 0, sysctl_sctp_check, "IU", - SCTPCTL_ECN_NONCE_DESC); - SYSCTL_PROC(_net_inet_sctp, OID_AUTO, strict_sacks, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_strict_sacks), 0, sysctl_sctp_check, "IU", SCTPCTL_STRICT_SACKS_DESC); #if !defined(SCTP_WITH_NO_CSUM) SYSCTL_PROC(_net_inet_sctp, OID_AUTO, loopback_nocsum, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback), 0, sysctl_sctp_check, "IU", SCTPCTL_LOOPBACK_NOCSUM_DESC); #endif SYSCTL_PROC(_net_inet_sctp, OID_AUTO, strict_init, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_strict_init), 0, sysctl_sctp_check, "IU", SCTPCTL_STRICT_INIT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, peer_chkoh, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_peer_chunk_oh), 0, sysctl_sctp_check, "IU", SCTPCTL_PEER_CHKOH_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, maxburst, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_max_burst_default), 0, sysctl_sctp_check, "IU", SCTPCTL_MAXBURST_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, maxchunks, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue), 0, sysctl_sctp_check, "IU", SCTPCTL_MAXCHUNKS_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, tcbhashsize, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_hashtblsize), 0, sysctl_sctp_check, "IU", SCTPCTL_TCBHASHSIZE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, pcbhashsize, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_pcbtblsize), 0, sysctl_sctp_check, "IU", SCTPCTL_PCBHASHSIZE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, min_split_point, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_min_split_point), 0, sysctl_sctp_check, "IU", SCTPCTL_MIN_SPLIT_POINT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, chunkscale, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_chunkscale), 0, sysctl_sctp_check, "IU", SCTPCTL_CHUNKSCALE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, delayed_sack_time, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default), 0, sysctl_sctp_check, "IU", SCTPCTL_DELAYED_SACK_TIME_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, sack_freq, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_sack_freq_default), 0, sysctl_sctp_check, "IU", SCTPCTL_SACK_FREQ_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, sys_resource, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_system_free_resc_limit), 0, sysctl_sctp_check, "IU", SCTPCTL_SYS_RESOURCE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, asoc_resource, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit), 0, sysctl_sctp_check, "IU", SCTPCTL_ASOC_RESOURCE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, heartbeat_interval, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default), 0, sysctl_sctp_check, "IU", SCTPCTL_HEARTBEAT_INTERVAL_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, pmtu_raise_time, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default), 0, sysctl_sctp_check, "IU", SCTPCTL_PMTU_RAISE_TIME_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, shutdown_guard_time, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default), 0, sysctl_sctp_check, "IU", SCTPCTL_SHUTDOWN_GUARD_TIME_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, secret_lifetime, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_secret_lifetime_default), 0, sysctl_sctp_check, "IU", SCTPCTL_SECRET_LIFETIME_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, rto_max, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_rto_max_default), 0, sysctl_sctp_check, "IU", SCTPCTL_RTO_MAX_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, rto_min, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_rto_min_default), 0, sysctl_sctp_check, "IU", SCTPCTL_RTO_MIN_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, rto_initial, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_rto_initial_default), 0, sysctl_sctp_check, "IU", SCTPCTL_RTO_INITIAL_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, init_rto_max, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_init_rto_max_default), 0, sysctl_sctp_check, "IU", SCTPCTL_INIT_RTO_MAX_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, valid_cookie_life, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default), 0, sysctl_sctp_check, "IU", SCTPCTL_VALID_COOKIE_LIFE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, init_rtx_max, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_init_rtx_max_default), 0, sysctl_sctp_check, "IU", SCTPCTL_INIT_RTX_MAX_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, assoc_rtx_max, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default), 0, sysctl_sctp_check, "IU", SCTPCTL_ASSOC_RTX_MAX_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, path_rtx_max, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_path_rtx_max_default), 0, sysctl_sctp_check, "IU", SCTPCTL_PATH_RTX_MAX_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, add_more_on_output, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_add_more_threshold), 0, sysctl_sctp_check, "IU", SCTPCTL_ADD_MORE_ON_OUTPUT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, outgoing_streams, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default), 0, sysctl_sctp_check, "IU", SCTPCTL_OUTGOING_STREAMS_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, cmt_on_off, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_cmt_on_off), 0, sysctl_sctp_check, "IU", SCTPCTL_CMT_ON_OFF_DESC); /* EY */ SYSCTL_PROC(_net_inet_sctp, OID_AUTO, nr_sack_on_off, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_nr_sack_on_off), 0, sysctl_sctp_check, "IU", SCTPCTL_NR_SACK_ON_OFF_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, cmt_use_dac, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_cmt_use_dac), 0, sysctl_sctp_check, "IU", SCTPCTL_CMT_USE_DAC_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, cmt_pf, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_cmt_pf), 0, sysctl_sctp_check, "IU", SCTPCTL_CMT_PF_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, cwnd_maxburst, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst), 0, sysctl_sctp_check, "IU", SCTPCTL_CWND_MAXBURST_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, early_fast_retran, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_early_fr), 0, sysctl_sctp_check, "IU", SCTPCTL_EARLY_FAST_RETRAN_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, early_fast_retran_msec, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_early_fr_msec), 0, sysctl_sctp_check, "IU", SCTPCTL_EARLY_FAST_RETRAN_MSEC_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, asconf_auth_nochk, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk), 0, sysctl_sctp_check, "IU", SCTPCTL_ASCONF_AUTH_NOCHK_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, auth_disable, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_auth_disable), 0, sysctl_sctp_check, "IU", SCTPCTL_AUTH_DISABLE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, nat_friendly, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_nat_friendly), 0, sysctl_sctp_check, "IU", SCTPCTL_NAT_FRIENDLY_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, abc_l_var, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_L2_abc_variable), 0, sysctl_sctp_check, "IU", SCTPCTL_ABC_L_VAR_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, max_chained_mbufs, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count), 0, sysctl_sctp_check, "IU", SCTPCTL_MAX_CHAINED_MBUFS_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, do_sctp_drain, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_do_drain), 0, sysctl_sctp_check, "IU", SCTPCTL_DO_SCTP_DRAIN_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, hb_max_burst, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_hb_maxburst), 0, sysctl_sctp_check, "IU", SCTPCTL_HB_MAX_BURST_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, abort_at_limit, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit), 0, sysctl_sctp_check, "IU", SCTPCTL_ABORT_AT_LIMIT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, strict_data_order, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_strict_data_order), 0, sysctl_sctp_check, "IU", SCTPCTL_STRICT_DATA_ORDER_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, min_residual, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_min_residual), 0, sysctl_sctp_check, "IU", SCTPCTL_MIN_RESIDUAL_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, max_retran_chunk, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_max_retran_chunk), 0, sysctl_sctp_check, "IU", SCTPCTL_MAX_RETRAN_CHUNK_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, log_level, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_logging_level), 0, sysctl_sctp_check, "IU", SCTPCTL_LOGGING_LEVEL_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, default_cc_module, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_default_cc_module), 0, sysctl_sctp_check, "IU", SCTPCTL_DEFAULT_CC_MODULE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, default_ss_module, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_default_ss_module), 0, sysctl_sctp_check, "IU", SCTPCTL_DEFAULT_SS_MODULE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, default_frag_interleave, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_default_frag_interleave), 0, sysctl_sctp_check, "IU", SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, mobility_base, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_mobility_base), 0, sysctl_sctp_check, "IU", SCTPCTL_MOBILITY_BASE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, mobility_fasthandoff, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff), 0, sysctl_sctp_check, "IU", SCTPCTL_MOBILITY_FASTHANDOFF_DESC); #if defined(SCTP_LOCAL_TRACE_BUF) SYSCTL_STRUCT(_net_inet_sctp, OID_AUTO, log, CTLFLAG_RD, &SCTP_BASE_SYSCTL(sctp_log), sctp_log, "SCTP logging (struct sctp_log)"); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, clear_trace, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_log), 0, sysctl_sctp_cleartrace, "IU", "Clear SCTP Logging buffer"); #endif SYSCTL_PROC(_net_inet_sctp, OID_AUTO, udp_tunneling_for_client_enable, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable), 0, sysctl_sctp_check, "IU", SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, udp_tunneling_port, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_udp_tunneling_port), 0, sysctl_sctp_udp_tunneling_check, "IU", SCTPCTL_UDP_TUNNELING_PORT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, enable_sack_immediately, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_enable_sack_immediately), 0, sysctl_sctp_check, "IU", SCTPCTL_SACK_IMMEDIATELY_ENABLE_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, nat_friendly_init, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly), 0, sysctl_sctp_check, "IU", SCTPCTL_NAT_FRIENDLY_INITS_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, vtag_time_wait, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_vtag_time_wait), 0, sysctl_sctp_check, "IU", SCTPCTL_TIME_WAIT_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, buffer_splitting, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_buffer_splitting), 0, sysctl_sctp_check, "IU", SCTPCTL_BUFFER_SPLITTING_DESC); SYSCTL_PROC(_net_inet_sctp, OID_AUTO, initial_cwnd, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_initial_cwnd), 0, sysctl_sctp_check, "IU", SCTPCTL_INITIAL_CWND_DESC); #ifdef SCTP_DEBUG SYSCTL_PROC(_net_inet_sctp, OID_AUTO, debug, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_debug_on), 0, sysctl_sctp_check, "IU", SCTPCTL_DEBUG_DESC); #endif /* SCTP_DEBUG */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SYSCTL_PROC(_net_inet_sctp, OID_AUTO, output_unlocked, CTLTYPE_UINT | CTLFLAG_RW, &SCTP_BASE_SYSCTL(sctp_output_unlocked), 0, sysctl_sctp_check, "IU", SCTPCTL_OUTPUT_UNLOCKED_DESC); #endif #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) SYSCTL_PROC(_net_inet_sctp, OID_AUTO, stats, CTLTYPE_STRUCT | CTLFLAG_RD, 0, 0, sysctl_stat_get, "S,sctpstat", "SCTP statistics (struct sctp_stat)"); #else SYSCTL_STRUCT(_net_inet_sctp, OID_AUTO, stats, CTLFLAG_RW, &SCTP_BASE_STATS_SYSCTL, sctpstat, "SCTP statistics (struct sctp_stat)"); #endif SYSCTL_PROC(_net_inet_sctp, OID_AUTO, assoclist, CTLTYPE_OPAQUE | CTLFLAG_RD, 0, 0, sctp_assoclist, "S,xassoc", "List of active SCTP associations"); diff --git a/sys/netinet/sctp_sysctl.h b/sys/netinet/sctp_sysctl.h index fe16ba883338..3f33d3da4b5f 100644 --- a/sys/netinet/sctp_sysctl.h +++ b/sys/netinet/sctp_sysctl.h @@ -1,540 +1,534 @@ /*- * Copyright (c) 2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_sysctl_h__ #define __sctp_sysctl_h__ #include #include struct sctp_sysctl { uint32_t sctp_sendspace; uint32_t sctp_recvspace; uint32_t sctp_auto_asconf; uint32_t sctp_multiple_asconfs; uint32_t sctp_ecn_enable; - uint32_t sctp_ecn_nonce; + uint32_t sctp_not_used; uint32_t sctp_strict_sacks; #if !defined(SCTP_WITH_NO_CSUM) uint32_t sctp_no_csum_on_loopback; #endif uint32_t sctp_strict_init; uint32_t sctp_peer_chunk_oh; uint32_t sctp_max_burst_default; uint32_t sctp_max_chunks_on_queue; uint32_t sctp_hashtblsize; uint32_t sctp_pcbtblsize; uint32_t sctp_min_split_point; uint32_t sctp_chunkscale; uint32_t sctp_delayed_sack_time_default; uint32_t sctp_sack_freq_default; uint32_t sctp_system_free_resc_limit; uint32_t sctp_asoc_free_resc_limit; uint32_t sctp_heartbeat_interval_default; uint32_t sctp_pmtu_raise_time_default; uint32_t sctp_shutdown_guard_time_default; uint32_t sctp_secret_lifetime_default; uint32_t sctp_rto_max_default; uint32_t sctp_rto_min_default; uint32_t sctp_rto_initial_default; uint32_t sctp_init_rto_max_default; uint32_t sctp_valid_cookie_life_default; uint32_t sctp_init_rtx_max_default; uint32_t sctp_assoc_rtx_max_default; uint32_t sctp_path_rtx_max_default; uint32_t sctp_add_more_threshold; uint32_t sctp_nr_outgoing_streams_default; uint32_t sctp_cmt_on_off; uint32_t sctp_cmt_use_dac; /* EY 5/5/08 - nr_sack flag variable */ uint32_t sctp_nr_sack_on_off; uint32_t sctp_cmt_pf; uint32_t sctp_use_cwnd_based_maxburst; uint32_t sctp_early_fr; uint32_t sctp_early_fr_msec; uint32_t sctp_asconf_auth_nochk; uint32_t sctp_auth_disable; uint32_t sctp_nat_friendly; uint32_t sctp_L2_abc_variable; uint32_t sctp_mbuf_threshold_count; uint32_t sctp_do_drain; uint32_t sctp_hb_maxburst; uint32_t sctp_abort_if_one_2_one_hits_limit; uint32_t sctp_strict_data_order; uint32_t sctp_min_residual; uint32_t sctp_max_retran_chunk; uint32_t sctp_logging_level; /* JRS - Variable for default congestion control module */ uint32_t sctp_default_cc_module; /* RS - Variable for default stream scheduling module */ uint32_t sctp_default_ss_module; uint32_t sctp_default_frag_interleave; uint32_t sctp_mobility_base; uint32_t sctp_mobility_fasthandoff; uint32_t sctp_inits_include_nat_friendly; #if defined(SCTP_LOCAL_TRACE_BUF) struct sctp_log sctp_log; #endif uint32_t sctp_udp_tunneling_for_client_enable; uint32_t sctp_udp_tunneling_port; uint32_t sctp_enable_sack_immediately; uint32_t sctp_vtag_time_wait; uint32_t sctp_buffer_splitting; uint32_t sctp_initial_cwnd; #if defined(SCTP_DEBUG) uint32_t sctp_debug_on; #endif #if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING) uint32_t sctp_output_unlocked; #endif }; /* * limits for the sysctl variables */ /* maxdgram: Maximum outgoing SCTP buffer size */ #define SCTPCTL_MAXDGRAM_DESC "Maximum outgoing SCTP buffer size" #define SCTPCTL_MAXDGRAM_MIN 0 #define SCTPCTL_MAXDGRAM_MAX 0xFFFFFFFF #define SCTPCTL_MAXDGRAM_DEFAULT 262144 /* 256k */ /* recvspace: Maximum incoming SCTP buffer size */ #define SCTPCTL_RECVSPACE_DESC "Maximum incoming SCTP buffer size" #define SCTPCTL_RECVSPACE_MIN 0 #define SCTPCTL_RECVSPACE_MAX 0xFFFFFFFF #define SCTPCTL_RECVSPACE_DEFAULT 262144 /* 256k */ /* autoasconf: Enable SCTP Auto-ASCONF */ #define SCTPCTL_AUTOASCONF_DESC "Enable SCTP Auto-ASCONF" #define SCTPCTL_AUTOASCONF_MIN 0 #define SCTPCTL_AUTOASCONF_MAX 1 #define SCTPCTL_AUTOASCONF_DEFAULT SCTP_DEFAULT_AUTO_ASCONF /* autoasconf: Enable SCTP Auto-ASCONF */ #define SCTPCTL_MULTIPLEASCONFS_DESC "Enable SCTP Muliple-ASCONFs" #define SCTPCTL_MULTIPLEASCONFS_MIN 0 #define SCTPCTL_MULTIPLEASCONFS_MAX 1 #define SCTPCTL_MULTIPLEASCONFS_DEFAULT SCTP_DEFAULT_MULTIPLE_ASCONFS /* ecn_enable: Enable SCTP ECN */ #define SCTPCTL_ECN_ENABLE_DESC "Enable SCTP ECN" #define SCTPCTL_ECN_ENABLE_MIN 0 #define SCTPCTL_ECN_ENABLE_MAX 1 #define SCTPCTL_ECN_ENABLE_DEFAULT 1 -/* ecn_nonce: Enable SCTP ECN Nonce */ -#define SCTPCTL_ECN_NONCE_DESC "Enable SCTP ECN Nonce" -#define SCTPCTL_ECN_NONCE_MIN 0 -#define SCTPCTL_ECN_NONCE_MAX 1 -#define SCTPCTL_ECN_NONCE_DEFAULT 0 - /* strict_sacks: Enable SCTP Strict SACK checking */ #define SCTPCTL_STRICT_SACKS_DESC "Enable SCTP Strict SACK checking" #define SCTPCTL_STRICT_SACKS_MIN 0 #define SCTPCTL_STRICT_SACKS_MAX 1 #define SCTPCTL_STRICT_SACKS_DEFAULT 1 /* loopback_nocsum: Enable NO Csum on packets sent on loopback */ #define SCTPCTL_LOOPBACK_NOCSUM_DESC "Enable NO Csum on packets sent on loopback" #define SCTPCTL_LOOPBACK_NOCSUM_MIN 0 #define SCTPCTL_LOOPBACK_NOCSUM_MAX 1 #define SCTPCTL_LOOPBACK_NOCSUM_DEFAULT 1 /* strict_init: Enable strict INIT/INIT-ACK singleton enforcement */ #define SCTPCTL_STRICT_INIT_DESC "Enable strict INIT/INIT-ACK singleton enforcement" #define SCTPCTL_STRICT_INIT_MIN 0 #define SCTPCTL_STRICT_INIT_MAX 1 #define SCTPCTL_STRICT_INIT_DEFAULT 1 /* peer_chkoh: Amount to debit peers rwnd per chunk sent */ #define SCTPCTL_PEER_CHKOH_DESC "Amount to debit peers rwnd per chunk sent" #define SCTPCTL_PEER_CHKOH_MIN 0 #define SCTPCTL_PEER_CHKOH_MAX 0xFFFFFFFF #define SCTPCTL_PEER_CHKOH_DEFAULT 256 /* maxburst: Default max burst for sctp endpoints */ #define SCTPCTL_MAXBURST_DESC "Default max burst for sctp endpoints" #define SCTPCTL_MAXBURST_MIN 0 #define SCTPCTL_MAXBURST_MAX 0xFFFFFFFF #define SCTPCTL_MAXBURST_DEFAULT SCTP_DEF_MAX_BURST /* maxchunks: Default max chunks on queue per asoc */ #define SCTPCTL_MAXCHUNKS_DESC "Default max chunks on queue per asoc" #define SCTPCTL_MAXCHUNKS_MIN 0 #define SCTPCTL_MAXCHUNKS_MAX 0xFFFFFFFF #define SCTPCTL_MAXCHUNKS_DEFAULT SCTP_ASOC_MAX_CHUNKS_ON_QUEUE /* tcbhashsize: Tuneable for Hash table sizes */ #define SCTPCTL_TCBHASHSIZE_DESC "Tunable for TCB hash table sizes" #define SCTPCTL_TCBHASHSIZE_MIN 1 #define SCTPCTL_TCBHASHSIZE_MAX 0xFFFFFFFF #define SCTPCTL_TCBHASHSIZE_DEFAULT SCTP_TCBHASHSIZE /* pcbhashsize: Tuneable for PCB Hash table sizes */ #define SCTPCTL_PCBHASHSIZE_DESC "Tunable for PCB hash table sizes" #define SCTPCTL_PCBHASHSIZE_MIN 1 #define SCTPCTL_PCBHASHSIZE_MAX 0xFFFFFFFF #define SCTPCTL_PCBHASHSIZE_DEFAULT SCTP_PCBHASHSIZE /* min_split_point: Minimum size when splitting a chunk */ #define SCTPCTL_MIN_SPLIT_POINT_DESC "Minimum size when splitting a chunk" #define SCTPCTL_MIN_SPLIT_POINT_MIN 0 #define SCTPCTL_MIN_SPLIT_POINT_MAX 0xFFFFFFFF #define SCTPCTL_MIN_SPLIT_POINT_DEFAULT SCTP_DEFAULT_SPLIT_POINT_MIN /* chunkscale: Tuneable for Scaling of number of chunks and messages */ #define SCTPCTL_CHUNKSCALE_DESC "Tuneable for Scaling of number of chunks and messages" #define SCTPCTL_CHUNKSCALE_MIN 1 #define SCTPCTL_CHUNKSCALE_MAX 0xFFFFFFFF #define SCTPCTL_CHUNKSCALE_DEFAULT SCTP_CHUNKQUEUE_SCALE /* delayed_sack_time: Default delayed SACK timer in msec */ #define SCTPCTL_DELAYED_SACK_TIME_DESC "Default delayed SACK timer in msec" #define SCTPCTL_DELAYED_SACK_TIME_MIN 0 #define SCTPCTL_DELAYED_SACK_TIME_MAX 0xFFFFFFFF #define SCTPCTL_DELAYED_SACK_TIME_DEFAULT SCTP_RECV_MSEC /* sack_freq: Default SACK frequency */ #define SCTPCTL_SACK_FREQ_DESC "Default SACK frequency" #define SCTPCTL_SACK_FREQ_MIN 0 #define SCTPCTL_SACK_FREQ_MAX 0xFFFFFFFF #define SCTPCTL_SACK_FREQ_DEFAULT SCTP_DEFAULT_SACK_FREQ /* sys_resource: Max number of cached resources in the system */ #define SCTPCTL_SYS_RESOURCE_DESC "Max number of cached resources in the system" #define SCTPCTL_SYS_RESOURCE_MIN 0 #define SCTPCTL_SYS_RESOURCE_MAX 0xFFFFFFFF #define SCTPCTL_SYS_RESOURCE_DEFAULT SCTP_DEF_SYSTEM_RESC_LIMIT /* asoc_resource: Max number of cached resources in an asoc */ #define SCTPCTL_ASOC_RESOURCE_DESC "Max number of cached resources in an asoc" #define SCTPCTL_ASOC_RESOURCE_MIN 0 #define SCTPCTL_ASOC_RESOURCE_MAX 0xFFFFFFFF #define SCTPCTL_ASOC_RESOURCE_DEFAULT SCTP_DEF_ASOC_RESC_LIMIT /* heartbeat_interval: Default heartbeat interval in msec */ #define SCTPCTL_HEARTBEAT_INTERVAL_DESC "Default heartbeat interval in msec" #define SCTPCTL_HEARTBEAT_INTERVAL_MIN 0 #define SCTPCTL_HEARTBEAT_INTERVAL_MAX 0xFFFFFFFF #define SCTPCTL_HEARTBEAT_INTERVAL_DEFAULT SCTP_HB_DEFAULT_MSEC /* pmtu_raise_time: Default PMTU raise timer in sec */ #define SCTPCTL_PMTU_RAISE_TIME_DESC "Default PMTU raise timer in sec" #define SCTPCTL_PMTU_RAISE_TIME_MIN 0 #define SCTPCTL_PMTU_RAISE_TIME_MAX 0xFFFFFFFF #define SCTPCTL_PMTU_RAISE_TIME_DEFAULT SCTP_DEF_PMTU_RAISE_SEC /* shutdown_guard_time: Default shutdown guard timer in sec */ #define SCTPCTL_SHUTDOWN_GUARD_TIME_DESC "Default shutdown guard timer in sec" #define SCTPCTL_SHUTDOWN_GUARD_TIME_MIN 0 #define SCTPCTL_SHUTDOWN_GUARD_TIME_MAX 0xFFFFFFFF #define SCTPCTL_SHUTDOWN_GUARD_TIME_DEFAULT SCTP_DEF_MAX_SHUTDOWN_SEC /* secret_lifetime: Default secret lifetime in sec */ #define SCTPCTL_SECRET_LIFETIME_DESC "Default secret lifetime in sec" #define SCTPCTL_SECRET_LIFETIME_MIN 0 #define SCTPCTL_SECRET_LIFETIME_MAX 0xFFFFFFFF #define SCTPCTL_SECRET_LIFETIME_DEFAULT SCTP_DEFAULT_SECRET_LIFE_SEC /* rto_max: Default maximum retransmission timeout in msec */ #define SCTPCTL_RTO_MAX_DESC "Default maximum retransmission timeout in msec" #define SCTPCTL_RTO_MAX_MIN 0 #define SCTPCTL_RTO_MAX_MAX 0xFFFFFFFF #define SCTPCTL_RTO_MAX_DEFAULT SCTP_RTO_UPPER_BOUND /* rto_min: Default minimum retransmission timeout in msec */ #define SCTPCTL_RTO_MIN_DESC "Default minimum retransmission timeout in msec" #define SCTPCTL_RTO_MIN_MIN 0 #define SCTPCTL_RTO_MIN_MAX 0xFFFFFFFF #define SCTPCTL_RTO_MIN_DEFAULT SCTP_RTO_LOWER_BOUND /* rto_initial: Default initial retransmission timeout in msec */ #define SCTPCTL_RTO_INITIAL_DESC "Default initial retransmission timeout in msec" #define SCTPCTL_RTO_INITIAL_MIN 0 #define SCTPCTL_RTO_INITIAL_MAX 0xFFFFFFFF #define SCTPCTL_RTO_INITIAL_DEFAULT SCTP_RTO_INITIAL /* init_rto_max: Default maximum retransmission timeout during association setup in msec */ #define SCTPCTL_INIT_RTO_MAX_DESC "Default maximum retransmission timeout during association setup in msec" #define SCTPCTL_INIT_RTO_MAX_MIN 0 #define SCTPCTL_INIT_RTO_MAX_MAX 0xFFFFFFFF #define SCTPCTL_INIT_RTO_MAX_DEFAULT SCTP_RTO_UPPER_BOUND /* valid_cookie_life: Default cookie lifetime in sec */ #define SCTPCTL_VALID_COOKIE_LIFE_DESC "Default cookie lifetime in sec" #define SCTPCTL_VALID_COOKIE_LIFE_MIN 0 #define SCTPCTL_VALID_COOKIE_LIFE_MAX 0xFFFFFFFF #define SCTPCTL_VALID_COOKIE_LIFE_DEFAULT SCTP_DEFAULT_COOKIE_LIFE /* init_rtx_max: Default maximum number of retransmission for INIT chunks */ #define SCTPCTL_INIT_RTX_MAX_DESC "Default maximum number of retransmission for INIT chunks" #define SCTPCTL_INIT_RTX_MAX_MIN 0 #define SCTPCTL_INIT_RTX_MAX_MAX 0xFFFFFFFF #define SCTPCTL_INIT_RTX_MAX_DEFAULT SCTP_DEF_MAX_INIT /* assoc_rtx_max: Default maximum number of retransmissions per association */ #define SCTPCTL_ASSOC_RTX_MAX_DESC "Default maximum number of retransmissions per association" #define SCTPCTL_ASSOC_RTX_MAX_MIN 0 #define SCTPCTL_ASSOC_RTX_MAX_MAX 0xFFFFFFFF #define SCTPCTL_ASSOC_RTX_MAX_DEFAULT SCTP_DEF_MAX_SEND /* path_rtx_max: Default maximum of retransmissions per path */ #define SCTPCTL_PATH_RTX_MAX_DESC "Default maximum of retransmissions per path" #define SCTPCTL_PATH_RTX_MAX_MIN 0 #define SCTPCTL_PATH_RTX_MAX_MAX 0xFFFFFFFF #define SCTPCTL_PATH_RTX_MAX_DEFAULT SCTP_DEF_MAX_PATH_RTX /* add_more_on_output: When space wise is it worthwhile to try to add more to a socket send buffer */ #define SCTPCTL_ADD_MORE_ON_OUTPUT_DESC "When space wise is it worthwhile to try to add more to a socket send buffer" #define SCTPCTL_ADD_MORE_ON_OUTPUT_MIN 0 #define SCTPCTL_ADD_MORE_ON_OUTPUT_MAX 0xFFFFFFFF #define SCTPCTL_ADD_MORE_ON_OUTPUT_DEFAULT SCTP_DEFAULT_ADD_MORE /* outgoing_streams: Default number of outgoing streams */ #define SCTPCTL_OUTGOING_STREAMS_DESC "Default number of outgoing streams" #define SCTPCTL_OUTGOING_STREAMS_MIN 1 #define SCTPCTL_OUTGOING_STREAMS_MAX 65535 #define SCTPCTL_OUTGOING_STREAMS_DEFAULT SCTP_OSTREAM_INITIAL /* cmt_on_off: CMT on/off flag */ #define SCTPCTL_CMT_ON_OFF_DESC "CMT settings" #define SCTPCTL_CMT_ON_OFF_MIN 0 #define SCTPCTL_CMT_ON_OFF_MAX 2 #define SCTPCTL_CMT_ON_OFF_DEFAULT 0 /* EY - nr_sack_on_off: NR_SACK on/off flag */ #define SCTPCTL_NR_SACK_ON_OFF_DESC "NR_SACK on/off flag" #define SCTPCTL_NR_SACK_ON_OFF_MIN 0 #define SCTPCTL_NR_SACK_ON_OFF_MAX 1 #define SCTPCTL_NR_SACK_ON_OFF_DEFAULT 0 /* cmt_use_dac: CMT DAC on/off flag */ #define SCTPCTL_CMT_USE_DAC_DESC "CMT DAC on/off flag" #define SCTPCTL_CMT_USE_DAC_MIN 0 #define SCTPCTL_CMT_USE_DAC_MAX 1 #define SCTPCTL_CMT_USE_DAC_DEFAULT 0 /* JRS 5/2107 - CMT PF type flag */ #define SCTPCTL_CMT_PF_DESC "CMT PF type flag" #define SCTPCTL_CMT_PF_MIN 0 #define SCTPCTL_CMT_PF_MAX 2 #define SCTPCTL_CMT_PF_DEFAULT 0 /* cwnd_maxburst: Use a CWND adjusting maxburst */ #define SCTPCTL_CWND_MAXBURST_DESC "Use a CWND adjusting maxburst" #define SCTPCTL_CWND_MAXBURST_MIN 0 #define SCTPCTL_CWND_MAXBURST_MAX 1 #define SCTPCTL_CWND_MAXBURST_DEFAULT 1 /* early_fast_retran: Early Fast Retransmit with timer */ #define SCTPCTL_EARLY_FAST_RETRAN_DESC "Early Fast Retransmit with timer" #define SCTPCTL_EARLY_FAST_RETRAN_MIN 0 #define SCTPCTL_EARLY_FAST_RETRAN_MAX 0xFFFFFFFF #define SCTPCTL_EARLY_FAST_RETRAN_DEFAULT 0 /* early_fast_retran_msec: Early Fast Retransmit minimum timer value */ #define SCTPCTL_EARLY_FAST_RETRAN_MSEC_DESC "Early Fast Retransmit minimum timer value" #define SCTPCTL_EARLY_FAST_RETRAN_MSEC_MIN 0 #define SCTPCTL_EARLY_FAST_RETRAN_MSEC_MAX 0xFFFFFFFF #define SCTPCTL_EARLY_FAST_RETRAN_MSEC_DEFAULT SCTP_MINFR_MSEC_TIMER /* asconf_auth_nochk: Disable SCTP ASCONF AUTH requirement */ #define SCTPCTL_ASCONF_AUTH_NOCHK_DESC "Disable SCTP ASCONF AUTH requirement" #define SCTPCTL_ASCONF_AUTH_NOCHK_MIN 0 #define SCTPCTL_ASCONF_AUTH_NOCHK_MAX 1 #define SCTPCTL_ASCONF_AUTH_NOCHK_DEFAULT 0 /* auth_disable: Disable SCTP AUTH function */ #define SCTPCTL_AUTH_DISABLE_DESC "Disable SCTP AUTH function" #define SCTPCTL_AUTH_DISABLE_MIN 0 #define SCTPCTL_AUTH_DISABLE_MAX 1 #define SCTPCTL_AUTH_DISABLE_DEFAULT 0 /* nat_friendly: SCTP NAT friendly operation */ #define SCTPCTL_NAT_FRIENDLY_DESC "SCTP NAT friendly operation" #define SCTPCTL_NAT_FRIENDLY_MIN 0 #define SCTPCTL_NAT_FRIENDLY_MAX 1 #define SCTPCTL_NAT_FRIENDLY_DEFAULT 1 /* abc_l_var: SCTP ABC max increase per SACK (L) */ #define SCTPCTL_ABC_L_VAR_DESC "SCTP ABC max increase per SACK (L)" #define SCTPCTL_ABC_L_VAR_MIN 0 #define SCTPCTL_ABC_L_VAR_MAX 0xFFFFFFFF #define SCTPCTL_ABC_L_VAR_DEFAULT 1 /* max_chained_mbufs: Default max number of small mbufs on a chain */ #define SCTPCTL_MAX_CHAINED_MBUFS_DESC "Default max number of small mbufs on a chain" #define SCTPCTL_MAX_CHAINED_MBUFS_MIN 0 #define SCTPCTL_MAX_CHAINED_MBUFS_MAX 0xFFFFFFFF #define SCTPCTL_MAX_CHAINED_MBUFS_DEFAULT SCTP_DEFAULT_MBUFS_IN_CHAIN /* do_sctp_drain: Should SCTP respond to the drain calls */ #define SCTPCTL_DO_SCTP_DRAIN_DESC "Should SCTP respond to the drain calls" #define SCTPCTL_DO_SCTP_DRAIN_MIN 0 #define SCTPCTL_DO_SCTP_DRAIN_MAX 1 #define SCTPCTL_DO_SCTP_DRAIN_DEFAULT 1 /* hb_max_burst: Confirmation Heartbeat max burst? */ #define SCTPCTL_HB_MAX_BURST_DESC "Confirmation Heartbeat max burst" #define SCTPCTL_HB_MAX_BURST_MIN 1 #define SCTPCTL_HB_MAX_BURST_MAX 0xFFFFFFFF #define SCTPCTL_HB_MAX_BURST_DEFAULT SCTP_DEF_MAX_BURST /* abort_at_limit: When one-2-one hits qlimit abort */ #define SCTPCTL_ABORT_AT_LIMIT_DESC "When one-2-one hits qlimit abort" #define SCTPCTL_ABORT_AT_LIMIT_MIN 0 #define SCTPCTL_ABORT_AT_LIMIT_MAX 1 #define SCTPCTL_ABORT_AT_LIMIT_DEFAULT 0 /* strict_data_order: Enforce strict data ordering, abort if control inside data */ #define SCTPCTL_STRICT_DATA_ORDER_DESC "Enforce strict data ordering, abort if control inside data" #define SCTPCTL_STRICT_DATA_ORDER_MIN 0 #define SCTPCTL_STRICT_DATA_ORDER_MAX 1 #define SCTPCTL_STRICT_DATA_ORDER_DEFAULT 0 /* min_residual: min residual in a data fragment leftover */ #define SCTPCTL_MIN_RESIDUAL_DESC "Minimum residual data chunk in second part of split" #define SCTPCTL_MIN_RESIDUAL_MIN 20 #define SCTPCTL_MIN_RESIDUAL_MAX 65535 #define SCTPCTL_MIN_RESIDUAL_DEFAULT 1452 /* max_retran_chunk: max chunk retransmissions */ #define SCTPCTL_MAX_RETRAN_CHUNK_DESC "Maximum times an unlucky chunk can be retran'd before assoc abort" #define SCTPCTL_MAX_RETRAN_CHUNK_MIN 0 #define SCTPCTL_MAX_RETRAN_CHUNK_MAX 65535 #define SCTPCTL_MAX_RETRAN_CHUNK_DEFAULT 30 /* sctp_logging: This gives us logging when the options are enabled */ #define SCTPCTL_LOGGING_LEVEL_DESC "Ltrace/KTR trace logging level" #define SCTPCTL_LOGGING_LEVEL_MIN 0 #define SCTPCTL_LOGGING_LEVEL_MAX 0xffffffff #define SCTPCTL_LOGGING_LEVEL_DEFAULT 0 /* JRS - default congestion control module sysctl */ #define SCTPCTL_DEFAULT_CC_MODULE_DESC "Default congestion control module" #define SCTPCTL_DEFAULT_CC_MODULE_MIN 0 #define SCTPCTL_DEFAULT_CC_MODULE_MAX 2 #define SCTPCTL_DEFAULT_CC_MODULE_DEFAULT 0 /* RS - default stream scheduling module sysctl */ #define SCTPCTL_DEFAULT_SS_MODULE_DESC "Default stream scheduling module" #define SCTPCTL_DEFAULT_SS_MODULE_MIN 0 #define SCTPCTL_DEFAULT_SS_MODULE_MAX 5 #define SCTPCTL_DEFAULT_SS_MODULE_DEFAULT 0 /* RRS - default fragment interleave */ #define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DESC "Default fragment interleave level" #define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MIN 0 #define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MAX 2 #define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DEFAULT 1 /* mobility_base: Enable SCTP mobility support */ #define SCTPCTL_MOBILITY_BASE_DESC "Enable SCTP base mobility" #define SCTPCTL_MOBILITY_BASE_MIN 0 #define SCTPCTL_MOBILITY_BASE_MAX 1 #define SCTPCTL_MOBILITY_BASE_DEFAULT SCTP_DEFAULT_MOBILITY_BASE /* mobility_fasthandoff: Enable SCTP fast handoff support */ #define SCTPCTL_MOBILITY_FASTHANDOFF_DESC "Enable SCTP fast handoff" #define SCTPCTL_MOBILITY_FASTHANDOFF_MIN 0 #define SCTPCTL_MOBILITY_FASTHANDOFF_MAX 1 #define SCTPCTL_MOBILITY_FASTHANDOFF_DEFAULT SCTP_DEFAULT_MOBILITY_FASTHANDOFF /* Enable SCTP/UDP tunneling for clients*/ #define SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_DESC "Enable SCTP/UDP tunneling for client" #define SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_MIN 0 #define SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_MAX 1 #define SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_DEFAULT SCTPCTL_UDP_TUNNELING_FOR_CLIENT_ENABLE_MIN /* Enable SCTP/UDP tunneling port */ #define SCTPCTL_UDP_TUNNELING_PORT_DESC "Set the SCTP/UDP tunneling port" #define SCTPCTL_UDP_TUNNELING_PORT_MIN 0 #define SCTPCTL_UDP_TUNNELING_PORT_MAX 65535 #define SCTPCTL_UDP_TUNNELING_PORT_DEFAULT SCTP_OVER_UDP_TUNNELING_PORT /* Enable sending of the SACK-IMMEDIATELY bit */ #define SCTPCTL_SACK_IMMEDIATELY_ENABLE_DESC "Enable sending of the SACK-IMMEDIATELY-bit." #define SCTPCTL_SACK_IMMEDIATELY_ENABLE_MIN 0 #define SCTPCTL_SACK_IMMEDIATELY_ENABLE_MAX 1 #define SCTPCTL_SACK_IMMEDIATELY_ENABLE_DEFAULT SCTPCTL_SACK_IMMEDIATELY_ENABLE_MIN /* Enable sending of the NAT-FRIENDLY message */ #define SCTPCTL_NAT_FRIENDLY_INITS_DESC "Enable sending of the nat-friendly SCTP option on INITs." #define SCTPCTL_NAT_FRIENDLY_INITS_MIN 0 #define SCTPCTL_NAT_FRIENDLY_INITS_MAX 1 #define SCTPCTL_NAT_FRIENDLY_INITS_DEFAULT SCTPCTL_NAT_FRIENDLY_INITS_MIN /* Vtag time wait in seconds */ #define SCTPCTL_TIME_WAIT_DESC "Vtag time wait time in seconds, 0 disables it." #define SCTPCTL_TIME_WAIT_MIN 0 #define SCTPCTL_TIME_WAIT_MAX 0xffffffff #define SCTPCTL_TIME_WAIT_DEFAULT SCTP_TIME_WAIT /* Enable Send/Receive buffer splitting */ #define SCTPCTL_BUFFER_SPLITTING_DESC "Enable send/receive buffer splitting." #define SCTPCTL_BUFFER_SPLITTING_MIN 0 #define SCTPCTL_BUFFER_SPLITTING_MAX 0x3 #define SCTPCTL_BUFFER_SPLITTING_DEFAULT SCTPCTL_BUFFER_SPLITTING_MIN /* Initial congestion window in MTU */ #define SCTPCTL_INITIAL_CWND_DESC "Initial congestion window in MTUs" #define SCTPCTL_INITIAL_CWND_MIN 0 #define SCTPCTL_INITIAL_CWND_MAX 0xffffffff #define SCTPCTL_INITIAL_CWND_DEFAULT 3 #if defined(SCTP_DEBUG) /* debug: Configure debug output */ #define SCTPCTL_DEBUG_DESC "Configure debug output" #define SCTPCTL_DEBUG_MIN 0 #define SCTPCTL_DEBUG_MAX 0xFFFFFFFF #define SCTPCTL_DEBUG_DEFAULT 0 #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) #define SCTPCTL_OUTPUT_UNLOCKED_DESC "Unlock socket when sending packets down to IP." #define SCTPCTL_OUTPUT_UNLOCKED_MIN 0 #define SCTPCTL_OUTPUT_UNLOCKED_MAX 1 #define SCTPCTL_OUTPUT_UNLOCKED_DEFAULT SCTPCTL_OUTPUT_UNLOCKED_MIN #endif #if defined(_KERNEL) || defined(__Userspace__) #if defined(SYSCTL_DECL) SYSCTL_DECL(_net_inet_sctp); #endif void sctp_init_sysctls(void); #endif /* _KERNEL */ #endif /* __sctp_sysctl_h__ */ diff --git a/sys/netinet/sctp_timer.c b/sys/netinet/sctp_timer.c index dcf8fd739369..1cb77f0f7c09 100644 --- a/sys/netinet/sctp_timer.c +++ b/sys/netinet/sctp_timer.c @@ -1,1788 +1,1770 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_timer.c,v 1.29 2005/03/06 16:04:18 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #define _IP_VHL #include #include #ifdef INET6 #endif #include #include #include #include #include #include #include #include #include #include #include #include void sctp_early_fr_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_tmit_chunk *chk, *pchk; struct timeval now, min_wait, tv; unsigned int cur_rtt, cnt = 0, cnt_resend = 0; /* an early FR is occuring. */ (void)SCTP_GETTIME_TIMEVAL(&now); /* get cur rto in micro-seconds */ if (net->lastsa == 0) { /* Hmm no rtt estimate yet? */ cur_rtt = stcb->asoc.initial_rto >> 2; } else { cur_rtt = ((net->lastsa >> 2) + net->lastsv) >> 1; } if (cur_rtt < SCTP_BASE_SYSCTL(sctp_early_fr_msec)) { cur_rtt = SCTP_BASE_SYSCTL(sctp_early_fr_msec); } cur_rtt *= 1000; tv.tv_sec = cur_rtt / 1000000; tv.tv_usec = cur_rtt % 1000000; min_wait = now; timevalsub(&min_wait, &tv); if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) { /* * if we hit here, we don't have enough seconds on the clock * to account for the RTO. We just let the lower seconds be * the bounds and don't worry about it. This may mean we * will mark a lot more than we should. */ min_wait.tv_sec = min_wait.tv_usec = 0; } TAILQ_FOREACH_REVERSE_SAFE(chk, &stcb->asoc.sent_queue, sctpchunk_listhead, sctp_next, pchk) { if (chk->whoTo != net) { continue; } if (chk->sent == SCTP_DATAGRAM_RESEND) cnt_resend++; else if ((chk->sent > SCTP_DATAGRAM_UNSENT) && (chk->sent < SCTP_DATAGRAM_RESEND)) { /* pending, may need retran */ if (chk->sent_rcv_time.tv_sec > min_wait.tv_sec) { /* * we have reached a chunk that was sent * some seconds past our min.. forget it we * will find no more to send. */ continue; } else if (chk->sent_rcv_time.tv_sec == min_wait.tv_sec) { /* * we must look at the micro seconds to * know. */ if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) { /* * ok it was sent after our boundary * time. */ continue; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_EARLYFR_LOGGING_ENABLE) { sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count, 4, SCTP_FR_MARKED_EARLY); } SCTP_STAT_INCR(sctps_earlyfrmrkretrans); chk->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); /* double book size since we are doing an early FR */ chk->book_size_scale++; cnt += chk->send_size; if ((cnt + net->flight_size) > net->cwnd) { /* Mark all we could possibly resend */ break; } } } if (cnt) { /* * JRS - Use the congestion control given in the congestion * control module */ stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer(inp, stcb, net); } else if (cnt_resend) { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED); } /* Restart it? */ if (net->flight_size < net->cwnd) { SCTP_STAT_INCR(sctps_earlyfrstrtmr); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net); } } void sctp_audit_retranmission_queue(struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n", asoc->sent_queue_retran_cnt, asoc->sent_queue_cnt); asoc->sent_queue_retran_cnt = 0; asoc->sent_queue_cnt = 0; TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(asoc->sent_queue_retran_cnt); } asoc->sent_queue_cnt++; } TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(asoc->sent_queue_retran_cnt); } } TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(asoc->sent_queue_retran_cnt); } } SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n", asoc->sent_queue_retran_cnt, asoc->sent_queue_cnt); } int sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint16_t threshold) { if (net) { net->error_count++; SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n", net, net->error_count, net->failure_threshold); if (net->error_count > net->failure_threshold) { /* We had a threshold failure */ if (net->dest_state & SCTP_ADDR_REACHABLE) { net->dest_state &= ~SCTP_ADDR_REACHABLE; net->dest_state |= SCTP_ADDR_NOT_REACHABLE; net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY; if (net == stcb->asoc.primary_destination) { net->dest_state |= SCTP_ADDR_WAS_PRIMARY; } /* * JRS 5/14/07 - If a destination is * unreachable, the PF bit is turned off. * This allows an unambiguous use of the PF * bit for destinations that are reachable * but potentially failed. If the * destination is set to the unreachable * state, also set the destination to the PF * state. */ /* * Add debug message here if destination is * not in PF state. */ /* Stop any running T3 timers here? */ if ((stcb->asoc.sctp_cmt_on_off > 0) && (stcb->asoc.sctp_cmt_pf > 0)) { net->dest_state &= ~SCTP_ADDR_PF; SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n", net); } sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, SCTP_FAILED_THRESHOLD, (void *)net, SCTP_SO_NOT_LOCKED); } } /*********HOLD THIS COMMENT FOR PATCH OF ALTERNATE *********ROUTING CODE */ /*********HOLD THIS COMMENT FOR END OF PATCH OF ALTERNATE *********ROUTING CODE */ } if (stcb == NULL) return (0); if (net) { if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_INCR, stcb->asoc.overall_error_count, (stcb->asoc.overall_error_count + 1), SCTP_FROM_SCTP_TIMER, __LINE__); } stcb->asoc.overall_error_count++; } } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { sctp_misc_ints(SCTP_THRESHOLD_INCR, stcb->asoc.overall_error_count, (stcb->asoc.overall_error_count + 1), SCTP_FROM_SCTP_TIMER, __LINE__); } stcb->asoc.overall_error_count++; } SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n", &stcb->asoc, stcb->asoc.overall_error_count, (uint32_t) threshold, ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state)); /* * We specifically do not do >= to give the assoc one more change * before we fail it. */ if (stcb->asoc.overall_error_count > threshold) { /* Abort notification sends a ULP notify */ struct mbuf *oper; oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1); } inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1; sctp_abort_an_association(inp, stcb, SCTP_FAILED_THRESHOLD, oper, SCTP_SO_NOT_LOCKED); return (1); } return (0); } /* * sctp_find_alternate_net() returns a non-NULL pointer as long * the argument net is non-NULL. */ struct sctp_nets * sctp_find_alternate_net(struct sctp_tcb *stcb, struct sctp_nets *net, int mode) { /* Find and return an alternate network if possible */ struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL; int once; /* JRS 5/14/07 - Initialize min_errors to an impossible value. */ int min_errors = -1; uint32_t max_cwnd = 0; if (stcb->asoc.numnets == 1) { /* No others but net */ return (TAILQ_FIRST(&stcb->asoc.nets)); } /* * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate * net algorithm. This algorithm chooses the active destination (not * in PF state) with the largest cwnd value. If all destinations are * in PF state, unreachable, or unconfirmed, choose the desination * that is in PF state with the lowest error count. In case of a * tie, choose the destination that was most recently active. */ if (mode == 2) { TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) { /* * JRS 5/14/07 - If the destination is unreachable * or unconfirmed, skip it. */ if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) || (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) { continue; } /* * JRS 5/14/07 - If the destination is reachable * but in PF state, compare the error count of the * destination to the minimum error count seen thus * far. Store the destination with the lower error * count. If the error counts are equal, store the * destination that was most recently active. */ if (mnet->dest_state & SCTP_ADDR_PF) { /* * JRS 5/14/07 - If the destination under * consideration is the current destination, * work as if the error count is one higher. * The actual error count will not be * incremented until later in the t3 * handler. */ if (mnet == net) { if (min_errors == -1) { min_errors = mnet->error_count + 1; min_errors_net = mnet; } else if (mnet->error_count + 1 < min_errors) { min_errors = mnet->error_count + 1; min_errors_net = mnet; } else if (mnet->error_count + 1 == min_errors && mnet->last_active > min_errors_net->last_active) { min_errors_net = mnet; min_errors = mnet->error_count + 1; } continue; } else { if (min_errors == -1) { min_errors = mnet->error_count; min_errors_net = mnet; } else if (mnet->error_count < min_errors) { min_errors = mnet->error_count; min_errors_net = mnet; } else if (mnet->error_count == min_errors && mnet->last_active > min_errors_net->last_active) { min_errors_net = mnet; min_errors = mnet->error_count; } continue; } } /* * JRS 5/14/07 - If the destination is reachable and * not in PF state, compare the cwnd of the * destination to the highest cwnd seen thus far. * Store the destination with the higher cwnd value. * If the cwnd values are equal, randomly choose one * of the two destinations. */ if (max_cwnd < mnet->cwnd) { max_cwnd_net = mnet; max_cwnd = mnet->cwnd; } else if (max_cwnd == mnet->cwnd) { uint32_t rndval; uint8_t this_random; if (stcb->asoc.hb_random_idx > 3) { rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); this_random = stcb->asoc.hb_random_values[0]; stcb->asoc.hb_random_idx++; stcb->asoc.hb_ect_randombit = 0; } else { this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; stcb->asoc.hb_random_idx++; stcb->asoc.hb_ect_randombit = 0; } if (this_random % 2 == 1) { max_cwnd_net = mnet; max_cwnd = mnet->cwnd; /* Useless? */ } } } /* * JRS 5/14/07 - After all destination have been considered * as alternates, check to see if there was some active * destination (not in PF state). If not, check to see if * there was some PF destination with the minimum number of * errors. If not, return the original destination. If * there is a min_errors_net, remove the PF flag from that * destination, set the cwnd to one or two MTUs, and return * the destination as an alt. If there was some active * destination with a highest cwnd, return the destination * as an alt. */ if (max_cwnd_net == NULL) { if (min_errors_net == NULL) { return (net); } min_errors_net->dest_state &= ~SCTP_ADDR_PF; min_errors_net->cwnd = min_errors_net->mtu * stcb->asoc.sctp_cmt_pf; if (SCTP_OS_TIMER_PENDING(&min_errors_net->rxt_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, min_errors_net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_2); } SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to active with %d errors.\n", min_errors_net, min_errors_net->error_count); return (min_errors_net); } else { return (max_cwnd_net); } } /* * JRS 5/14/07 - If mode is set to 1, use the CMT policy for * choosing an alternate net. */ else if (mode == 1) { TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) { if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) || (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) { /* * will skip ones that are not-reachable or * unconfirmed */ continue; } if (max_cwnd < mnet->cwnd) { max_cwnd_net = mnet; max_cwnd = mnet->cwnd; } else if (max_cwnd == mnet->cwnd) { uint32_t rndval; uint8_t this_random; if (stcb->asoc.hb_random_idx > 3) { rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); this_random = stcb->asoc.hb_random_values[0]; stcb->asoc.hb_random_idx = 0; stcb->asoc.hb_ect_randombit = 0; } else { this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; stcb->asoc.hb_random_idx++; stcb->asoc.hb_ect_randombit = 0; } if (this_random % 2) { max_cwnd_net = mnet; max_cwnd = mnet->cwnd; } } } if (max_cwnd_net) { return (max_cwnd_net); } } mnet = net; once = 0; if (mnet == NULL) { mnet = TAILQ_FIRST(&stcb->asoc.nets); if (mnet == NULL) { return (NULL); } } do { alt = TAILQ_NEXT(mnet, sctp_next); if (alt == NULL) { once++; if (once > 1) { break; } alt = TAILQ_FIRST(&stcb->asoc.nets); if (alt == NULL) { return (NULL); } } if (alt->ro.ro_rt == NULL) { if (alt->ro._s_addr) { sctp_free_ifa(alt->ro._s_addr); alt->ro._s_addr = NULL; } alt->src_addr_selected = 0; } /* sa_ignore NO_NULL_CHK */ if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) && (alt->ro.ro_rt != NULL) && (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) { /* Found a reachable address */ break; } mnet = alt; } while (alt != NULL); if (alt == NULL) { /* Case where NO insv network exists (dormant state) */ /* we rotate destinations */ once = 0; mnet = net; do { if (mnet == NULL) { return (TAILQ_FIRST(&stcb->asoc.nets)); } alt = TAILQ_NEXT(mnet, sctp_next); if (alt == NULL) { once++; if (once > 1) { break; } alt = TAILQ_FIRST(&stcb->asoc.nets); } /* sa_ignore NO_NULL_CHK */ if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) && (alt != net)) { /* Found an alternate address */ break; } mnet = alt; } while (alt != NULL); } if (alt == NULL) { return (net); } return (alt); } static void sctp_backoff_on_timeout(struct sctp_tcb *stcb, struct sctp_nets *net, int win_probe, int num_marked, int num_abandoned) { if (net->RTO == 0) { net->RTO = stcb->asoc.minrto; } net->RTO <<= 1; if (net->RTO > stcb->asoc.maxrto) { net->RTO = stcb->asoc.maxrto; } if ((win_probe == 0) && (num_marked || num_abandoned)) { /* We don't apply penalty to window probe scenarios */ /* JRS - Use the congestion control given in the CC module */ stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net); } } #ifndef INVARIANTS static void sctp_recover_sent_list(struct sctp_tcb *stcb) { struct sctp_tmit_chunk *chk, *nchk; struct sctp_association *asoc; asoc = &stcb->asoc; TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) { SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n", chk, chk->rec.data.TSN_seq, asoc->last_acked_seq); TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); if (chk->pr_sctp_on) { if (asoc->pr_sctp_cnt != 0) asoc->pr_sctp_cnt--; } if (chk->data) { /* sa_ignore NO_NULL_CHK */ sctp_free_bufspace(stcb, asoc, chk, 1); sctp_m_freem(chk->data); chk->data = NULL; if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) { asoc->sent_queue_cnt_removeable--; } } asoc->sent_queue_cnt--; sctp_free_a_chunk(stcb, chk); } } SCTP_PRINTF("after recover order is as follows\n"); TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { SCTP_PRINTF("chk:%p TSN:%x\n", chk, chk->rec.data.TSN_seq); } } #endif static int sctp_mark_all_for_resend(struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_nets *alt, int window_probe, int *num_marked, int *num_abandoned) { /* * Mark all chunks (well not all) that were sent to *net for * retransmission. Move them to alt for there destination as well... * We only mark chunks that have been outstanding long enough to * have received feed-back. */ struct sctp_tmit_chunk *chk, *nchk; struct sctp_nets *lnets; struct timeval now, min_wait, tv; int cur_rtt; int cnt_abandoned; int audit_tf, num_mk, fir; unsigned int cnt_mk; uint32_t orig_flight, orig_tf; uint32_t tsnlast, tsnfirst; int recovery_cnt = 0; /* none in flight now */ audit_tf = 0; fir = 0; /* * figure out how long a data chunk must be pending before we can * mark it .. */ (void)SCTP_GETTIME_TIMEVAL(&now); /* get cur rto in micro-seconds */ cur_rtt = (((net->lastsa >> 2) + net->lastsv) >> 1); cur_rtt *= 1000; if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(cur_rtt, stcb->asoc.peers_rwnd, window_probe, SCTP_FR_T3_MARK_TIME); sctp_log_fr(net->flight_size, SCTP_OS_TIMER_PENDING(&net->fr_timer.timer), SCTP_OS_TIMER_ACTIVE(&net->fr_timer.timer), SCTP_FR_CWND_REPORT); sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT); } tv.tv_sec = cur_rtt / 1000000; tv.tv_usec = cur_rtt % 1000000; min_wait = now; timevalsub(&min_wait, &tv); if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) { /* * if we hit here, we don't have enough seconds on the clock * to account for the RTO. We just let the lower seconds be * the bounds and don't worry about it. This may mean we * will mark a lot more than we should. */ min_wait.tv_sec = min_wait.tv_usec = 0; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(cur_rtt, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME); sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME); } /* * Our rwnd will be incorrect here since we are not adding back the * cnt * mbuf but we will fix that down below. */ orig_flight = net->flight_size; orig_tf = stcb->asoc.total_flight; net->fast_retran_ip = 0; /* Now on to each chunk */ cnt_abandoned = 0; num_mk = cnt_mk = 0; tsnfirst = tsnlast = 0; #ifndef INVARIANTS start_again: #endif TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) { if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) { /* Strange case our list got out of order? */ SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x", (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq); recovery_cnt++; #ifdef INVARIANTS panic("last acked >= chk on sent-Q"); #else SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt); sctp_recover_sent_list(stcb); if (recovery_cnt < 10) { goto start_again; } else { SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt); } #endif } if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) { /* * found one to mark: If it is less than * DATAGRAM_ACKED it MUST not be a skipped or marked * TSN but instead one that is either already set * for retransmission OR one that needs * retransmission. */ /* validate its been outstanding long enough */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(chk->rec.data.TSN_seq, chk->sent_rcv_time.tv_sec, chk->sent_rcv_time.tv_usec, SCTP_FR_T3_MARK_TIME); } if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) { /* * we have reached a chunk that was sent * some seconds past our min.. forget it we * will find no more to send. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(0, chk->sent_rcv_time.tv_sec, chk->sent_rcv_time.tv_usec, SCTP_FR_T3_STOPPED); } continue; } else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) && (window_probe == 0)) { /* * we must look at the micro seconds to * know. */ if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) { /* * ok it was sent after our boundary * time. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(0, chk->sent_rcv_time.tv_sec, chk->sent_rcv_time.tv_usec, SCTP_FR_T3_STOPPED); } continue; } } if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) { /* Is it expired? */ if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) { /* Yes so drop it */ if (chk->data) { (void)sctp_release_pr_sctp_chunk(stcb, chk, (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT), SCTP_SO_NOT_LOCKED); cnt_abandoned++; } continue; } } if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) { /* Has it been retransmitted tv_sec times? */ if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) { if (chk->data) { (void)sctp_release_pr_sctp_chunk(stcb, chk, (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT), SCTP_SO_NOT_LOCKED); cnt_abandoned++; } continue; } } if (chk->sent < SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); num_mk++; if (fir == 0) { fir = 1; tsnfirst = chk->rec.data.TSN_seq; } tsnlast = chk->rec.data.TSN_seq; if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count, 0, SCTP_FR_T3_MARKED); } if (chk->rec.data.chunk_was_revoked) { /* deflate the cwnd */ chk->whoTo->cwnd -= chk->book_size; chk->rec.data.chunk_was_revoked = 0; } net->marked_retrans++; stcb->asoc.marked_retrans++; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO, chk->whoTo->flight_size, chk->book_size, (uintptr_t) chk->whoTo, chk->rec.data.TSN_seq); } sctp_flight_size_decrease(chk); sctp_total_flight_decrease(stcb, chk); stcb->asoc.peers_rwnd += chk->send_size; stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh); } chk->sent = SCTP_DATAGRAM_RESEND; SCTP_STAT_INCR(sctps_markedretrans); /* reset the TSN for striking and other FR stuff */ chk->rec.data.doing_fast_retransmit = 0; /* Clear any time so NO RTT is being done */ chk->do_rtt = 0; if (alt != net) { sctp_free_remote_addr(chk->whoTo); chk->no_fr_allowed = 1; chk->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } else { chk->no_fr_allowed = 0; if (TAILQ_EMPTY(&stcb->asoc.send_queue)) { chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq; } else { chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq; } } /* * CMT: Do not allow FRs on retransmitted TSNs. */ if (stcb->asoc.sctp_cmt_on_off > 0) { chk->no_fr_allowed = 1; } #ifdef THIS_SHOULD_NOT_BE_DONE } else if (chk->sent == SCTP_DATAGRAM_ACKED) { /* remember highest acked one */ could_be_sent = chk; #endif } if (chk->sent == SCTP_DATAGRAM_RESEND) { cnt_mk++; } } if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) { /* we did not subtract the same things? */ audit_tf = 1; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) { sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT); } #ifdef SCTP_DEBUG if (num_mk) { SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n", tsnlast); SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n", num_mk, (u_long)stcb->asoc.peers_rwnd); SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n", tsnlast); SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n", num_mk, (int)stcb->asoc.peers_rwnd); } #endif *num_marked = num_mk; *num_abandoned = cnt_abandoned; /* * Now check for a ECN Echo that may be stranded And include the * cnt_mk'd to have all resends in the control queue. */ TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { cnt_mk++; } if ((chk->whoTo == net) && (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = alt; if (chk->sent != SCTP_DATAGRAM_RESEND) { chk->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); cnt_mk++; } atomic_add_int(&alt->ref_count, 1); } } #ifdef THIS_SHOULD_NOT_BE_DONE if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) { /* fix it so we retransmit the highest acked anyway */ sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); cnt_mk++; could_be_sent->sent = SCTP_DATAGRAM_RESEND; } #endif if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) { #ifdef INVARIANTS SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n", cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk); #endif #ifndef SCTP_AUDITING_ENABLED stcb->asoc.sent_queue_retran_cnt = cnt_mk; #endif } if (audit_tf) { SCTPDBG(SCTP_DEBUG_TIMER4, "Audit total flight due to negative value net:%p\n", net); stcb->asoc.total_flight = 0; stcb->asoc.total_flight_count = 0; /* Clear all networks flight size */ TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) { lnets->flight_size = 0; SCTPDBG(SCTP_DEBUG_TIMER4, "Net:%p c-f cwnd:%d ssthresh:%d\n", lnets, lnets->cwnd, lnets->ssthresh); } TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if (chk->sent < SCTP_DATAGRAM_RESEND) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP, chk->whoTo->flight_size, chk->book_size, (uintptr_t) chk->whoTo, chk->rec.data.TSN_seq); } sctp_flight_size_increase(chk); sctp_total_flight_increase(stcb, chk); } } } - /* - * Setup the ecn nonce re-sync point. We do this since - * retranmissions are NOT setup for ECN. This means that do to - * Karn's rule, we don't know the total of the peers ecn bits. - */ - chk = TAILQ_FIRST(&stcb->asoc.send_queue); - if (chk == NULL) { - stcb->asoc.nonce_resync_tsn = stcb->asoc.sending_seq; - } else { - stcb->asoc.nonce_resync_tsn = chk->rec.data.TSN_seq; - } - stcb->asoc.nonce_wait_for_ecne = 0; - stcb->asoc.nonce_sum_check = 0; /* We return 1 if we only have a window probe outstanding */ return (0); } int sctp_t3rxt_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; int win_probe, num_mk, num_abandoned; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { struct sctp_nets *lnet; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if (net == lnet) { sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3); } else { sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3); } } } /* Find an alternate and mark those for retransmission */ if ((stcb->asoc.peers_rwnd == 0) && (stcb->asoc.total_flight < net->mtu)) { SCTP_STAT_INCR(sctps_timowindowprobe); win_probe = 1; } else { win_probe = 0; } /* * JRS 5/14/07 - If CMT PF is on and the destination if not already * in PF state, set the destination to PF state and store the * current time as the time that the destination was last active. In * addition, find an alternate destination with PF-based * find_alt_net(). */ if ((stcb->asoc.sctp_cmt_on_off > 0) && (stcb->asoc.sctp_cmt_pf > 0)) { if ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF) { net->dest_state |= SCTP_ADDR_PF; net->last_active = sctp_get_tick_count(); SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from active to PF.\n", net); } alt = sctp_find_alternate_net(stcb, net, 2); } else if (stcb->asoc.sctp_cmt_on_off > 0) { /* * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is being * used, then pick dest with largest ssthresh for any * retransmission. */ alt = sctp_find_alternate_net(stcb, net, 1); /* * CUCv2: If a different dest is picked for the * retransmission, then new (rtx-)pseudo_cumack needs to be * tracked for orig dest. Let CUCv2 track new (rtx-) * pseudo-cumack always. */ net->find_pseudo_cumack = 1; net->find_rtx_pseudo_cumack = 1; } else { /* CMT is OFF */ alt = sctp_find_alternate_net(stcb, net, 0); } num_mk = 0; num_abandoned = 0; (void)sctp_mark_all_for_resend(stcb, net, alt, win_probe, &num_mk, &num_abandoned); /* FR Loss recovery just ended with the T3. */ stcb->asoc.fast_retran_loss_recovery = 0; /* CMT FR loss recovery ended with the T3 */ net->fast_retran_loss_recovery = 0; /* * setup the sat loss recovery that prevents satellite cwnd advance. */ stcb->asoc.sat_t3_loss_recovery = 1; stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq; /* Backoff the timer and cwnd */ sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned); if (win_probe == 0) { /* We don't do normal threshold management on window probes */ if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { /* Association was destroyed */ return (1); } else { if (net != stcb->asoc.primary_destination) { /* send a immediate HB if our RTO is stale */ struct timeval now; unsigned int ms_goneby; (void)SCTP_GETTIME_TIMEVAL(&now); if (net->last_sent_time.tv_sec) { ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000; } else { ms_goneby = 0; } if ((ms_goneby > net->RTO) || (net->RTO == 0)) { /* * no recent feed back in an RTO or * more, request a RTT update */ if (sctp_send_hb(stcb, 1, net) < 0) /* * Less than 0 means we lost * the assoc */ return (1); } } } } else { /* * For a window probe we don't penalize the net's but only * the association. This may fail it if SACKs are not coming * back. If sack's are coming with rwnd locked at 0, we will * continue to hold things waiting for rwnd to raise */ if (sctp_threshold_management(inp, stcb, NULL, stcb->asoc.max_send_times)) { /* Association was destroyed */ return (1); } } if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) { /* Move all pending over too */ sctp_move_chunks_from_net(stcb, net); /* * Get the address that failed, to force a new src address * selecton and a route allocation. */ if (net->ro._s_addr) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; /* Force a route allocation too */ if (net->ro.ro_rt) { RTFREE(net->ro.ro_rt); net->ro.ro_rt = NULL; } /* Was it our primary? */ if ((stcb->asoc.primary_destination == net) && (alt != net)) { /* * Yes, note it as such and find an alternate note: * this means HB code must use this to resent the * primary if it goes active AND if someone does a * change-primary then this flag must be cleared * from any net structures. */ if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, alt) == 0) { net->dest_state |= SCTP_ADDR_WAS_PRIMARY; } } } else if ((stcb->asoc.sctp_cmt_on_off > 0) && (stcb->asoc.sctp_cmt_pf > 0) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { /* * JRS 5/14/07 - If the destination hasn't failed completely * but is in PF state, a PF-heartbeat needs to be sent * manually. */ if (sctp_send_hb(stcb, 1, net) < 0) /* Return less than 0 means we lost the association */ return (1); } /* * Special case for cookie-echo'ed case, we don't do output but must * await the COOKIE-ACK before retransmission */ if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) { /* * Here we just reset the timer and start again since we * have not established the asoc */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); return (0); } if (stcb->asoc.peer_supports_prsctp) { struct sctp_tmit_chunk *lchk; lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc); /* C3. See if we need to send a Fwd-TSN */ if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) { - /* - * ISSUE with ECN, see FWD-TSN processing for notes - * on issues that will occur when the ECN NONCE - * stuff is put into SCTP for cross checking. - */ send_forward_tsn(stcb, &stcb->asoc); if (lchk) { /* Assure a timer is up */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo); } } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX); } return (0); } int sctp_t1init_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { /* bump the thresholds */ if (stcb->asoc.delayed_connection) { /* * special hook for delayed connection. The library did NOT * complete the rest of its sends. */ stcb->asoc.delayed_connection = 0; sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED); return (0); } if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) { return (0); } if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_init_times)) { /* Association was destroyed */ return (1); } stcb->asoc.dropped_special_cnt = 0; sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0); if (stcb->asoc.initial_init_rto_max < net->RTO) { net->RTO = stcb->asoc.initial_init_rto_max; } if (stcb->asoc.numnets > 1) { /* If we have more than one addr use it */ struct sctp_nets *alt; alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0); if (alt != stcb->asoc.primary_destination) { sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination); stcb->asoc.primary_destination = alt; } } /* Send out a new init */ sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED); return (0); } /* * For cookie and asconf we actually need to find and mark for resend, then * increment the resend counter (after all the threshold management stuff of * course). */ int sctp_cookie_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; struct sctp_tmit_chunk *cookie; /* first before all else we must find the cookie */ TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) { if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) { break; } } if (cookie == NULL) { if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) { /* FOOBAR! */ struct mbuf *oper; oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (oper) { struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(oper, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); ph->param_length = htons(SCTP_BUF_LEN(oper)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3); } inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4; sctp_abort_an_association(inp, stcb, SCTP_INTERNAL_ERROR, oper, SCTP_SO_NOT_LOCKED); } else { #ifdef INVARIANTS panic("Cookie timer expires in wrong state?"); #else SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc)); return (0); #endif } return (0); } /* Ok we found the cookie, threshold management next */ if (sctp_threshold_management(inp, stcb, cookie->whoTo, stcb->asoc.max_init_times)) { /* Assoc is over */ return (1); } /* * cleared theshold management now lets backoff the address & select * an alternate */ stcb->asoc.dropped_special_cnt = 0; sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0); alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0); if (alt != cookie->whoTo) { sctp_free_remote_addr(cookie->whoTo); cookie->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } /* Now mark the retran info */ if (cookie->sent != SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); } cookie->sent = SCTP_DATAGRAM_RESEND; /* * Now call the output routine to kick out the cookie again, Note we * don't mark any chunks for retran so that FR will need to kick in * to move these (or a send timer). */ return (0); } int sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; struct sctp_tmit_chunk *strrst = NULL, *chk = NULL; if (stcb->asoc.stream_reset_outstanding == 0) { return (0); } /* find the existing STRRESET, we use the seq number we sent out on */ (void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst); if (strrst == NULL) { return (0); } /* do threshold management */ if (sctp_threshold_management(inp, stcb, strrst->whoTo, stcb->asoc.max_send_times)) { /* Assoc is over */ return (1); } /* * cleared theshold management now lets backoff the address & select * an alternate */ sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0); alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0); sctp_free_remote_addr(strrst->whoTo); strrst->whoTo = alt; atomic_add_int(&alt->ref_count, 1); /* See if a ECN Echo is also stranded */ TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { if ((chk->whoTo == net) && (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { sctp_free_remote_addr(chk->whoTo); if (chk->sent != SCTP_DATAGRAM_RESEND) { chk->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); } chk->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } } if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) { /* * If the address went un-reachable, we need to move to * alternates for ALL chk's in queue */ sctp_move_chunks_from_net(stcb, net); } /* mark the retran info */ if (strrst->sent != SCTP_DATAGRAM_RESEND) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); strrst->sent = SCTP_DATAGRAM_RESEND; /* restart the timer */ sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo); return (0); } int sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; struct sctp_tmit_chunk *asconf, *chk; /* is this a first send, or a retransmission? */ if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) { /* compose a new ASCONF chunk and send it */ sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED); } else { /* * Retransmission of the existing ASCONF is needed */ /* find the existing ASCONF */ asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue); if (asconf == NULL) { return (0); } /* do threshold management */ if (sctp_threshold_management(inp, stcb, asconf->whoTo, stcb->asoc.max_send_times)) { /* Assoc is over */ return (1); } if (asconf->snd_count > stcb->asoc.max_send_times) { /* * Something is rotten: our peer is not responding * to ASCONFs but apparently is to other chunks. * i.e. it is not properly handling the chunk type * upper bits. Mark this peer as ASCONF incapable * and cleanup. */ SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n"); sctp_asconf_cleanup(stcb, net); return (0); } /* * cleared threshold management, so now backoff the net and * select an alternate */ sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0); alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0); if (asconf->whoTo != alt) { sctp_free_remote_addr(asconf->whoTo); asconf->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } /* See if an ECN Echo is also stranded */ TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { if ((chk->whoTo == net) && (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = alt; if (chk->sent != SCTP_DATAGRAM_RESEND) { chk->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); } atomic_add_int(&alt->ref_count, 1); } } TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) { if (chk->whoTo != alt) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = alt; atomic_add_int(&alt->ref_count, 1); } if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); chk->sent = SCTP_DATAGRAM_RESEND; } if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) { /* * If the address went un-reachable, we need to move * to the alternate for ALL chunks in queue */ sctp_move_chunks_from_net(stcb, net); } /* mark the retran info */ if (asconf->sent != SCTP_DATAGRAM_RESEND) sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); asconf->sent = SCTP_DATAGRAM_RESEND; /* send another ASCONF if any and we can do */ sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED); } return (0); } /* Mobility adaptation */ void sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { if (stcb->asoc.deleted_primary == NULL) { SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n"); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); return; } SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary "); SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa); sctp_free_remote_addr(stcb->asoc.deleted_primary); stcb->asoc.deleted_primary = NULL; sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); return; } /* * For the shutdown and shutdown-ack, we do not keep one around on the * control queue. This means we must generate a new one and call the general * chunk output routine, AFTER having done threshold management. * It is assumed that net is non-NULL. */ int sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; /* first threshold managment */ if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { /* Assoc is over */ return (1); } sctp_backoff_on_timeout(stcb, net, 1, 0, 0); /* second select an alternative */ alt = sctp_find_alternate_net(stcb, net, 0); /* third generate a shutdown into the queue for out net */ sctp_send_shutdown(stcb, alt); /* fourth restart timer */ sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt); return (0); } int sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_nets *alt; /* first threshold managment */ if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { /* Assoc is over */ return (1); } sctp_backoff_on_timeout(stcb, net, 1, 0, 0); /* second select an alternative */ alt = sctp_find_alternate_net(stcb, net, 0); /* third generate a shutdown into the queue for out net */ sctp_send_shutdown_ack(stcb, alt); /* fourth restart timer */ sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt); return (0); } static void sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { struct sctp_stream_queue_pending *sp; unsigned int i, chks_in_queue = 0; int being_filled = 0; /* * This function is ONLY called when the send/sent queues are empty. */ if ((stcb == NULL) || (inp == NULL)) return; if (stcb->asoc.sent_queue_retran_cnt) { SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n", stcb->asoc.sent_queue_retran_cnt); stcb->asoc.sent_queue_retran_cnt = 0; } SCTP_TCB_SEND_LOCK(stcb); if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) { int cnt = 0; /* Check to see if a spoke fell off the wheel */ for (i = 0; i < stcb->asoc.streamoutcnt; i++) { if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { stcb->asoc.ss_functions.sctp_ss_add_to_stream(stcb, &stcb->asoc, &stcb->asoc.strmout[i], NULL, 1); cnt++; } } if (cnt) { /* yep, we lost a spoke or two */ SCTP_PRINTF("Found an additional %d streams NOT on outwheel, corrected\n", cnt); } else { /* no spokes lost, */ stcb->asoc.total_output_queue_size = 0; } SCTP_TCB_SEND_UNLOCK(stcb); return; } SCTP_TCB_SEND_UNLOCK(stcb); /* Check to see if some data queued, if so report it */ for (i = 0; i < stcb->asoc.streamoutcnt; i++) { if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) { if (sp->msg_is_complete) being_filled++; chks_in_queue++; } } } if (chks_in_queue != stcb->asoc.stream_queue_cnt) { SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n", stcb->asoc.stream_queue_cnt, chks_in_queue); } if (chks_in_queue) { /* call the output queue function */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) && (TAILQ_EMPTY(&stcb->asoc.sent_queue))) { /* * Probably should go in and make it go back through * and add fragments allowed */ if (being_filled == 0) { SCTP_PRINTF("Still nothing moved %d chunks are stuck\n", chks_in_queue); } } } else { SCTP_PRINTF("Found no chunks on any queue tot:%lu\n", (u_long)stcb->asoc.total_output_queue_size); stcb->asoc.total_output_queue_size = 0; } } int sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, int cnt_of_unconf) { int ret; if (net) { if (net->hb_responded == 0) { if (net->ro._s_addr) { /* * Invalidate the src address if we did not * get a response last time. */ sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } sctp_backoff_on_timeout(stcb, net, 1, 0, 0); } /* Zero PBA, if it needs it */ if (net->partial_bytes_acked) { net->partial_bytes_acked = 0; } } if ((stcb->asoc.total_output_queue_size > 0) && (TAILQ_EMPTY(&stcb->asoc.send_queue)) && (TAILQ_EMPTY(&stcb->asoc.sent_queue))) { sctp_audit_stream_queues_for_size(inp, stcb); } /* Send a new HB, this will do threshold managment, pick a new dest */ if (cnt_of_unconf == 0) { if (sctp_send_hb(stcb, 0, NULL) < 0) { return (1); } } else { /* * this will send out extra hb's up to maxburst if there are * any unconfirmed addresses. */ uint32_t cnt_sent = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) && (net->dest_state & SCTP_ADDR_REACHABLE)) { cnt_sent++; if (net->hb_responded == 0) { /* Did we respond last time? */ if (net->ro._s_addr) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } } ret = sctp_send_hb(stcb, 1, net); if (ret < 0) return 1; else if (ret == 0) { break; } if (cnt_sent >= SCTP_BASE_SYSCTL(sctp_hb_maxburst)) break; } } } return (0); } void sctp_pathmtu_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { uint32_t next_mtu, mtu; next_mtu = sctp_get_next_mtu(inp, net->mtu); if ((next_mtu > net->mtu) && (net->port == 0)) { if ((net->src_addr_selected == 0) || (net->ro._s_addr == NULL) || (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) { if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } else if (net->ro._s_addr == NULL) { #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE) if (net->ro._l_addr.sa.sa_family == AF_INET6) { struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; /* KAME hack: embed scopeid */ (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); } #endif net->ro._s_addr = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, stcb->asoc.vrf_id); #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE) if (net->ro._l_addr.sa.sa_family == AF_INET6) { struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_recoverscope(sin6); } #endif /* INET6 */ } if (net->ro._s_addr) net->src_addr_selected = 1; } if (net->ro._s_addr) { mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt); if (net->port) { mtu -= sizeof(struct udphdr); } if (mtu > next_mtu) { net->mtu = next_mtu; } } } /* restart the timer */ sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); } void sctp_autoclose_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { struct timeval tn, *tim_touse; struct sctp_association *asoc; int ticks_gone_by; (void)SCTP_GETTIME_TIMEVAL(&tn); if (stcb->asoc.sctp_autoclose_ticks && sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { /* Auto close is on */ asoc = &stcb->asoc; /* pick the time to use */ if (asoc->time_last_rcvd.tv_sec > asoc->time_last_sent.tv_sec) { tim_touse = &asoc->time_last_rcvd; } else { tim_touse = &asoc->time_last_sent; } /* Now has long enough transpired to autoclose? */ ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec); if ((ticks_gone_by > 0) && (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) { /* * autoclose time has hit, call the output routine, * which should do nothing just to be SURE we don't * have hanging data. We can then safely check the * queues and know that we are clear to send * shutdown */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED); /* Are we clean? */ if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) { /* * there is nothing queued to send, so I'm * done... */ if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) { /* only send SHUTDOWN 1st time thru */ sctp_send_shutdown(stcb, stcb->asoc.primary_destination); if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } } else { /* * No auto close at this time, reset t-o to check * later */ int tmp; /* fool the timer startup to use the time left */ tmp = asoc->sctp_autoclose_ticks; asoc->sctp_autoclose_ticks -= ticks_gone_by; sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, net); /* restore the real tick value */ asoc->sctp_autoclose_ticks = tmp; } } } diff --git a/sys/netinet/sctputil.c b/sys/netinet/sctputil.c index 5917578095fc..823eea2dc5cc 100644 --- a/sys/netinet/sctputil.c +++ b/sys/netinet/sctputil.c @@ -1,6900 +1,6893 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctputil.c,v 1.37 2005/03/07 23:26:09 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #ifdef INET6 #endif #include #include #include #include #include /* for sctp_deliver_data() */ #include #include #include #ifndef KTR_SCTP #define KTR_SCTP KTR_SUBSYS #endif extern struct sctp_cc_functions sctp_cc_functions[]; extern struct sctp_ss_functions sctp_ss_functions[]; void sctp_sblog(struct sockbuf *sb, struct sctp_tcb *stcb, int from, int incr) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.sb.stcb = stcb; sctp_clog.x.sb.so_sbcc = sb->sb_cc; if (stcb) sctp_clog.x.sb.stcb_sbcc = stcb->asoc.sb_cc; else sctp_clog.x.sb.stcb_sbcc = 0; sctp_clog.x.sb.incr = incr; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_SB, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_closing(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int16_t loc) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.close.inp = (void *)inp; sctp_clog.x.close.sctp_flags = inp->sctp_flags; if (stcb) { sctp_clog.x.close.stcb = (void *)stcb; sctp_clog.x.close.state = (uint16_t) stcb->asoc.state; } else { sctp_clog.x.close.stcb = 0; sctp_clog.x.close.state = 0; } sctp_clog.x.close.loc = loc; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_CLOSE, 0, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void rto_logging(struct sctp_nets *net, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.rto.net = (void *)net; sctp_clog.x.rto.rtt = net->prev_rtt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RTT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_strm_del_alt(struct sctp_tcb *stcb, uint32_t tsn, uint16_t sseq, uint16_t stream, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.strlog.stcb = stcb; sctp_clog.x.strlog.n_tsn = tsn; sctp_clog.x.strlog.n_sseq = sseq; sctp_clog.x.strlog.e_tsn = 0; sctp_clog.x.strlog.e_sseq = 0; sctp_clog.x.strlog.strm = stream; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_STRM, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_nagle_event(struct sctp_tcb *stcb, int action) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.nagle.stcb = (void *)stcb; sctp_clog.x.nagle.total_flight = stcb->asoc.total_flight; sctp_clog.x.nagle.total_in_queue = stcb->asoc.total_output_queue_size; sctp_clog.x.nagle.count_in_queue = stcb->asoc.chunks_on_out_queue; sctp_clog.x.nagle.count_in_flight = stcb->asoc.total_flight_count; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_NAGLE, action, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_sack(uint32_t old_cumack, uint32_t cumack, uint32_t tsn, uint16_t gaps, uint16_t dups, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.sack.cumack = cumack; sctp_clog.x.sack.oldcumack = old_cumack; sctp_clog.x.sack.tsn = tsn; sctp_clog.x.sack.numGaps = gaps; sctp_clog.x.sack.numDups = dups; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_SACK, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_map(uint32_t map, uint32_t cum, uint32_t high, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.map.base = map; sctp_clog.x.map.cum = cum; sctp_clog.x.map.high = high; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MAP, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_fr(uint32_t biggest_tsn, uint32_t biggest_new_tsn, uint32_t tsn, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.fr.largest_tsn = biggest_tsn; sctp_clog.x.fr.largest_new_tsn = biggest_new_tsn; sctp_clog.x.fr.tsn = tsn; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_FR, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_mb(struct mbuf *m, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.mb.mp = m; sctp_clog.x.mb.mbuf_flags = (uint8_t) (SCTP_BUF_GET_FLAGS(m)); sctp_clog.x.mb.size = (uint16_t) (SCTP_BUF_LEN(m)); sctp_clog.x.mb.data = SCTP_BUF_AT(m, 0); if (SCTP_BUF_IS_EXTENDED(m)) { sctp_clog.x.mb.ext = SCTP_BUF_EXTEND_BASE(m); sctp_clog.x.mb.refcnt = (uint8_t) (SCTP_BUF_EXTEND_REFCNT(m)); } else { sctp_clog.x.mb.ext = 0; sctp_clog.x.mb.refcnt = 0; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MBUF, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_strm_del(struct sctp_queued_to_read *control, struct sctp_queued_to_read *poschk, int from) { struct sctp_cwnd_log sctp_clog; if (control == NULL) { SCTP_PRINTF("Gak log of NULL?\n"); return; } sctp_clog.x.strlog.stcb = control->stcb; sctp_clog.x.strlog.n_tsn = control->sinfo_tsn; sctp_clog.x.strlog.n_sseq = control->sinfo_ssn; sctp_clog.x.strlog.strm = control->sinfo_stream; if (poschk != NULL) { sctp_clog.x.strlog.e_tsn = poschk->sinfo_tsn; sctp_clog.x.strlog.e_sseq = poschk->sinfo_ssn; } else { sctp_clog.x.strlog.e_tsn = 0; sctp_clog.x.strlog.e_sseq = 0; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_STRM, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net, int augment, uint8_t from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.cwnd.net = net; if (stcb->asoc.send_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_send = 255; else sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; if (stcb->asoc.stream_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_str = 255; else sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; if (net) { sctp_clog.x.cwnd.cwnd_new_value = net->cwnd; sctp_clog.x.cwnd.inflight = net->flight_size; sctp_clog.x.cwnd.pseudo_cumack = net->pseudo_cumack; sctp_clog.x.cwnd.meets_pseudo_cumack = net->new_pseudo_cumack; sctp_clog.x.cwnd.need_new_pseudo_cumack = net->find_pseudo_cumack; } if (SCTP_CWNDLOG_PRESEND == from) { sctp_clog.x.cwnd.meets_pseudo_cumack = stcb->asoc.peers_rwnd; } sctp_clog.x.cwnd.cwnd_augment = augment; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_CWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_lock(struct sctp_inpcb *inp, struct sctp_tcb *stcb, uint8_t from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); if (inp) { sctp_clog.x.lock.sock = (void *)inp->sctp_socket; } else { sctp_clog.x.lock.sock = (void *)NULL; } sctp_clog.x.lock.inp = (void *)inp; if (stcb) { sctp_clog.x.lock.tcb_lock = mtx_owned(&stcb->tcb_mtx); } else { sctp_clog.x.lock.tcb_lock = SCTP_LOCK_UNKNOWN; } if (inp) { sctp_clog.x.lock.inp_lock = mtx_owned(&inp->inp_mtx); sctp_clog.x.lock.create_lock = mtx_owned(&inp->inp_create_mtx); } else { sctp_clog.x.lock.inp_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.create_lock = SCTP_LOCK_UNKNOWN; } sctp_clog.x.lock.info_lock = rw_wowned(&SCTP_BASE_INFO(ipi_ep_mtx)); if (inp && (inp->sctp_socket)) { sctp_clog.x.lock.sock_lock = mtx_owned(&(inp->sctp_socket->so_rcv.sb_mtx)); sctp_clog.x.lock.sockrcvbuf_lock = mtx_owned(&(inp->sctp_socket->so_rcv.sb_mtx)); sctp_clog.x.lock.socksndbuf_lock = mtx_owned(&(inp->sctp_socket->so_snd.sb_mtx)); } else { sctp_clog.x.lock.sock_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.sockrcvbuf_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.socksndbuf_lock = SCTP_LOCK_UNKNOWN; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_LOCK_EVENT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_maxburst(struct sctp_tcb *stcb, struct sctp_nets *net, int error, int burst, uint8_t from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.cwnd.net = net; sctp_clog.x.cwnd.cwnd_new_value = error; sctp_clog.x.cwnd.inflight = net->flight_size; sctp_clog.x.cwnd.cwnd_augment = burst; if (stcb->asoc.send_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_send = 255; else sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; if (stcb->asoc.stream_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_str = 255; else sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MAXBURST, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_rwnd(uint8_t from, uint32_t peers_rwnd, uint32_t snd_size, uint32_t overhead) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.rwnd.rwnd = peers_rwnd; sctp_clog.x.rwnd.send_size = snd_size; sctp_clog.x.rwnd.overhead = overhead; sctp_clog.x.rwnd.new_rwnd = 0; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_rwnd_set(uint8_t from, uint32_t peers_rwnd, uint32_t flight_size, uint32_t overhead, uint32_t a_rwndval) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.rwnd.rwnd = peers_rwnd; sctp_clog.x.rwnd.send_size = flight_size; sctp_clog.x.rwnd.overhead = overhead; sctp_clog.x.rwnd.new_rwnd = a_rwndval; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_mbcnt(uint8_t from, uint32_t total_oq, uint32_t book, uint32_t total_mbcnt_q, uint32_t mbcnt) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.mbcnt.total_queue_size = total_oq; sctp_clog.x.mbcnt.size_change = book; sctp_clog.x.mbcnt.total_queue_mb_size = total_mbcnt_q; sctp_clog.x.mbcnt.mbcnt_change = mbcnt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MBCNT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_misc_ints(uint8_t from, uint32_t a, uint32_t b, uint32_t c, uint32_t d) { SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_MISC_EVENT, from, a, b, c, d); } void sctp_wakeup_log(struct sctp_tcb *stcb, uint32_t cumtsn, uint32_t wake_cnt, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.wake.stcb = (void *)stcb; sctp_clog.x.wake.wake_cnt = wake_cnt; sctp_clog.x.wake.flight = stcb->asoc.total_flight_count; sctp_clog.x.wake.send_q = stcb->asoc.send_queue_cnt; sctp_clog.x.wake.sent_q = stcb->asoc.sent_queue_cnt; if (stcb->asoc.stream_queue_cnt < 0xff) sctp_clog.x.wake.stream_qcnt = (uint8_t) stcb->asoc.stream_queue_cnt; else sctp_clog.x.wake.stream_qcnt = 0xff; if (stcb->asoc.chunks_on_out_queue < 0xff) sctp_clog.x.wake.chunks_on_oque = (uint8_t) stcb->asoc.chunks_on_out_queue; else sctp_clog.x.wake.chunks_on_oque = 0xff; sctp_clog.x.wake.sctpflags = 0; /* set in the defered mode stuff */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) sctp_clog.x.wake.sctpflags |= 1; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) sctp_clog.x.wake.sctpflags |= 2; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) sctp_clog.x.wake.sctpflags |= 4; /* what about the sb */ if (stcb->sctp_socket) { struct socket *so = stcb->sctp_socket; sctp_clog.x.wake.sbflags = (uint8_t) ((so->so_snd.sb_flags & 0x00ff)); } else { sctp_clog.x.wake.sbflags = 0xff; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_WAKE, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_block(uint8_t from, struct socket *so, struct sctp_association *asoc, int sendlen) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.blk.onsb = asoc->total_output_queue_size; sctp_clog.x.blk.send_sent_qcnt = (uint16_t) (asoc->send_queue_cnt + asoc->sent_queue_cnt); sctp_clog.x.blk.peer_rwnd = asoc->peers_rwnd; sctp_clog.x.blk.stream_qcnt = (uint16_t) asoc->stream_queue_cnt; sctp_clog.x.blk.chunks_on_oque = (uint16_t) asoc->chunks_on_out_queue; sctp_clog.x.blk.flight_size = (uint16_t) (asoc->total_flight / 1024); sctp_clog.x.blk.sndlen = sendlen; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_BLOCK, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } int sctp_fill_stat_log(void *optval, size_t *optsize) { /* May need to fix this if ktrdump does not work */ return (0); } #ifdef SCTP_AUDITING_ENABLED uint8_t sctp_audit_data[SCTP_AUDIT_SIZE][2]; static int sctp_audit_indx = 0; static void sctp_print_audit_report(void) { int i; int cnt; cnt = 0; for (i = sctp_audit_indx; i < SCTP_AUDIT_SIZE; i++) { if ((sctp_audit_data[i][0] == 0xe0) && (sctp_audit_data[i][1] == 0x01)) { cnt = 0; SCTP_PRINTF("\n"); } else if (sctp_audit_data[i][0] == 0xf0) { cnt = 0; SCTP_PRINTF("\n"); } else if ((sctp_audit_data[i][0] == 0xc0) && (sctp_audit_data[i][1] == 0x01)) { SCTP_PRINTF("\n"); cnt = 0; } SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], (uint32_t) sctp_audit_data[i][1]); cnt++; if ((cnt % 14) == 0) SCTP_PRINTF("\n"); } for (i = 0; i < sctp_audit_indx; i++) { if ((sctp_audit_data[i][0] == 0xe0) && (sctp_audit_data[i][1] == 0x01)) { cnt = 0; SCTP_PRINTF("\n"); } else if (sctp_audit_data[i][0] == 0xf0) { cnt = 0; SCTP_PRINTF("\n"); } else if ((sctp_audit_data[i][0] == 0xc0) && (sctp_audit_data[i][1] == 0x01)) { SCTP_PRINTF("\n"); cnt = 0; } SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], (uint32_t) sctp_audit_data[i][1]); cnt++; if ((cnt % 14) == 0) SCTP_PRINTF("\n"); } SCTP_PRINTF("\n"); } void sctp_auditing(int from, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int resend_cnt, tot_out, rep, tot_book_cnt; struct sctp_nets *lnet; struct sctp_tmit_chunk *chk; sctp_audit_data[sctp_audit_indx][0] = 0xAA; sctp_audit_data[sctp_audit_indx][1] = 0x000000ff & from; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } if (inp == NULL) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0x01; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } return; } if (stcb == NULL) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0x02; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } return; } sctp_audit_data[sctp_audit_indx][0] = 0xA1; sctp_audit_data[sctp_audit_indx][1] = (0x000000ff & stcb->asoc.sent_queue_retran_cnt); sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 0; tot_book_cnt = 0; resend_cnt = tot_out = 0; TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { resend_cnt++; } else if (chk->sent < SCTP_DATAGRAM_RESEND) { tot_out += chk->book_size; tot_book_cnt++; } } if (resend_cnt != stcb->asoc.sent_queue_retran_cnt) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA1; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } SCTP_PRINTF("resend_cnt:%d asoc-tot:%d\n", resend_cnt, stcb->asoc.sent_queue_retran_cnt); rep = 1; stcb->asoc.sent_queue_retran_cnt = resend_cnt; sctp_audit_data[sctp_audit_indx][0] = 0xA2; sctp_audit_data[sctp_audit_indx][1] = (0x000000ff & stcb->asoc.sent_queue_retran_cnt); sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } } if (tot_out != stcb->asoc.total_flight) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA2; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("tot_flt:%d asoc_tot:%d\n", tot_out, (int)stcb->asoc.total_flight); stcb->asoc.total_flight = tot_out; } if (tot_book_cnt != stcb->asoc.total_flight_count) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA5; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("tot_flt_book:%d\n", tot_book_cnt); stcb->asoc.total_flight_count = tot_book_cnt; } tot_out = 0; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { tot_out += lnet->flight_size; } if (tot_out != stcb->asoc.total_flight) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA3; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("real flight:%d net total was %d\n", stcb->asoc.total_flight, tot_out); /* now corrective action */ TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { tot_out = 0; TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if ((chk->whoTo == lnet) && (chk->sent < SCTP_DATAGRAM_RESEND)) { tot_out += chk->book_size; } } if (lnet->flight_size != tot_out) { SCTP_PRINTF("net:%p flight was %d corrected to %d\n", lnet, lnet->flight_size, tot_out); lnet->flight_size = tot_out; } } } if (rep) { sctp_print_audit_report(); } } void sctp_audit_log(uint8_t ev, uint8_t fd) { sctp_audit_data[sctp_audit_indx][0] = ev; sctp_audit_data[sctp_audit_indx][1] = fd; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } } #endif /* * sctp_stop_timers_for_shutdown() should be called * when entering the SHUTDOWN_SENT or SHUTDOWN_ACK_SENT * state to make sure that all timers are stopped. */ void sctp_stop_timers_for_shutdown(struct sctp_tcb *stcb) { struct sctp_association *asoc; struct sctp_nets *net; asoc = &stcb->asoc; (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); } } /* * a list of sizes based on typical mtu's, used only if next hop size not * returned. */ static uint32_t sctp_mtu_sizes[] = { 68, 296, 508, 512, 544, 576, 1006, 1492, 1500, 1536, 2002, 2048, 4352, 4464, 8166, 17914, 32000, 65535 }; /* * Return the largest MTU smaller than val. If there is no * entry, just return val. */ uint32_t sctp_get_prev_mtu(uint32_t val) { uint32_t i; if (val <= sctp_mtu_sizes[0]) { return (val); } for (i = 1; i < (sizeof(sctp_mtu_sizes) / sizeof(uint32_t)); i++) { if (val <= sctp_mtu_sizes[i]) { break; } } return (sctp_mtu_sizes[i - 1]); } /* * Return the smallest MTU larger than val. If there is no * entry, just return val. */ uint32_t sctp_get_next_mtu(struct sctp_inpcb *inp, uint32_t val) { /* select another MTU that is just bigger than this one */ uint32_t i; for (i = 0; i < (sizeof(sctp_mtu_sizes) / sizeof(uint32_t)); i++) { if (val < sctp_mtu_sizes[i]) { return (sctp_mtu_sizes[i]); } } return (val); } void sctp_fill_random_store(struct sctp_pcb *m) { /* * Here we use the MD5/SHA-1 to hash with our good randomNumbers and * our counter. The result becomes our good random numbers and we * then setup to give these out. Note that we do no locking to * protect this. This is ok, since if competing folks call this we * will get more gobbled gook in the random store which is what we * want. There is a danger that two guys will use the same random * numbers, but thats ok too since that is random as well :-> */ m->store_at = 0; (void)sctp_hmac(SCTP_HMAC, (uint8_t *) m->random_numbers, sizeof(m->random_numbers), (uint8_t *) & m->random_counter, sizeof(m->random_counter), (uint8_t *) m->random_store); m->random_counter++; } uint32_t sctp_select_initial_TSN(struct sctp_pcb *inp) { /* * A true implementation should use random selection process to get * the initial stream sequence number, using RFC1750 as a good * guideline */ uint32_t x, *xp; uint8_t *p; int store_at, new_store; if (inp->initial_sequence_debug != 0) { uint32_t ret; ret = inp->initial_sequence_debug; inp->initial_sequence_debug++; return (ret); } retry: store_at = inp->store_at; new_store = store_at + sizeof(uint32_t); if (new_store >= (SCTP_SIGNATURE_SIZE - 3)) { new_store = 0; } if (!atomic_cmpset_int(&inp->store_at, store_at, new_store)) { goto retry; } if (new_store == 0) { /* Refill the random store */ sctp_fill_random_store(inp); } p = &inp->random_store[store_at]; xp = (uint32_t *) p; x = *xp; return (x); } uint32_t sctp_select_a_tag(struct sctp_inpcb *inp, uint16_t lport, uint16_t rport, int save_in_twait) { uint32_t x, not_done; struct timeval now; (void)SCTP_GETTIME_TIMEVAL(&now); not_done = 1; while (not_done) { x = sctp_select_initial_TSN(&inp->sctp_ep); if (x == 0) { /* we never use 0 */ continue; } if (sctp_is_vtag_good(inp, x, lport, rport, &now, save_in_twait)) { not_done = 0; } } return (x); } int sctp_init_asoc(struct sctp_inpcb *m, struct sctp_tcb *stcb, uint32_t override_tag, uint32_t vrf_id) { struct sctp_association *asoc; /* * Anything set to zero is taken care of by the allocation routine's * bzero */ /* * Up front select what scoping to apply on addresses I tell my peer * Not sure what to do with these right now, we will need to come up * with a way to set them. We may need to pass them through from the * caller in the sctp_aloc_assoc() function. */ int i; asoc = &stcb->asoc; /* init all variables to a known value. */ SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_INUSE); asoc->max_burst = m->sctp_ep.max_burst; asoc->heart_beat_delay = TICKS_TO_MSEC(m->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); asoc->cookie_life = m->sctp_ep.def_cookie_life; asoc->sctp_cmt_on_off = m->sctp_cmt_on_off; asoc->sctp_nr_sack_on_off = (uint8_t) SCTP_BASE_SYSCTL(sctp_nr_sack_on_off); asoc->sctp_cmt_pf = (uint8_t) SCTP_BASE_SYSCTL(sctp_cmt_pf); asoc->sctp_frag_point = m->sctp_frag_point; #ifdef INET asoc->default_tos = m->ip_inp.inp.inp_ip_tos; #else asoc->default_tos = 0; #endif #ifdef INET6 asoc->default_flowlabel = ((struct in6pcb *)m)->in6p_flowinfo; #else asoc->default_flowlabel = 0; #endif asoc->sb_send_resv = 0; if (override_tag) { asoc->my_vtag = override_tag; } else { asoc->my_vtag = sctp_select_a_tag(m, stcb->sctp_ep->sctp_lport, stcb->rport, 1); } /* Get the nonce tags */ asoc->my_vtag_nonce = sctp_select_a_tag(m, stcb->sctp_ep->sctp_lport, stcb->rport, 0); asoc->peer_vtag_nonce = sctp_select_a_tag(m, stcb->sctp_ep->sctp_lport, stcb->rport, 0); asoc->vrf_id = vrf_id; if (sctp_is_feature_on(m, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) asoc->hb_is_disabled = 1; else asoc->hb_is_disabled = 0; #ifdef SCTP_ASOCLOG_OF_TSNS asoc->tsn_in_at = 0; asoc->tsn_out_at = 0; asoc->tsn_in_wrapped = 0; asoc->tsn_out_wrapped = 0; asoc->cumack_log_at = 0; asoc->cumack_log_atsnt = 0; #endif #ifdef SCTP_FS_SPEC_LOG asoc->fs_index = 0; #endif asoc->refcnt = 0; asoc->assoc_up_sent = 0; asoc->asconf_seq_out = asoc->str_reset_seq_out = asoc->init_seq_number = asoc->sending_seq = sctp_select_initial_TSN(&m->sctp_ep); asoc->asconf_seq_out_acked = asoc->asconf_seq_out - 1; /* we are optimisitic here */ asoc->peer_supports_pktdrop = 1; asoc->peer_supports_nat = 0; asoc->sent_queue_retran_cnt = 0; /* for CMT */ asoc->last_net_cmt_send_started = NULL; /* This will need to be adjusted */ asoc->last_acked_seq = asoc->init_seq_number - 1; asoc->advanced_peer_ack_point = asoc->last_acked_seq; asoc->asconf_seq_in = asoc->last_acked_seq; /* here we are different, we hold the next one we expect */ asoc->str_reset_seq_in = asoc->last_acked_seq + 1; asoc->initial_init_rto_max = m->sctp_ep.initial_init_rto_max; asoc->initial_rto = m->sctp_ep.initial_rto; asoc->max_init_times = m->sctp_ep.max_init_times; asoc->max_send_times = m->sctp_ep.max_send_times; asoc->def_net_failure = m->sctp_ep.def_net_failure; asoc->free_chunk_cnt = 0; asoc->iam_blocking = 0; - /* ECN Nonce initialization */ + asoc->context = m->sctp_context; asoc->def_send = m->def_send; - asoc->ecn_nonce_allowed = 0; - asoc->receiver_nonce_sum = 1; - asoc->nonce_sum_expect_base = 1; - asoc->nonce_sum_check = 1; - asoc->nonce_resync_tsn = 0; - asoc->nonce_wait_for_ecne = 0; - asoc->nonce_wait_tsn = 0; asoc->delayed_ack = TICKS_TO_MSEC(m->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); asoc->sack_freq = m->sctp_ep.sctp_sack_freq; asoc->pr_sctp_cnt = 0; asoc->total_output_queue_size = 0; if (m->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { struct in6pcb *inp6; /* Its a V6 socket */ inp6 = (struct in6pcb *)m; asoc->ipv6_addr_legal = 1; /* Now look at the binding flag to see if V4 will be legal */ if (SCTP_IPV6_V6ONLY(inp6) == 0) { asoc->ipv4_addr_legal = 1; } else { /* V4 addresses are NOT legal on the association */ asoc->ipv4_addr_legal = 0; } } else { /* Its a V4 socket, no - V6 */ asoc->ipv4_addr_legal = 1; asoc->ipv6_addr_legal = 0; } asoc->my_rwnd = max(SCTP_SB_LIMIT_RCV(m->sctp_socket), SCTP_MINIMAL_RWND); asoc->peers_rwnd = SCTP_SB_LIMIT_RCV(m->sctp_socket); asoc->smallest_mtu = m->sctp_frag_point; asoc->minrto = m->sctp_ep.sctp_minrto; asoc->maxrto = m->sctp_ep.sctp_maxrto; asoc->locked_on_sending = NULL; asoc->stream_locked_on = 0; asoc->ecn_echo_cnt_onq = 0; asoc->stream_locked = 0; asoc->send_sack = 1; LIST_INIT(&asoc->sctp_restricted_addrs); TAILQ_INIT(&asoc->nets); TAILQ_INIT(&asoc->pending_reply_queue); TAILQ_INIT(&asoc->asconf_ack_sent); /* Setup to fill the hb random cache at first HB */ asoc->hb_random_idx = 4; asoc->sctp_autoclose_ticks = m->sctp_ep.auto_close_time; stcb->asoc.congestion_control_module = m->sctp_ep.sctp_default_cc_module; stcb->asoc.cc_functions = sctp_cc_functions[m->sctp_ep.sctp_default_cc_module]; stcb->asoc.stream_scheduling_module = m->sctp_ep.sctp_default_ss_module; stcb->asoc.ss_functions = sctp_ss_functions[m->sctp_ep.sctp_default_ss_module]; /* * Now the stream parameters, here we allocate space for all streams * that we request by default. */ asoc->strm_realoutsize = asoc->streamoutcnt = asoc->pre_open_streams = m->sctp_ep.pre_open_stream_count; SCTP_MALLOC(asoc->strmout, struct sctp_stream_out *, asoc->streamoutcnt * sizeof(struct sctp_stream_out), SCTP_M_STRMO); if (asoc->strmout == NULL) { /* big trouble no memory */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } for (i = 0; i < asoc->streamoutcnt; i++) { /* * inbound side must be set to 0xffff, also NOTE when we get * the INIT-ACK back (for INIT sender) we MUST reduce the * count (streamoutcnt) but first check if we sent to any of * the upper streams that were dropped (if some were). Those * that were dropped must be notified to the upper layer as * failed to send. */ asoc->strmout[i].next_sequence_sent = 0x0; TAILQ_INIT(&asoc->strmout[i].outqueue); asoc->strmout[i].stream_no = i; asoc->strmout[i].last_msg_incomplete = 0; asoc->ss_functions.sctp_ss_init_stream(&asoc->strmout[i]); } asoc->ss_functions.sctp_ss_init(stcb, asoc, 0); /* Now the mapping array */ asoc->mapping_array_size = SCTP_INITIAL_MAPPING_ARRAY; SCTP_MALLOC(asoc->mapping_array, uint8_t *, asoc->mapping_array_size, SCTP_M_MAP); if (asoc->mapping_array == NULL) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } memset(asoc->mapping_array, 0, asoc->mapping_array_size); SCTP_MALLOC(asoc->nr_mapping_array, uint8_t *, asoc->mapping_array_size, SCTP_M_MAP); if (asoc->nr_mapping_array == NULL) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } memset(asoc->nr_mapping_array, 0, asoc->mapping_array_size); /* Now the init of the other outqueues */ TAILQ_INIT(&asoc->free_chunks); TAILQ_INIT(&asoc->control_send_queue); TAILQ_INIT(&asoc->asconf_send_queue); TAILQ_INIT(&asoc->send_queue); TAILQ_INIT(&asoc->sent_queue); TAILQ_INIT(&asoc->reasmqueue); TAILQ_INIT(&asoc->resetHead); asoc->max_inbound_streams = m->sctp_ep.max_open_streams_intome; TAILQ_INIT(&asoc->asconf_queue); /* authentication fields */ asoc->authinfo.random = NULL; asoc->authinfo.active_keyid = 0; asoc->authinfo.assoc_key = NULL; asoc->authinfo.assoc_keyid = 0; asoc->authinfo.recv_key = NULL; asoc->authinfo.recv_keyid = 0; LIST_INIT(&asoc->shared_keys); asoc->marked_retrans = 0; asoc->timoinit = 0; asoc->timodata = 0; asoc->timosack = 0; asoc->timoshutdown = 0; asoc->timoheartbeat = 0; asoc->timocookie = 0; asoc->timoshutdownack = 0; (void)SCTP_GETTIME_TIMEVAL(&asoc->start_time); asoc->discontinuity_time = asoc->start_time; /* * sa_ignore MEMLEAK {memory is put in the assoc mapping array and * freed later when the association is freed. */ return (0); } void sctp_print_mapping_array(struct sctp_association *asoc) { unsigned int i, limit; printf("Mapping array size: %d, baseTSN: %8.8x, cumAck: %8.8x, highestTSN: (%8.8x, %8.8x).\n", asoc->mapping_array_size, asoc->mapping_array_base_tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, asoc->highest_tsn_inside_nr_map); for (limit = asoc->mapping_array_size; limit > 1; limit--) { if (asoc->mapping_array[limit - 1]) { break; } } printf("Renegable mapping array (last %d entries are zero):\n", asoc->mapping_array_size - limit); for (i = 0; i < limit; i++) { printf("%2.2x%c", asoc->mapping_array[i], ((i + 1) % 16) ? ' ' : '\n'); } if (limit % 16) printf("\n"); for (limit = asoc->mapping_array_size; limit > 1; limit--) { if (asoc->nr_mapping_array[limit - 1]) { break; } } printf("Non renegable mapping array (last %d entries are zero):\n", asoc->mapping_array_size - limit); for (i = 0; i < limit; i++) { printf("%2.2x%c", asoc->nr_mapping_array[i], ((i + 1) % 16) ? ' ' : '\n'); } if (limit % 16) printf("\n"); } int sctp_expand_mapping_array(struct sctp_association *asoc, uint32_t needed) { /* mapping array needs to grow */ uint8_t *new_array1, *new_array2; uint32_t new_size; new_size = asoc->mapping_array_size + ((needed + 7) / 8 + SCTP_MAPPING_ARRAY_INCR); SCTP_MALLOC(new_array1, uint8_t *, new_size, SCTP_M_MAP); SCTP_MALLOC(new_array2, uint8_t *, new_size, SCTP_M_MAP); if ((new_array1 == NULL) || (new_array2 == NULL)) { /* can't get more, forget it */ SCTP_PRINTF("No memory for expansion of SCTP mapping array %d\n", new_size); if (new_array1) { SCTP_FREE(new_array1, SCTP_M_MAP); } if (new_array2) { SCTP_FREE(new_array2, SCTP_M_MAP); } return (-1); } memset(new_array1, 0, new_size); memset(new_array2, 0, new_size); memcpy(new_array1, asoc->mapping_array, asoc->mapping_array_size); memcpy(new_array2, asoc->nr_mapping_array, asoc->mapping_array_size); SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); asoc->mapping_array = new_array1; asoc->nr_mapping_array = new_array2; asoc->mapping_array_size = new_size; return (0); } static void sctp_iterator_work(struct sctp_iterator *it) { int iteration_count = 0; int inp_skip = 0; int first_in = 1; struct sctp_inpcb *tinp; SCTP_INP_INFO_RLOCK(); SCTP_ITERATOR_LOCK(); if (it->inp) { SCTP_INP_RLOCK(it->inp); SCTP_INP_DECR_REF(it->inp); } if (it->inp == NULL) { /* iterator is complete */ done_with_iterator: SCTP_ITERATOR_UNLOCK(); SCTP_INP_INFO_RUNLOCK(); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); return; } select_a_new_ep: if (first_in) { first_in = 0; } else { SCTP_INP_RLOCK(it->inp); } while (((it->pcb_flags) && ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) || ((it->pcb_features) && ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) { /* endpoint flags or features don't match, so keep looking */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { SCTP_INP_RUNLOCK(it->inp); goto done_with_iterator; } tinp = it->inp; it->inp = LIST_NEXT(it->inp, sctp_list); SCTP_INP_RUNLOCK(tinp); if (it->inp == NULL) { goto done_with_iterator; } SCTP_INP_RLOCK(it->inp); } /* now go through each assoc which is in the desired state */ if (it->done_current_ep == 0) { if (it->function_inp != NULL) inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val); it->done_current_ep = 1; } if (it->stcb == NULL) { /* run the per instance function */ it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list); } if ((inp_skip) || it->stcb == NULL) { if (it->function_inp_end != NULL) { inp_skip = (*it->function_inp_end) (it->inp, it->pointer, it->val); } SCTP_INP_RUNLOCK(it->inp); goto no_stcb; } while (it->stcb) { SCTP_TCB_LOCK(it->stcb); if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) { /* not in the right state... keep looking */ SCTP_TCB_UNLOCK(it->stcb); goto next_assoc; } /* see if we have limited out the iterator loop */ iteration_count++; if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) { /* Pause to let others grab the lock */ atomic_add_int(&it->stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(it->stcb); SCTP_INP_INCR_REF(it->inp); SCTP_INP_RUNLOCK(it->inp); SCTP_ITERATOR_UNLOCK(); SCTP_INP_INFO_RUNLOCK(); SCTP_INP_INFO_RLOCK(); SCTP_ITERATOR_LOCK(); if (sctp_it_ctl.iterator_flags) { /* We won't be staying here */ SCTP_INP_DECR_REF(it->inp); atomic_add_int(&it->stcb->asoc.refcnt, -1); if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_MUST_EXIT) { goto done_with_iterator; } if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_STOP_CUR_IT) { sctp_it_ctl.iterator_flags &= ~SCTP_ITERATOR_STOP_CUR_IT; goto done_with_iterator; } if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_STOP_CUR_INP) { sctp_it_ctl.iterator_flags &= ~SCTP_ITERATOR_STOP_CUR_INP; goto no_stcb; } /* If we reach here huh? */ printf("Unknown it ctl flag %x\n", sctp_it_ctl.iterator_flags); sctp_it_ctl.iterator_flags = 0; } SCTP_INP_RLOCK(it->inp); SCTP_INP_DECR_REF(it->inp); SCTP_TCB_LOCK(it->stcb); atomic_add_int(&it->stcb->asoc.refcnt, -1); iteration_count = 0; } /* run function on this one */ (*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val); /* * we lie here, it really needs to have its own type but * first I must verify that this won't effect things :-0 */ if (it->no_chunk_output == 0) sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); SCTP_TCB_UNLOCK(it->stcb); next_assoc: it->stcb = LIST_NEXT(it->stcb, sctp_tcblist); if (it->stcb == NULL) { /* Run last function */ if (it->function_inp_end != NULL) { inp_skip = (*it->function_inp_end) (it->inp, it->pointer, it->val); } } } SCTP_INP_RUNLOCK(it->inp); no_stcb: /* done with all assocs on this endpoint, move on to next endpoint */ it->done_current_ep = 0; if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; } else { it->inp = LIST_NEXT(it->inp, sctp_list); } if (it->inp == NULL) { goto done_with_iterator; } goto select_a_new_ep; } void sctp_iterator_worker(void) { struct sctp_iterator *it, *nit; /* This function is called with the WQ lock in place */ sctp_it_ctl.iterator_running = 1; TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { sctp_it_ctl.cur_it = it; /* now lets work on this one */ TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); SCTP_IPI_ITERATOR_WQ_UNLOCK(); CURVNET_SET(it->vn); sctp_iterator_work(it); CURVNET_RESTORE(); SCTP_IPI_ITERATOR_WQ_LOCK(); if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_MUST_EXIT) { sctp_it_ctl.cur_it = NULL; break; } /* sa_ignore FREED_MEMORY */ } sctp_it_ctl.iterator_running = 0; return; } static void sctp_handle_addr_wq(void) { /* deal with the ADDR wq from the rtsock calls */ struct sctp_laddr *wi, *nwi; struct sctp_asconf_iterator *asc; SCTP_MALLOC(asc, struct sctp_asconf_iterator *, sizeof(struct sctp_asconf_iterator), SCTP_M_ASC_IT); if (asc == NULL) { /* Try later, no memory */ sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); return; } LIST_INIT(&asc->list_of_work); asc->cnt = 0; SCTP_WQ_ADDR_LOCK(); LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { LIST_REMOVE(wi, sctp_nxt_addr); LIST_INSERT_HEAD(&asc->list_of_work, wi, sctp_nxt_addr); asc->cnt++; } SCTP_WQ_ADDR_UNLOCK(); if (asc->cnt == 0) { SCTP_FREE(asc, SCTP_M_ASC_IT); } else { (void)sctp_initiate_iterator(sctp_asconf_iterator_ep, sctp_asconf_iterator_stcb, NULL, /* No ep end for boundall */ SCTP_PCB_FLAGS_BOUNDALL, SCTP_PCB_ANY_FEATURES, SCTP_ASOC_ANY_STATE, (void *)asc, 0, sctp_asconf_iterator_end, NULL, 0); } } int retcode = 0; int cur_oerr = 0; void sctp_timeout_handler(void *t) { struct sctp_inpcb *inp; struct sctp_tcb *stcb; struct sctp_nets *net; struct sctp_timer *tmr; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif int did_output, type; tmr = (struct sctp_timer *)t; inp = (struct sctp_inpcb *)tmr->ep; stcb = (struct sctp_tcb *)tmr->tcb; net = (struct sctp_nets *)tmr->net; CURVNET_SET((struct vnet *)tmr->vnet); did_output = 1; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xF0, (uint8_t) tmr->type); sctp_auditing(3, inp, stcb, net); #endif /* sanity checks... */ if (tmr->self != (void *)tmr) { /* * SCTP_PRINTF("Stale SCTP timer fired (%p), ignoring...\n", * tmr); */ CURVNET_RESTORE(); return; } tmr->stopped_from = 0xa001; if (!SCTP_IS_TIMER_TYPE_VALID(tmr->type)) { /* * SCTP_PRINTF("SCTP timer fired with invalid type: 0x%x\n", * tmr->type); */ CURVNET_RESTORE(); return; } tmr->stopped_from = 0xa002; if ((tmr->type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) { CURVNET_RESTORE(); return; } /* if this is an iterator timeout, get the struct and clear inp */ tmr->stopped_from = 0xa003; type = tmr->type; if (inp) { SCTP_INP_INCR_REF(inp); if ((inp->sctp_socket == 0) && ((tmr->type != SCTP_TIMER_TYPE_INPKILL) && (tmr->type != SCTP_TIMER_TYPE_INIT) && (tmr->type != SCTP_TIMER_TYPE_SEND) && (tmr->type != SCTP_TIMER_TYPE_RECV) && (tmr->type != SCTP_TIMER_TYPE_HEARTBEAT) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWN) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWNACK) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWNGUARD) && (tmr->type != SCTP_TIMER_TYPE_ASOCKILL)) ) { SCTP_INP_DECR_REF(inp); CURVNET_RESTORE(); return; } } tmr->stopped_from = 0xa004; if (stcb) { atomic_add_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state == 0) { atomic_add_int(&stcb->asoc.refcnt, -1); if (inp) { SCTP_INP_DECR_REF(inp); } CURVNET_RESTORE(); return; } } tmr->stopped_from = 0xa005; SCTPDBG(SCTP_DEBUG_TIMER1, "Timer type %d goes off\n", tmr->type); if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { if (inp) { SCTP_INP_DECR_REF(inp); } if (stcb) { atomic_add_int(&stcb->asoc.refcnt, -1); } CURVNET_RESTORE(); return; } tmr->stopped_from = 0xa006; if (stcb) { SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((tmr->type != SCTP_TIMER_TYPE_ASOCKILL) && ((stcb->asoc.state == 0) || (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED))) { SCTP_TCB_UNLOCK(stcb); if (inp) { SCTP_INP_DECR_REF(inp); } CURVNET_RESTORE(); return; } } /* record in stopped what t-o occured */ tmr->stopped_from = tmr->type; /* mark as being serviced now */ if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { /* * Callout has been rescheduled. */ goto get_out; } if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { /* * Not active, so no action. */ goto get_out; } SCTP_OS_TIMER_DEACTIVATE(&tmr->timer); /* call the handler for the appropriate timer type */ switch (tmr->type) { case SCTP_TIMER_TYPE_ZERO_COPY: if (inp == NULL) { break; } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: if (inp == NULL) { break; } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_SENDQ_EVENT(inp, inp->sctp_socket); } break; case SCTP_TIMER_TYPE_ADDR_WQ: sctp_handle_addr_wq(); break; case SCTP_TIMER_TYPE_SEND: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timodata); stcb->asoc.timodata++; stcb->asoc.num_send_timers_up--; if (stcb->asoc.num_send_timers_up < 0) { stcb->asoc.num_send_timers_up = 0; } SCTP_TCB_LOCK_ASSERT(stcb); cur_oerr = stcb->asoc.overall_error_count; retcode = sctp_t3rxt_timer(inp, stcb, net); if (retcode) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_TCB_LOCK_ASSERT(stcb); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); if ((stcb->asoc.num_send_timers_up == 0) && (stcb->asoc.sent_queue_cnt > 0)) { struct sctp_tmit_chunk *chk; /* * safeguard. If there on some on the sent queue * somewhere but no timers running something is * wrong... so we start a timer on the first chunk * on the send queue on whatever net it is sent to. */ chk = TAILQ_FIRST(&stcb->asoc.sent_queue); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, chk->whoTo); } break; case SCTP_TIMER_TYPE_INIT: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoinit); stcb->asoc.timoinit++; if (sctp_t1init_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } /* We do output but not here */ did_output = 0; break; case SCTP_TIMER_TYPE_RECV: if ((stcb == NULL) || (inp == NULL)) { break; } { SCTP_STAT_INCR(sctps_timosack); stcb->asoc.timosack++; sctp_send_sack(stcb); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SACK_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_SHUTDOWN: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_shutdown_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoshutdown); stcb->asoc.timoshutdown++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_HEARTBEAT: { struct sctp_nets *lnet; int cnt_of_unconf = 0; if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoheartbeat); stcb->asoc.timoheartbeat++; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { cnt_of_unconf++; } } if (cnt_of_unconf == 0) { if (sctp_heartbeat_timer(inp, stcb, lnet, cnt_of_unconf)) { /* no need to unlock on tcb its gone */ goto out_decr; } } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, lnet); #endif sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, lnet); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_HB_TMR, SCTP_SO_NOT_LOCKED); } break; case SCTP_TIMER_TYPE_COOKIE: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_cookie_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timocookie); stcb->asoc.timocookie++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif /* * We consider T3 and Cookie timer pretty much the same with * respect to where from in chunk_output. */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_NEWCOOKIE: { struct timeval tv; int i, secret; if (inp == NULL) { break; } SCTP_STAT_INCR(sctps_timosecret); (void)SCTP_GETTIME_TIMEVAL(&tv); SCTP_INP_WLOCK(inp); inp->sctp_ep.time_of_secret_change = tv.tv_sec; inp->sctp_ep.last_secret_number = inp->sctp_ep.current_secret_number; inp->sctp_ep.current_secret_number++; if (inp->sctp_ep.current_secret_number >= SCTP_HOW_MANY_SECRETS) { inp->sctp_ep.current_secret_number = 0; } secret = (int)inp->sctp_ep.current_secret_number; for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { inp->sctp_ep.secret_key[secret][i] = sctp_select_initial_TSN(&inp->sctp_ep); } SCTP_INP_WUNLOCK(inp); sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, stcb, net); } did_output = 0; break; case SCTP_TIMER_TYPE_PATHMTURAISE: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timopathmtu); sctp_pathmtu_timer(inp, stcb, net); did_output = 0; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_shutdownack_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoshutdownack); stcb->asoc.timoshutdownack++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_ACK_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoshutdownguard); sctp_abort_an_association(inp, stcb, SCTP_SHUTDOWN_GUARD_EXPIRES, NULL, SCTP_SO_NOT_LOCKED); /* no need to unlock on tcb its gone */ goto out_decr; case SCTP_TIMER_TYPE_STRRESET: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_strreset_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timostrmrst); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_EARLYFR: /* Need to do FR of things for net */ if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoearlyfr); sctp_early_fr_timer(inp, stcb, net); break; case SCTP_TIMER_TYPE_ASCONF: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_asconf_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoasconf); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_ASCONF_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_PRIM_DELETED: if ((stcb == NULL) || (inp == NULL)) { break; } sctp_delete_prim_timer(inp, stcb, net); SCTP_STAT_INCR(sctps_timodelprim); break; case SCTP_TIMER_TYPE_AUTOCLOSE: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoautoclose); sctp_autoclose_timer(inp, stcb, net); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED); did_output = 0; break; case SCTP_TIMER_TYPE_ASOCKILL: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoassockill); /* Can we free it yet? */ SCTP_INP_DECR_REF(inp); sctp_timer_stop(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_2); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif /* * free asoc, always unlocks (or destroy's) so prevent * duplicate unlock or unlock of a free mtx :-0 */ stcb = NULL; goto out_no_decr; case SCTP_TIMER_TYPE_INPKILL: SCTP_STAT_INCR(sctps_timoinpkill); if (inp == NULL) { break; } /* * special case, take away our increment since WE are the * killer */ SCTP_INP_DECR_REF(inp); sctp_timer_stop(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_3); sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_FROM_INPKILL_TIMER); inp = NULL; goto out_no_decr; default: SCTPDBG(SCTP_DEBUG_TIMER1, "sctp_timeout_handler:unknown timer %d\n", tmr->type); break; }; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xF1, (uint8_t) tmr->type); if (inp) sctp_auditing(5, inp, stcb, net); #endif if ((did_output) && stcb) { /* * Now we need to clean up the control chunk chain if an * ECNE is on it. It must be marked as UNSENT again so next * call will continue to send it until such time that we get * a CWR, to remove it. It is, however, less likely that we * will find a ecn echo on the chain though. */ sctp_fix_ecn_echo(&stcb->asoc); } get_out: if (stcb) { SCTP_TCB_UNLOCK(stcb); } out_decr: if (inp) { SCTP_INP_DECR_REF(inp); } out_no_decr: SCTPDBG(SCTP_DEBUG_TIMER1, "Timer now complete (type %d)\n", type); CURVNET_RESTORE(); } void sctp_timer_start(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int to_ticks; struct sctp_timer *tmr; if ((t_type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) return; to_ticks = 0; tmr = NULL; if (stcb) { SCTP_TCB_LOCK_ASSERT(stcb); } switch (t_type) { case SCTP_TIMER_TYPE_ZERO_COPY: tmr = &inp->sctp_ep.zero_copy_timer; to_ticks = SCTP_ZERO_COPY_TICK_DELAY; break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: tmr = &inp->sctp_ep.zero_copy_sendq_timer; to_ticks = SCTP_ZERO_COPY_SENDQ_TICK_DELAY; break; case SCTP_TIMER_TYPE_ADDR_WQ: /* Only 1 tick away :-) */ tmr = &SCTP_BASE_INFO(addr_wq_timer); to_ticks = SCTP_ADDRESS_TICK_DELAY; break; case SCTP_TIMER_TYPE_SEND: /* Here we use the RTO timer */ { int rto_val; if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; if (net->RTO == 0) { rto_val = stcb->asoc.initial_rto; } else { rto_val = net->RTO; } to_ticks = MSEC_TO_TICKS(rto_val); } break; case SCTP_TIMER_TYPE_INIT: /* * Here we use the INIT timer default usually about 1 * minute. */ if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } break; case SCTP_TIMER_TYPE_RECV: /* * Here we use the Delayed-Ack timer value from the inp * ususually about 200ms. */ if (stcb == NULL) { return; } tmr = &stcb->asoc.dack_timer; to_ticks = MSEC_TO_TICKS(stcb->asoc.delayed_ack); break; case SCTP_TIMER_TYPE_SHUTDOWN: /* Here we use the RTO of the destination. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_HEARTBEAT: /* * the net is used here so that we can add in the RTO. Even * though we use a different timer. We also add the HB timer * PLUS a random jitter. */ if ((inp == NULL) || (stcb == NULL)) { return; } else { uint32_t rndval; uint8_t this_random; int cnt_of_unconf = 0; struct sctp_nets *lnet; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { cnt_of_unconf++; } } if (cnt_of_unconf) { net = lnet = NULL; (void)sctp_heartbeat_timer(inp, stcb, lnet, cnt_of_unconf); } if (stcb->asoc.hb_random_idx > 3) { rndval = sctp_select_initial_TSN(&inp->sctp_ep); memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); stcb->asoc.hb_random_idx = 0; } this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; stcb->asoc.hb_random_idx++; stcb->asoc.hb_ect_randombit = 0; /* * this_random will be 0 - 256 ms RTO is in ms. */ if ((stcb->asoc.hb_is_disabled) && (cnt_of_unconf == 0)) { return; } if (net) { int delay; delay = stcb->asoc.heart_beat_delay; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && ((lnet->dest_state & SCTP_ADDR_OUT_OF_SCOPE) == 0) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { delay = 0; } } if (net->RTO == 0) { /* Never been checked */ to_ticks = this_random + stcb->asoc.initial_rto + delay; } else { /* set rto_val to the ms */ to_ticks = delay + net->RTO + this_random; } } else { if (cnt_of_unconf) { to_ticks = this_random + stcb->asoc.initial_rto; } else { to_ticks = stcb->asoc.heart_beat_delay + this_random + stcb->asoc.initial_rto; } } /* * Now we must convert the to_ticks that are now in * ms to ticks. */ to_ticks = MSEC_TO_TICKS(to_ticks); tmr = &stcb->asoc.hb_timer; } break; case SCTP_TIMER_TYPE_COOKIE: /* * Here we can use the RTO timer from the network since one * RTT was compelete. If a retran happened then we will be * using the RTO initial value. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_NEWCOOKIE: /* * nothing needed but the endpoint here ususually about 60 * minutes. */ if (inp == NULL) { return; } tmr = &inp->sctp_ep.signature_change; to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_SIGNATURE]; break; case SCTP_TIMER_TYPE_ASOCKILL: if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; to_ticks = MSEC_TO_TICKS(SCTP_ASOC_KILL_TIMEOUT); break; case SCTP_TIMER_TYPE_INPKILL: /* * The inp is setup to die. We re-use the signature_chage * timer since that has stopped and we are in the GONE * state. */ if (inp == NULL) { return; } tmr = &inp->sctp_ep.signature_change; to_ticks = MSEC_TO_TICKS(SCTP_INP_KILL_TIMEOUT); break; case SCTP_TIMER_TYPE_PATHMTURAISE: /* * Here we use the value found in the EP for PMTU ususually * about 10 minutes. */ if ((stcb == NULL) || (inp == NULL)) { return; } if (net == NULL) { return; } to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_PMTU]; tmr = &net->pmtu_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: /* Here we use the RTO of the destination */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: /* * Here we use the endpoints shutdown guard timer usually * about 3 minutes. */ if ((inp == NULL) || (stcb == NULL)) { return; } to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN]; tmr = &stcb->asoc.shut_guard_timer; break; case SCTP_TIMER_TYPE_STRRESET: /* * Here the timer comes from the stcb but its value is from * the net's RTO. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_EARLYFR: { unsigned int msec; if ((stcb == NULL) || (net == NULL)) { return; } if (net->flight_size > net->cwnd) { /* no need to start */ return; } SCTP_STAT_INCR(sctps_earlyfrstart); if (net->lastsa == 0) { /* Hmm no rtt estimate yet? */ msec = stcb->asoc.initial_rto >> 2; } else { msec = ((net->lastsa >> 2) + net->lastsv) >> 1; } if (msec < SCTP_BASE_SYSCTL(sctp_early_fr_msec)) { msec = SCTP_BASE_SYSCTL(sctp_early_fr_msec); if (msec < SCTP_MINFR_MSEC_FLOOR) { msec = SCTP_MINFR_MSEC_FLOOR; } } to_ticks = MSEC_TO_TICKS(msec); tmr = &net->fr_timer; } break; case SCTP_TIMER_TYPE_ASCONF: /* * Here the timer comes from the stcb but its value is from * the net's RTO. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &stcb->asoc.asconf_timer; break; case SCTP_TIMER_TYPE_PRIM_DELETED: if ((stcb == NULL) || (net != NULL)) { return; } to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); tmr = &stcb->asoc.delete_prim_timer; break; case SCTP_TIMER_TYPE_AUTOCLOSE: if (stcb == NULL) { return; } if (stcb->asoc.sctp_autoclose_ticks == 0) { /* * Really an error since stcb is NOT set to * autoclose */ return; } to_ticks = stcb->asoc.sctp_autoclose_ticks; tmr = &stcb->asoc.autoclose_timer; break; default: SCTPDBG(SCTP_DEBUG_TIMER1, "%s: Unknown timer type %d\n", __FUNCTION__, t_type); return; break; }; if ((to_ticks <= 0) || (tmr == NULL)) { SCTPDBG(SCTP_DEBUG_TIMER1, "%s: %d:software error to_ticks:%d tmr:%p not set ??\n", __FUNCTION__, t_type, to_ticks, tmr); return; } if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { /* * we do NOT allow you to have it already running. if it is * we leave the current one up unchanged */ return; } /* At this point we can proceed */ if (t_type == SCTP_TIMER_TYPE_SEND) { stcb->asoc.num_send_timers_up++; } tmr->stopped_from = 0; tmr->type = t_type; tmr->ep = (void *)inp; tmr->tcb = (void *)stcb; tmr->net = (void *)net; tmr->self = (void *)tmr; tmr->vnet = (void *)curvnet; tmr->ticks = sctp_get_tick_count(); (void)SCTP_OS_TIMER_START(&tmr->timer, to_ticks, sctp_timeout_handler, tmr); return; } void sctp_timer_stop(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t from) { struct sctp_timer *tmr; if ((t_type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) return; tmr = NULL; if (stcb) { SCTP_TCB_LOCK_ASSERT(stcb); } switch (t_type) { case SCTP_TIMER_TYPE_ZERO_COPY: tmr = &inp->sctp_ep.zero_copy_timer; break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: tmr = &inp->sctp_ep.zero_copy_sendq_timer; break; case SCTP_TIMER_TYPE_ADDR_WQ: tmr = &SCTP_BASE_INFO(addr_wq_timer); break; case SCTP_TIMER_TYPE_EARLYFR: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->fr_timer; SCTP_STAT_INCR(sctps_earlyfrstop); break; case SCTP_TIMER_TYPE_SEND: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_INIT: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_RECV: if (stcb == NULL) { return; } tmr = &stcb->asoc.dack_timer; break; case SCTP_TIMER_TYPE_SHUTDOWN: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_HEARTBEAT: if (stcb == NULL) { return; } tmr = &stcb->asoc.hb_timer; break; case SCTP_TIMER_TYPE_COOKIE: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_NEWCOOKIE: /* nothing needed but the endpoint here */ tmr = &inp->sctp_ep.signature_change; /* * We re-use the newcookie timer for the INP kill timer. We * must assure that we do not kill it by accident. */ break; case SCTP_TIMER_TYPE_ASOCKILL: /* * Stop the asoc kill timer. */ if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_INPKILL: /* * The inp is setup to die. We re-use the signature_chage * timer since that has stopped and we are in the GONE * state. */ tmr = &inp->sctp_ep.signature_change; break; case SCTP_TIMER_TYPE_PATHMTURAISE: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->pmtu_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: if (stcb == NULL) { return; } tmr = &stcb->asoc.shut_guard_timer; break; case SCTP_TIMER_TYPE_STRRESET: if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_ASCONF: if (stcb == NULL) { return; } tmr = &stcb->asoc.asconf_timer; break; case SCTP_TIMER_TYPE_PRIM_DELETED: if (stcb == NULL) { return; } tmr = &stcb->asoc.delete_prim_timer; break; case SCTP_TIMER_TYPE_AUTOCLOSE: if (stcb == NULL) { return; } tmr = &stcb->asoc.autoclose_timer; break; default: SCTPDBG(SCTP_DEBUG_TIMER1, "%s: Unknown timer type %d\n", __FUNCTION__, t_type); break; }; if (tmr == NULL) { return; } if ((tmr->type != t_type) && tmr->type) { /* * Ok we have a timer that is under joint use. Cookie timer * per chance with the SEND timer. We therefore are NOT * running the timer that the caller wants stopped. So just * return. */ return; } if ((t_type == SCTP_TIMER_TYPE_SEND) && (stcb != NULL)) { stcb->asoc.num_send_timers_up--; if (stcb->asoc.num_send_timers_up < 0) { stcb->asoc.num_send_timers_up = 0; } } tmr->self = NULL; tmr->stopped_from = from; (void)SCTP_OS_TIMER_STOP(&tmr->timer); return; } uint32_t sctp_calculate_len(struct mbuf *m) { uint32_t tlen = 0; struct mbuf *at; at = m; while (at) { tlen += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } return (tlen); } void sctp_mtu_size_reset(struct sctp_inpcb *inp, struct sctp_association *asoc, uint32_t mtu) { /* * Reset the P-MTU size on this association, this involves changing * the asoc MTU, going through ANY chunk+overhead larger than mtu to * allow the DF flag to be cleared. */ struct sctp_tmit_chunk *chk; unsigned int eff_mtu, ovh; asoc->smallest_mtu = mtu; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MIN_OVERHEAD; } else { ovh = SCTP_MIN_V4_OVERHEAD; } eff_mtu = mtu - ovh; TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { if (chk->send_size > eff_mtu) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->send_size > eff_mtu) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } } /* * given an association and starting time of the current RTT period return * RTO in number of msecs net should point to the current network */ uint32_t sctp_calculate_rto(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_nets *net, struct timeval *told, int safe) { /*- * given an association and the starting time of the current RTT * period (in value1/value2) return RTO in number of msecs. */ int calc_time = 0; int o_calctime; uint32_t new_rto = 0; int first_measure = 0; struct timeval now, then, *old; /* Copy it out for sparc64 */ if (safe == sctp_align_unsafe_makecopy) { old = &then; memcpy(&then, told, sizeof(struct timeval)); } else if (safe == sctp_align_safe_nocopy) { old = told; } else { /* error */ SCTP_PRINTF("Huh, bad rto calc call\n"); return (0); } /************************/ /* 1. calculate new RTT */ /************************/ /* get the current time */ (void)SCTP_GETTIME_TIMEVAL(&now); /* * Record the real time of the last RTT for use in DC-CC. */ net->last_measured_rtt = now; timevalsub(&net->last_measured_rtt, old); /* compute the RTT value */ if ((u_long)now.tv_sec > (u_long)old->tv_sec) { calc_time = ((u_long)now.tv_sec - (u_long)old->tv_sec) * 1000; if ((u_long)now.tv_usec > (u_long)old->tv_usec) { calc_time += (((u_long)now.tv_usec - (u_long)old->tv_usec) / 1000); } else if ((u_long)now.tv_usec < (u_long)old->tv_usec) { /* Borrow 1,000ms from current calculation */ calc_time -= 1000; /* Add in the slop over */ calc_time += ((int)now.tv_usec / 1000); /* Add in the pre-second ms's */ calc_time += (((int)1000000 - (int)old->tv_usec) / 1000); } } else if ((u_long)now.tv_sec == (u_long)old->tv_sec) { if ((u_long)now.tv_usec > (u_long)old->tv_usec) { calc_time = ((u_long)now.tv_usec - (u_long)old->tv_usec) / 1000; } else if ((u_long)now.tv_usec < (u_long)old->tv_usec) { /* impossible .. garbage in nothing out */ goto calc_rto; } else if ((u_long)now.tv_usec == (u_long)old->tv_usec) { /* * We have to have 1 usec :-D this must be the * loopback. */ calc_time = 1; } else { /* impossible .. garbage in nothing out */ goto calc_rto; } } else { /* Clock wrapped? */ goto calc_rto; } /***************************/ /* 2. update RTTVAR & SRTT */ /***************************/ net->rtt = o_calctime = calc_time; /* this is Van Jacobson's integer version */ if (net->RTO_measured) { calc_time -= (net->lastsa >> SCTP_RTT_SHIFT); /* take away 1/8th when * shift=3 */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { rto_logging(net, SCTP_LOG_RTTVAR); } net->prev_rtt = o_calctime; net->lastsa += calc_time; /* add 7/8th into sa when * shift=3 */ if (calc_time < 0) { calc_time = -calc_time; } calc_time -= (net->lastsv >> SCTP_RTT_VAR_SHIFT); /* take away 1/4 when * VAR shift=2 */ net->lastsv += calc_time; if (net->lastsv == 0) { net->lastsv = SCTP_CLOCK_GRANULARITY; } } else { /* First RTO measurment */ net->RTO_measured = 1; net->lastsa = calc_time << SCTP_RTT_SHIFT; /* Multiply by 8 when * shift=3 */ net->lastsv = calc_time; if (net->lastsv == 0) { net->lastsv = SCTP_CLOCK_GRANULARITY; } first_measure = 1; net->prev_rtt = o_calctime; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { rto_logging(net, SCTP_LOG_INITIAL_RTT); } } calc_rto: new_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; if ((new_rto > SCTP_SAT_NETWORK_MIN) && (stcb->asoc.sat_network_lockout == 0)) { stcb->asoc.sat_network = 1; } else if ((!first_measure) && stcb->asoc.sat_network) { stcb->asoc.sat_network = 0; stcb->asoc.sat_network_lockout = 1; } /* bound it, per C6/C7 in Section 5.3.1 */ if (new_rto < stcb->asoc.minrto) { new_rto = stcb->asoc.minrto; } if (new_rto > stcb->asoc.maxrto) { new_rto = stcb->asoc.maxrto; } /* we are now returning the RTO */ return (new_rto); } /* * return a pointer to a contiguous piece of data from the given mbuf chain * starting at 'off' for 'len' bytes. If the desired piece spans more than * one mbuf, a copy is made at 'ptr'. caller must ensure that the buffer size * is >= 'len' returns NULL if there there isn't 'len' bytes in the chain. */ caddr_t sctp_m_getptr(struct mbuf *m, int off, int len, uint8_t * in_ptr) { uint32_t count; uint8_t *ptr; ptr = in_ptr; if ((off < 0) || (len <= 0)) return (NULL); /* find the desired start location */ while ((m != NULL) && (off > 0)) { if (off < SCTP_BUF_LEN(m)) break; off -= SCTP_BUF_LEN(m); m = SCTP_BUF_NEXT(m); } if (m == NULL) return (NULL); /* is the current mbuf large enough (eg. contiguous)? */ if ((SCTP_BUF_LEN(m) - off) >= len) { return (mtod(m, caddr_t)+off); } else { /* else, it spans more than one mbuf, so save a temp copy... */ while ((m != NULL) && (len > 0)) { count = min(SCTP_BUF_LEN(m) - off, len); bcopy(mtod(m, caddr_t)+off, ptr, count); len -= count; ptr += count; off = 0; m = SCTP_BUF_NEXT(m); } if ((m == NULL) && (len > 0)) return (NULL); else return ((caddr_t)in_ptr); } } struct sctp_paramhdr * sctp_get_next_param(struct mbuf *m, int offset, struct sctp_paramhdr *pull, int pull_limit) { /* This just provides a typed signature to Peter's Pull routine */ return ((struct sctp_paramhdr *)sctp_m_getptr(m, offset, pull_limit, (uint8_t *) pull)); } int sctp_add_pad_tombuf(struct mbuf *m, int padlen) { /* * add padlen bytes of 0 filled padding to the end of the mbuf. If * padlen is > 3 this routine will fail. */ uint8_t *dp; int i; if (padlen > 3) { SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); return (ENOBUFS); } if (padlen <= M_TRAILINGSPACE(m)) { /* * The easy way. We hope the majority of the time we hit * here :) */ dp = (uint8_t *) (mtod(m, caddr_t)+SCTP_BUF_LEN(m)); SCTP_BUF_LEN(m) += padlen; } else { /* Hard way we must grow the mbuf */ struct mbuf *tmp; tmp = sctp_get_mbuf_for_msg(padlen, 0, M_DONTWAIT, 1, MT_DATA); if (tmp == NULL) { /* Out of space GAK! we are in big trouble. */ SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (ENOSPC); } /* setup and insert in middle */ SCTP_BUF_LEN(tmp) = padlen; SCTP_BUF_NEXT(tmp) = NULL; SCTP_BUF_NEXT(m) = tmp; dp = mtod(tmp, uint8_t *); } /* zero out the pad */ for (i = 0; i < padlen; i++) { *dp = 0; dp++; } return (0); } int sctp_pad_lastmbuf(struct mbuf *m, int padval, struct mbuf *last_mbuf) { /* find the last mbuf in chain and pad it */ struct mbuf *m_at; m_at = m; if (last_mbuf) { return (sctp_add_pad_tombuf(last_mbuf, padval)); } else { while (m_at) { if (SCTP_BUF_NEXT(m_at) == NULL) { return (sctp_add_pad_tombuf(m_at, padval)); } m_at = SCTP_BUF_NEXT(m_at); } } SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EFAULT); return (EFAULT); } static void sctp_notify_assoc_change(uint32_t event, struct sctp_tcb *stcb, uint32_t error, void *data, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_assoc_change *sac; struct sctp_queued_to_read *control; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif /* * For TCP model AND UDP connected sockets we will send an error up * when an ABORT comes in. */ if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && ((event == SCTP_COMM_LOST) || (event == SCTP_CANT_STR_ASSOC))) { if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_WAIT) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNREFUSED); stcb->sctp_socket->so_error = ECONNREFUSED; } else { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); stcb->sctp_socket->so_error = ECONNRESET; } /* Wake ANY sleepers */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif socantrcvmore(stcb->sctp_socket); sorwakeup(stcb->sctp_socket); sowwakeup(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVASSOCEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_assoc_change), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; sac = mtod(m_notify, struct sctp_assoc_change *); sac->sac_type = SCTP_ASSOC_CHANGE; sac->sac_flags = 0; sac->sac_length = sizeof(struct sctp_assoc_change); sac->sac_state = event; sac->sac_error = error; /* XXX verify these stream counts */ sac->sac_outbound_streams = stcb->asoc.streamoutcnt; sac->sac_inbound_streams = stcb->asoc.streamincnt; sac->sac_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_assoc_change); SCTP_BUF_NEXT(m_notify) = NULL; control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); if (event == SCTP_COMM_LOST) { /* Wake up any sleeper */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sctp_sowwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } } static void sctp_notify_peer_addr_change(struct sctp_tcb *stcb, uint32_t state, struct sockaddr *sa, uint32_t error) { struct mbuf *m_notify; struct sctp_paddr_change *spc; struct sctp_queued_to_read *control; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVPADDREVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_paddr_change), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) return; SCTP_BUF_LEN(m_notify) = 0; spc = mtod(m_notify, struct sctp_paddr_change *); spc->spc_type = SCTP_PEER_ADDR_CHANGE; spc->spc_flags = 0; spc->spc_length = sizeof(struct sctp_paddr_change); switch (sa->sa_family) { case AF_INET: memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in)); break; #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in6)); sin6 = (struct sockaddr_in6 *)&spc->spc_aaddr; if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) { if (sin6->sin6_scope_id == 0) { /* recover scope_id for user */ (void)sa6_recoverscope(sin6); } else { /* clear embedded scope_id for user */ in6_clearscope(&sin6->sin6_addr); } } break; } #endif default: /* TSNH */ break; } spc->spc_state = state; spc->spc_error = error; spc->spc_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_paddr_change); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); control->spec_flags = M_NOTIFICATION; /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } static void sctp_notify_send_failed(struct sctp_tcb *stcb, uint32_t error, struct sctp_tmit_chunk *chk, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_send_failed *ssf; struct sctp_queued_to_read *control; int length; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSENDFAILEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_send_failed), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; length = sizeof(struct sctp_send_failed) + chk->send_size; length -= sizeof(struct sctp_data_chunk); SCTP_BUF_LEN(m_notify) = 0; ssf = mtod(m_notify, struct sctp_send_failed *); ssf->ssf_type = SCTP_SEND_FAILED; if (error == SCTP_NOTIFY_DATAGRAM_UNSENT) ssf->ssf_flags = SCTP_DATA_UNSENT; else ssf->ssf_flags = SCTP_DATA_SENT; ssf->ssf_length = length; ssf->ssf_error = error; /* not exactly what the user sent in, but should be close :) */ bzero(&ssf->ssf_info, sizeof(ssf->ssf_info)); ssf->ssf_info.sinfo_stream = chk->rec.data.stream_number; ssf->ssf_info.sinfo_ssn = chk->rec.data.stream_seq; ssf->ssf_info.sinfo_flags = chk->rec.data.rcv_flags; ssf->ssf_info.sinfo_ppid = chk->rec.data.payloadtype; ssf->ssf_info.sinfo_context = chk->rec.data.context; ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); ssf->ssf_assoc_id = sctp_get_associd(stcb); if (chk->data) { /* * trim off the sctp chunk header(it should be there) */ if (chk->send_size >= sizeof(struct sctp_data_chunk)) { m_adj(chk->data, sizeof(struct sctp_data_chunk)); sctp_mbuf_crush(chk->data); chk->send_size -= sizeof(struct sctp_data_chunk); } } SCTP_BUF_NEXT(m_notify) = chk->data; SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_send_failed); /* Steal off the mbuf */ chk->data = NULL; /* * For this case, we check the actual socket buffer, since the assoc * is going away we don't want to overfill the socket buffer for a * non-reader */ if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); } static void sctp_notify_send_failed2(struct sctp_tcb *stcb, uint32_t error, struct sctp_stream_queue_pending *sp, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_send_failed *ssf; struct sctp_queued_to_read *control; int length; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSENDFAILEVNT)) { /* event not enabled */ return; } length = sizeof(struct sctp_send_failed) + sp->length; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_send_failed), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; ssf = mtod(m_notify, struct sctp_send_failed *); ssf->ssf_type = SCTP_SEND_FAILED; if (error == SCTP_NOTIFY_DATAGRAM_UNSENT) ssf->ssf_flags = SCTP_DATA_UNSENT; else ssf->ssf_flags = SCTP_DATA_SENT; ssf->ssf_length = length; ssf->ssf_error = error; /* not exactly what the user sent in, but should be close :) */ bzero(&ssf->ssf_info, sizeof(ssf->ssf_info)); ssf->ssf_info.sinfo_stream = sp->stream; ssf->ssf_info.sinfo_ssn = sp->strseq; if (sp->some_taken) { ssf->ssf_info.sinfo_flags = SCTP_DATA_LAST_FRAG; } else { ssf->ssf_info.sinfo_flags = SCTP_DATA_NOT_FRAG; } ssf->ssf_info.sinfo_ppid = sp->ppid; ssf->ssf_info.sinfo_context = sp->context; ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); ssf->ssf_assoc_id = sctp_get_associd(stcb); SCTP_BUF_NEXT(m_notify) = sp->data; SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_send_failed); /* Steal off the mbuf */ sp->data = NULL; /* * For this case, we check the actual socket buffer, since the assoc * is going away we don't want to overfill the socket buffer for a * non-reader */ if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); } static void sctp_notify_adaptation_layer(struct sctp_tcb *stcb, uint32_t error) { struct mbuf *m_notify; struct sctp_adaptation_event *sai; struct sctp_queued_to_read *control; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_ADAPTATIONEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_adaption_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; sai = mtod(m_notify, struct sctp_adaptation_event *); sai->sai_type = SCTP_ADAPTATION_INDICATION; sai->sai_flags = 0; sai->sai_length = sizeof(struct sctp_adaptation_event); sai->sai_adaptation_ind = stcb->asoc.peers_adaptation; sai->sai_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_adaptation_event); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); control->spec_flags = M_NOTIFICATION; /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } /* This always must be called with the read-queue LOCKED in the INP */ static void sctp_notify_partial_delivery_indication(struct sctp_tcb *stcb, uint32_t error, uint32_t val, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_pdapi_event *pdapi; struct sctp_queued_to_read *control; struct sockbuf *sb; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_PDAPIEVNT)) { /* event not enabled */ return; } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_CANT_READ) { return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_pdapi_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; pdapi = mtod(m_notify, struct sctp_pdapi_event *); pdapi->pdapi_type = SCTP_PARTIAL_DELIVERY_EVENT; pdapi->pdapi_flags = 0; pdapi->pdapi_length = sizeof(struct sctp_pdapi_event); pdapi->pdapi_indication = error; pdapi->pdapi_stream = (val >> 16); pdapi->pdapi_seq = (val & 0x0000ffff); pdapi->pdapi_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_pdapi_event); SCTP_BUF_NEXT(m_notify) = NULL; control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; control->held_length = 0; control->length = 0; sb = &stcb->sctp_socket->so_rcv; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m_notify)); } sctp_sballoc(stcb, sb, m_notify); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_add_int(&control->length, SCTP_BUF_LEN(m_notify)); control->end_added = 1; if (stcb->asoc.control_pdapi) TAILQ_INSERT_AFTER(&stcb->sctp_ep->read_queue, stcb->asoc.control_pdapi, control, next); else { /* we really should not see this case */ TAILQ_INSERT_TAIL(&stcb->sctp_ep->read_queue, control, next); } if (stcb->sctp_ep && stcb->sctp_socket) { /* This should always be the case */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } } static void sctp_notify_shutdown_event(struct sctp_tcb *stcb) { struct mbuf *m_notify; struct sctp_shutdown_event *sse; struct sctp_queued_to_read *control; /* * For TCP model AND UDP connected sockets we will send an error up * when an SHUTDOWN completes */ if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* mark socket closed for read/write and wakeup! */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } #endif socantsendmore(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; sse = mtod(m_notify, struct sctp_shutdown_event *); sse->sse_type = SCTP_SHUTDOWN_EVENT; sse->sse_flags = 0; sse->sse_length = sizeof(struct sctp_shutdown_event); sse->sse_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_shutdown_event); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } static void sctp_notify_sender_dry_event(struct sctp_tcb *stcb, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_sender_dry_event *event; struct sctp_queued_to_read *control; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_DRYEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_sender_dry_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) { /* no space left */ return; } SCTP_BUF_LEN(m_notify) = 0; event = mtod(m_notify, struct sctp_sender_dry_event *); event->sender_dry_type = SCTP_SENDER_DRY_EVENT; event->sender_dry_flags = 0; event->sender_dry_length = sizeof(struct sctp_sender_dry_event); event->sender_dry_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_sender_dry_event); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); control->spec_flags = M_NOTIFICATION; /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); } static void sctp_notify_stream_reset_add(struct sctp_tcb *stcb, int number_entries, int flag) { struct mbuf *m_notify; struct sctp_queued_to_read *control; struct sctp_stream_reset_event *strreset; int len; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; len = sizeof(struct sctp_stream_reset_event) + (number_entries * sizeof(uint16_t)); if (len > M_TRAILINGSPACE(m_notify)) { /* never enough room */ sctp_m_freem(m_notify); return; } strreset = mtod(m_notify, struct sctp_stream_reset_event *); strreset->strreset_type = SCTP_STREAM_RESET_EVENT; strreset->strreset_flags = SCTP_STRRESET_ADD_STREAM | flag; strreset->strreset_length = len; strreset->strreset_assoc_id = sctp_get_associd(stcb); strreset->strreset_list[0] = number_entries; SCTP_BUF_LEN(m_notify) = len; SCTP_BUF_NEXT(m_notify) = NULL; if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { /* no space */ sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } static void sctp_notify_stream_reset(struct sctp_tcb *stcb, int number_entries, uint16_t * list, int flag) { struct mbuf *m_notify; struct sctp_queued_to_read *control; struct sctp_stream_reset_event *strreset; int len; if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; len = sizeof(struct sctp_stream_reset_event) + (number_entries * sizeof(uint16_t)); if (len > M_TRAILINGSPACE(m_notify)) { /* never enough room */ sctp_m_freem(m_notify); return; } strreset = mtod(m_notify, struct sctp_stream_reset_event *); strreset->strreset_type = SCTP_STREAM_RESET_EVENT; if (number_entries == 0) { strreset->strreset_flags = flag | SCTP_STRRESET_ALL_STREAMS; } else { strreset->strreset_flags = flag | SCTP_STRRESET_STREAM_LIST; } strreset->strreset_length = len; strreset->strreset_assoc_id = sctp_get_associd(stcb); if (number_entries) { int i; for (i = 0; i < number_entries; i++) { strreset->strreset_list[i] = ntohs(list[i]); } } SCTP_BUF_LEN(m_notify) = len; SCTP_BUF_NEXT(m_notify) = NULL; if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { /* no space */ sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); } void sctp_ulp_notify(uint32_t notification, struct sctp_tcb *stcb, uint32_t error, void *data, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { if ((stcb == NULL) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { /* If the socket is gone we are out of here */ return; } if (stcb->sctp_socket->so_rcv.sb_state & SBS_CANTRCVMORE) { return; } if (stcb && ((stcb->asoc.state & SCTP_STATE_COOKIE_WAIT) || (stcb->asoc.state & SCTP_STATE_COOKIE_ECHOED))) { if ((notification == SCTP_NOTIFY_INTERFACE_DOWN) || (notification == SCTP_NOTIFY_INTERFACE_UP) || (notification == SCTP_NOTIFY_INTERFACE_CONFIRMED)) { /* Don't report these in front states */ return; } } switch (notification) { case SCTP_NOTIFY_ASSOC_UP: if (stcb->asoc.assoc_up_sent == 0) { sctp_notify_assoc_change(SCTP_COMM_UP, stcb, error, NULL, so_locked); stcb->asoc.assoc_up_sent = 1; } if (stcb->asoc.adaptation_needed && (stcb->asoc.adaptation_sent == 0)) { sctp_notify_adaptation_layer(stcb, error); } if (stcb->asoc.peer_supports_auth == 0) { sctp_ulp_notify(SCTP_NOTIFY_NO_PEER_AUTH, stcb, 0, NULL, so_locked); } break; case SCTP_NOTIFY_ASSOC_DOWN: sctp_notify_assoc_change(SCTP_SHUTDOWN_COMP, stcb, error, NULL, so_locked); break; case SCTP_NOTIFY_INTERFACE_DOWN: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_UNREACHABLE, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_INTERFACE_UP: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_AVAILABLE, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_INTERFACE_CONFIRMED: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_CONFIRMED, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_SPECIAL_SP_FAIL: sctp_notify_send_failed2(stcb, error, (struct sctp_stream_queue_pending *)data, so_locked); break; case SCTP_NOTIFY_DG_FAIL: sctp_notify_send_failed(stcb, error, (struct sctp_tmit_chunk *)data, so_locked); break; case SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION: { uint32_t val; val = *((uint32_t *) data); sctp_notify_partial_delivery_indication(stcb, error, val, so_locked); break; } case SCTP_NOTIFY_STRDATA_ERR: break; case SCTP_NOTIFY_ASSOC_ABORTED: if ((stcb) && (((stcb->asoc.state & SCTP_STATE_MASK) == SCTP_STATE_COOKIE_WAIT) || ((stcb->asoc.state & SCTP_STATE_MASK) == SCTP_STATE_COOKIE_ECHOED))) { sctp_notify_assoc_change(SCTP_CANT_STR_ASSOC, stcb, error, NULL, so_locked); } else { sctp_notify_assoc_change(SCTP_COMM_LOST, stcb, error, NULL, so_locked); } break; case SCTP_NOTIFY_PEER_OPENED_STREAM: break; case SCTP_NOTIFY_STREAM_OPENED_OK: break; case SCTP_NOTIFY_ASSOC_RESTART: sctp_notify_assoc_change(SCTP_RESTART, stcb, error, data, so_locked); if (stcb->asoc.peer_supports_auth == 0) { sctp_ulp_notify(SCTP_NOTIFY_NO_PEER_AUTH, stcb, 0, NULL, so_locked); } break; case SCTP_NOTIFY_HB_RESP: break; case SCTP_NOTIFY_STR_RESET_INSTREAM_ADD_OK: sctp_notify_stream_reset_add(stcb, error, SCTP_STRRESET_INBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_ADD_OK: sctp_notify_stream_reset_add(stcb, error, SCTP_STRRESET_OUTBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_ADD_FAIL: sctp_notify_stream_reset_add(stcb, error, (SCTP_STRRESET_FAILED | SCTP_STRRESET_OUTBOUND_STR)); break; case SCTP_NOTIFY_STR_RESET_SEND: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STRRESET_OUTBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_RECV: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STRRESET_INBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_FAILED_OUT: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), (SCTP_STRRESET_OUTBOUND_STR | SCTP_STRRESET_FAILED)); break; case SCTP_NOTIFY_STR_RESET_FAILED_IN: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), (SCTP_STRRESET_INBOUND_STR | SCTP_STRRESET_FAILED)); break; case SCTP_NOTIFY_ASCONF_ADD_IP: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_ADDED, data, error); break; case SCTP_NOTIFY_ASCONF_DELETE_IP: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_REMOVED, data, error); break; case SCTP_NOTIFY_ASCONF_SET_PRIMARY: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_MADE_PRIM, data, error); break; case SCTP_NOTIFY_ASCONF_SUCCESS: break; case SCTP_NOTIFY_ASCONF_FAILED: break; case SCTP_NOTIFY_PEER_SHUTDOWN: sctp_notify_shutdown_event(stcb); break; case SCTP_NOTIFY_AUTH_NEW_KEY: sctp_notify_authentication(stcb, SCTP_AUTH_NEWKEY, error, (uint16_t) (uintptr_t) data, so_locked); break; case SCTP_NOTIFY_AUTH_FREE_KEY: sctp_notify_authentication(stcb, SCTP_AUTH_FREE_KEY, error, (uint16_t) (uintptr_t) data, so_locked); break; case SCTP_NOTIFY_NO_PEER_AUTH: sctp_notify_authentication(stcb, SCTP_AUTH_NO_AUTH, error, (uint16_t) (uintptr_t) data, so_locked); break; case SCTP_NOTIFY_SENDER_DRY: sctp_notify_sender_dry_event(stcb, so_locked); break; default: SCTPDBG(SCTP_DEBUG_UTIL1, "%s: unknown notification %xh (%u)\n", __FUNCTION__, notification, notification); break; } /* end switch */ } void sctp_report_all_outbound(struct sctp_tcb *stcb, int holds_lock, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct sctp_association *asoc; struct sctp_stream_out *outs; struct sctp_tmit_chunk *chk, *nchk; struct sctp_stream_queue_pending *sp, *nsp; int i; if (stcb == NULL) { return; } asoc = &stcb->asoc; if (asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* already being freed */ return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (asoc->state & SCTP_STATE_CLOSED_SOCKET)) { return; } /* now through all the gunk freeing chunks */ if (holds_lock == 0) { SCTP_TCB_SEND_LOCK(stcb); } /* sent queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); asoc->sent_queue_cnt--; if (chk->data != NULL) { sctp_free_bufspace(stcb, asoc, chk, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_SENT, chk, so_locked); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ } /* pending send queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); asoc->send_queue_cnt--; if (chk->data != NULL) { sctp_free_bufspace(stcb, asoc, chk, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, chk, so_locked); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ } for (i = 0; i < asoc->streamoutcnt; i++) { /* For each stream */ outs = &asoc->strmout[i]; /* clean up any sends there */ asoc->locked_on_sending = NULL; TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { asoc->stream_queue_cnt--; TAILQ_REMOVE(&outs->outqueue, sp, next); sctp_free_spbufspace(stcb, asoc, sp); if (sp->data) { sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, (void *)sp, so_locked); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } } if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } /* Free the chunk */ sctp_free_a_strmoq(stcb, sp); /* sa_ignore FREED_MEMORY */ } } if (holds_lock == 0) { SCTP_TCB_SEND_UNLOCK(stcb); } } void sctp_abort_notification(struct sctp_tcb *stcb, int error, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { if (stcb == NULL) { return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { return; } /* Tell them we lost the asoc */ sctp_report_all_outbound(stcb, 1, so_locked); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_WAS_ABORTED; } sctp_ulp_notify(SCTP_NOTIFY_ASSOC_ABORTED, stcb, error, NULL, so_locked); } void sctp_abort_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct mbuf *m, int iphlen, struct sctphdr *sh, struct mbuf *op_err, uint32_t vrf_id, uint16_t port) { uint32_t vtag; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif vtag = 0; if (stcb != NULL) { /* We have a TCB to abort, send notification too */ vtag = stcb->asoc.peer_vtag; sctp_abort_notification(stcb, 0, SCTP_SO_NOT_LOCKED); /* get the assoc vrf id and table id */ vrf_id = stcb->asoc.vrf_id; stcb->asoc.state |= SCTP_STATE_WAS_ABORTED; } sctp_send_abort(m, iphlen, sh, vtag, op_err, vrf_id, port); if (stcb != NULL) { /* Ok, now lets free it */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_4); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } } #ifdef SCTP_ASOCLOG_OF_TSNS void sctp_print_out_track_log(struct sctp_tcb *stcb) { #ifdef NOSIY_PRINTS int i; SCTP_PRINTF("Last ep reason:%x\n", stcb->sctp_ep->last_abort_code); SCTP_PRINTF("IN bound TSN log-aaa\n"); if ((stcb->asoc.tsn_in_at == 0) && (stcb->asoc.tsn_in_wrapped == 0)) { SCTP_PRINTF("None rcvd\n"); goto none_in; } if (stcb->asoc.tsn_in_wrapped) { for (i = stcb->asoc.tsn_in_at; i < SCTP_TSN_LOG_SIZE; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.in_tsnlog[i].tsn, stcb->asoc.in_tsnlog[i].strm, stcb->asoc.in_tsnlog[i].seq, stcb->asoc.in_tsnlog[i].flgs, stcb->asoc.in_tsnlog[i].sz); } } if (stcb->asoc.tsn_in_at) { for (i = 0; i < stcb->asoc.tsn_in_at; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.in_tsnlog[i].tsn, stcb->asoc.in_tsnlog[i].strm, stcb->asoc.in_tsnlog[i].seq, stcb->asoc.in_tsnlog[i].flgs, stcb->asoc.in_tsnlog[i].sz); } } none_in: SCTP_PRINTF("OUT bound TSN log-aaa\n"); if ((stcb->asoc.tsn_out_at == 0) && (stcb->asoc.tsn_out_wrapped == 0)) { SCTP_PRINTF("None sent\n"); } if (stcb->asoc.tsn_out_wrapped) { for (i = stcb->asoc.tsn_out_at; i < SCTP_TSN_LOG_SIZE; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.out_tsnlog[i].tsn, stcb->asoc.out_tsnlog[i].strm, stcb->asoc.out_tsnlog[i].seq, stcb->asoc.out_tsnlog[i].flgs, stcb->asoc.out_tsnlog[i].sz); } } if (stcb->asoc.tsn_out_at) { for (i = 0; i < stcb->asoc.tsn_out_at; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.out_tsnlog[i].tsn, stcb->asoc.out_tsnlog[i].strm, stcb->asoc.out_tsnlog[i].seq, stcb->asoc.out_tsnlog[i].flgs, stcb->asoc.out_tsnlog[i].sz); } } #endif } #endif void sctp_abort_an_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int error, struct mbuf *op_err, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { uint32_t vtag; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); #endif if (stcb == NULL) { /* Got to have a TCB */ if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { if (LIST_FIRST(&inp->sctp_asoc_list) == NULL) { sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); } } return; } else { stcb->asoc.state |= SCTP_STATE_WAS_ABORTED; } vtag = stcb->asoc.peer_vtag; /* notify the ulp */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) sctp_abort_notification(stcb, error, so_locked); /* notify the peer */ #if defined(SCTP_PANIC_ON_ABORT) panic("aborting an association"); #endif sctp_send_abort_tcb(stcb, op_err, so_locked); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } /* now free the asoc */ #ifdef SCTP_ASOCLOG_OF_TSNS sctp_print_out_track_log(stcb); #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_5); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } void sctp_handle_ootb(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb *inp, struct mbuf *op_err, uint32_t vrf_id, uint16_t port) { struct sctp_chunkhdr *ch, chunk_buf; unsigned int chk_length; SCTP_STAT_INCR_COUNTER32(sctps_outoftheblue); /* Generate a TO address for future reference */ if (inp && (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { if (LIST_FIRST(&inp->sctp_asoc_list) == NULL) { sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); } } ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* break to abort land */ break; } switch (ch->chunk_type) { case SCTP_COOKIE_ECHO: /* We hit here only if the assoc is being freed */ return; case SCTP_PACKET_DROPPED: /* we don't respond to pkt-dropped */ return; case SCTP_ABORT_ASSOCIATION: /* we don't respond with an ABORT to an ABORT */ return; case SCTP_SHUTDOWN_COMPLETE: /* * we ignore it since we are not waiting for it and * peer is gone */ return; case SCTP_SHUTDOWN_ACK: sctp_send_shutdown_complete2(m, iphlen, sh, vrf_id, port); return; default: break; } offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } sctp_send_abort(m, iphlen, sh, 0, op_err, vrf_id, port); } /* * check the inbound datagram to make sure there is not an abort inside it, * if there is return 1, else return 0. */ int sctp_is_there_an_abort_here(struct mbuf *m, int iphlen, uint32_t * vtagfill) { struct sctp_chunkhdr *ch; struct sctp_init_chunk *init_chk, chunk_buf; int offset; unsigned int chk_length; offset = iphlen + sizeof(struct sctphdr); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* packet is probably corrupt */ break; } /* we seem to be ok, is it an abort? */ if (ch->chunk_type == SCTP_ABORT_ASSOCIATION) { /* yep, tell them */ return (1); } if (ch->chunk_type == SCTP_INITIATION) { /* need to update the Vtag */ init_chk = (struct sctp_init_chunk *)sctp_m_getptr(m, offset, sizeof(*init_chk), (uint8_t *) & chunk_buf); if (init_chk != NULL) { *vtagfill = ntohl(init_chk->init.initiate_tag); } } /* Nope, move to the next chunk */ offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } return (0); } /* * currently (2/02), ifa_addr embeds scope_id's and don't have sin6_scope_id * set (i.e. it's 0) so, create this function to compare link local scopes */ #ifdef INET6 uint32_t sctp_is_same_scope(struct sockaddr_in6 *addr1, struct sockaddr_in6 *addr2) { struct sockaddr_in6 a, b; /* save copies */ a = *addr1; b = *addr2; if (a.sin6_scope_id == 0) if (sa6_recoverscope(&a)) { /* can't get scope, so can't match */ return (0); } if (b.sin6_scope_id == 0) if (sa6_recoverscope(&b)) { /* can't get scope, so can't match */ return (0); } if (a.sin6_scope_id != b.sin6_scope_id) return (0); return (1); } /* * returns a sockaddr_in6 with embedded scope recovered and removed */ struct sockaddr_in6 * sctp_recover_scope(struct sockaddr_in6 *addr, struct sockaddr_in6 *store) { /* check and strip embedded scope junk */ if (addr->sin6_family == AF_INET6) { if (IN6_IS_SCOPE_LINKLOCAL(&addr->sin6_addr)) { if (addr->sin6_scope_id == 0) { *store = *addr; if (!sa6_recoverscope(store)) { /* use the recovered scope */ addr = store; } } else { /* else, return the original "to" addr */ in6_clearscope(&addr->sin6_addr); } } } return (addr); } #endif /* * are the two addresses the same? currently a "scopeless" check returns: 1 * if same, 0 if not */ int sctp_cmpaddr(struct sockaddr *sa1, struct sockaddr *sa2) { /* must be valid */ if (sa1 == NULL || sa2 == NULL) return (0); /* must be the same family */ if (sa1->sa_family != sa2->sa_family) return (0); switch (sa1->sa_family) { #ifdef INET6 case AF_INET6: { /* IPv6 addresses */ struct sockaddr_in6 *sin6_1, *sin6_2; sin6_1 = (struct sockaddr_in6 *)sa1; sin6_2 = (struct sockaddr_in6 *)sa2; return (SCTP6_ARE_ADDR_EQUAL(sin6_1, sin6_2)); } #endif case AF_INET: { /* IPv4 addresses */ struct sockaddr_in *sin_1, *sin_2; sin_1 = (struct sockaddr_in *)sa1; sin_2 = (struct sockaddr_in *)sa2; return (sin_1->sin_addr.s_addr == sin_2->sin_addr.s_addr); } default: /* we don't do these... */ return (0); } } void sctp_print_address(struct sockaddr *sa) { #ifdef INET6 char ip6buf[INET6_ADDRSTRLEN]; ip6buf[0] = 0; #endif switch (sa->sa_family) { #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; SCTP_PRINTF("IPv6 address: %s:port:%d scope:%u\n", ip6_sprintf(ip6buf, &sin6->sin6_addr), ntohs(sin6->sin6_port), sin6->sin6_scope_id); break; } #endif case AF_INET: { struct sockaddr_in *sin; unsigned char *p; sin = (struct sockaddr_in *)sa; p = (unsigned char *)&sin->sin_addr; SCTP_PRINTF("IPv4 address: %u.%u.%u.%u:%d\n", p[0], p[1], p[2], p[3], ntohs(sin->sin_port)); break; } default: SCTP_PRINTF("?\n"); break; } } void sctp_print_address_pkt(struct ip *iph, struct sctphdr *sh) { switch (iph->ip_v) { case IPVERSION: { struct sockaddr_in lsa, fsa; bzero(&lsa, sizeof(lsa)); lsa.sin_len = sizeof(lsa); lsa.sin_family = AF_INET; lsa.sin_addr = iph->ip_src; lsa.sin_port = sh->src_port; bzero(&fsa, sizeof(fsa)); fsa.sin_len = sizeof(fsa); fsa.sin_family = AF_INET; fsa.sin_addr = iph->ip_dst; fsa.sin_port = sh->dest_port; SCTP_PRINTF("src: "); sctp_print_address((struct sockaddr *)&lsa); SCTP_PRINTF("dest: "); sctp_print_address((struct sockaddr *)&fsa); break; } #ifdef INET6 case IPV6_VERSION >> 4: { struct ip6_hdr *ip6; struct sockaddr_in6 lsa6, fsa6; ip6 = (struct ip6_hdr *)iph; bzero(&lsa6, sizeof(lsa6)); lsa6.sin6_len = sizeof(lsa6); lsa6.sin6_family = AF_INET6; lsa6.sin6_addr = ip6->ip6_src; lsa6.sin6_port = sh->src_port; bzero(&fsa6, sizeof(fsa6)); fsa6.sin6_len = sizeof(fsa6); fsa6.sin6_family = AF_INET6; fsa6.sin6_addr = ip6->ip6_dst; fsa6.sin6_port = sh->dest_port; SCTP_PRINTF("src: "); sctp_print_address((struct sockaddr *)&lsa6); SCTP_PRINTF("dest: "); sctp_print_address((struct sockaddr *)&fsa6); break; } #endif default: /* TSNH */ break; } } void sctp_pull_off_control_to_new_inp(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, struct sctp_tcb *stcb, int waitflags) { /* * go through our old INP and pull off any control structures that * belong to stcb and move then to the new inp. */ struct socket *old_so, *new_so; struct sctp_queued_to_read *control, *nctl; struct sctp_readhead tmp_queue; struct mbuf *m; int error = 0; old_so = old_inp->sctp_socket; new_so = new_inp->sctp_socket; TAILQ_INIT(&tmp_queue); error = sblock(&old_so->so_rcv, waitflags); if (error) { /* * Gak, can't get sblock, we have a problem. data will be * left stranded.. and we don't dare look at it since the * other thread may be reading something. Oh well, its a * screwed up app that does a peeloff OR a accept while * reading from the main socket... actually its only the * peeloff() case, since I think read will fail on a * listening socket.. */ return; } /* lock the socket buffers */ SCTP_INP_READ_LOCK(old_inp); TAILQ_FOREACH_SAFE(control, &old_inp->read_queue, next, nctl) { /* Pull off all for out target stcb */ if (control->stcb == stcb) { /* remove it we want it */ TAILQ_REMOVE(&old_inp->read_queue, control, next); TAILQ_INSERT_TAIL(&tmp_queue, control, next); m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &old_so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } } } SCTP_INP_READ_UNLOCK(old_inp); /* Remove the sb-lock on the old socket */ sbunlock(&old_so->so_rcv); /* Now we move them over to the new socket buffer */ SCTP_INP_READ_LOCK(new_inp); TAILQ_FOREACH_SAFE(control, &tmp_queue, next, nctl) { TAILQ_INSERT_TAIL(&new_inp->read_queue, control, next); m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); } sctp_sballoc(stcb, &new_so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } } SCTP_INP_READ_UNLOCK(new_inp); } void sctp_add_to_readq(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_queued_to_read *control, struct sockbuf *sb, int end, int inp_read_lock_held, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /* * Here we must place the control on the end of the socket read * queue AND increment sb_cc so that select will work properly on * read. */ struct mbuf *m, *prev = NULL; if (inp == NULL) { /* Gak, TSNH!! */ #ifdef INVARIANTS panic("Gak, inp NULL on add_to_readq"); #endif return; } if (inp_read_lock_held == 0) SCTP_INP_READ_LOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_CANT_READ) { sctp_free_remote_addr(control->whoFrom); if (control->data) { sctp_m_freem(control->data); control->data = NULL; } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), control); if (inp_read_lock_held == 0) SCTP_INP_READ_UNLOCK(inp); return; } if (!(control->spec_flags & M_NOTIFICATION)) { atomic_add_int(&inp->total_recvs, 1); if (!control->do_not_ref_stcb) { atomic_add_int(&stcb->total_recvs, 1); } } m = control->data; control->held_length = 0; control->length = 0; while (m) { if (SCTP_BUF_LEN(m) == 0) { /* Skip mbufs with NO length */ if (prev == NULL) { /* First one */ control->data = sctp_m_free(m); m = control->data; } else { SCTP_BUF_NEXT(prev) = sctp_m_free(m); m = SCTP_BUF_NEXT(prev); } if (m == NULL) { control->tail_mbuf = prev; } continue; } prev = m; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); } sctp_sballoc(stcb, sb, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_add_int(&control->length, SCTP_BUF_LEN(m)); m = SCTP_BUF_NEXT(m); } if (prev != NULL) { control->tail_mbuf = prev; } else { /* Everything got collapsed out?? */ sctp_free_remote_addr(control->whoFrom); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), control); if (inp_read_lock_held == 0) SCTP_INP_READ_UNLOCK(inp); return; } if (end) { control->end_added = 1; } TAILQ_INSERT_TAIL(&inp->read_queue, control, next); if (inp_read_lock_held == 0) SCTP_INP_READ_UNLOCK(inp); if (inp && inp->sctp_socket) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } else { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(inp); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sctp_sorwakeup(inp, inp->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } } } int sctp_append_to_readq(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_queued_to_read *control, struct mbuf *m, int end, int ctls_cumack, struct sockbuf *sb) { /* * A partial delivery API event is underway. OR we are appending on * the reassembly queue. * * If PDAPI this means we need to add m to the end of the data. * Increase the length in the control AND increment the sb_cc. * Otherwise sb is NULL and all we need to do is put it at the end * of the mbuf chain. */ int len = 0; struct mbuf *mm, *tail = NULL, *prev = NULL; if (inp) { SCTP_INP_READ_LOCK(inp); } if (control == NULL) { get_out: if (inp) { SCTP_INP_READ_UNLOCK(inp); } return (-1); } if (inp && (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_CANT_READ)) { SCTP_INP_READ_UNLOCK(inp); return 0; } if (control->end_added) { /* huh this one is complete? */ goto get_out; } mm = m; if (mm == NULL) { goto get_out; } while (mm) { if (SCTP_BUF_LEN(mm) == 0) { /* Skip mbufs with NO lenght */ if (prev == NULL) { /* First one */ m = sctp_m_free(mm); mm = m; } else { SCTP_BUF_NEXT(prev) = sctp_m_free(mm); mm = SCTP_BUF_NEXT(prev); } continue; } prev = mm; len += SCTP_BUF_LEN(mm); if (sb) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(mm)); } sctp_sballoc(stcb, sb, mm); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } } mm = SCTP_BUF_NEXT(mm); } if (prev) { tail = prev; } else { /* Really there should always be a prev */ if (m == NULL) { /* Huh nothing left? */ #ifdef INVARIANTS panic("Nothing left to add?"); #else goto get_out; #endif } tail = m; } if (control->tail_mbuf) { /* append */ SCTP_BUF_NEXT(control->tail_mbuf) = m; control->tail_mbuf = tail; } else { /* nothing there */ #ifdef INVARIANTS if (control->data != NULL) { panic("This should NOT happen"); } #endif control->data = m; control->tail_mbuf = tail; } atomic_add_int(&control->length, len); if (end) { /* message is complete */ if (stcb && (control == stcb->asoc.control_pdapi)) { stcb->asoc.control_pdapi = NULL; } control->held_length = 0; control->end_added = 1; } if (stcb == NULL) { control->do_not_ref_stcb = 1; } /* * When we are appending in partial delivery, the cum-ack is used * for the actual pd-api highest tsn on this mbuf. The true cum-ack * is populated in the outbound sinfo structure from the true cumack * if the association exists... */ control->sinfo_tsn = control->sinfo_cumtsn = ctls_cumack; if (inp) { SCTP_INP_READ_UNLOCK(inp); } if (inp && inp->sctp_socket) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } else { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { SCTP_SOCKET_UNLOCK(so, 1); return (0); } #endif sctp_sorwakeup(inp, inp->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } } return (0); } /*************HOLD THIS COMMENT FOR PATCH FILE OF *************ALTERNATE ROUTING CODE */ /*************HOLD THIS COMMENT FOR END OF PATCH FILE OF *************ALTERNATE ROUTING CODE */ struct mbuf * sctp_generate_invmanparam(int err) { /* Return a MBUF with a invalid mandatory parameter */ struct mbuf *m; m = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (m) { struct sctp_paramhdr *ph; SCTP_BUF_LEN(m) = sizeof(struct sctp_paramhdr); ph = mtod(m, struct sctp_paramhdr *); ph->param_length = htons(sizeof(struct sctp_paramhdr)); ph->param_type = htons(err); } return (m); } #ifdef SCTP_MBCNT_LOGGING void sctp_free_bufspace(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_tmit_chunk *tp1, int chk_cnt) { if (tp1->data == NULL) { return; } asoc->chunks_on_out_queue -= chk_cnt; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBCNT_LOGGING_ENABLE) { sctp_log_mbcnt(SCTP_LOG_MBCNT_DECREASE, asoc->total_output_queue_size, tp1->book_size, 0, tp1->mbcnt); } if (asoc->total_output_queue_size >= tp1->book_size) { atomic_add_int(&asoc->total_output_queue_size, -tp1->book_size); } else { asoc->total_output_queue_size = 0; } if (stcb->sctp_socket && (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) || ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)))) { if (stcb->sctp_socket->so_snd.sb_cc >= tp1->book_size) { stcb->sctp_socket->so_snd.sb_cc -= tp1->book_size; } else { stcb->sctp_socket->so_snd.sb_cc = 0; } } } #endif int sctp_release_pr_sctp_chunk(struct sctp_tcb *stcb, struct sctp_tmit_chunk *tp1, int reason, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct sctp_stream_out *strq; struct sctp_tmit_chunk *chk = NULL, *tp2; struct sctp_stream_queue_pending *sp; uint16_t stream = 0, seq = 0; uint8_t foundeom = 0; int ret_sz = 0; int notdone; int do_wakeup_routine = 0; stream = tp1->rec.data.stream_number; seq = tp1->rec.data.stream_seq; do { ret_sz += tp1->book_size; if (tp1->data != NULL) { if (tp1->sent < SCTP_DATAGRAM_RESEND) { sctp_flight_size_decrease(tp1); sctp_total_flight_decrease(stcb, tp1); } sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); stcb->asoc.peers_rwnd += tp1->send_size; stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, reason, tp1, so_locked); if (tp1->data) { sctp_m_freem(tp1->data); tp1->data = NULL; } do_wakeup_routine = 1; if (PR_SCTP_BUF_ENABLED(tp1->flags)) { stcb->asoc.sent_queue_cnt_removeable--; } } tp1->sent = SCTP_FORWARD_TSN_SKIP; if ((tp1->rec.data.rcv_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { /* not frag'ed we ae done */ notdone = 0; foundeom = 1; } else if (tp1->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { /* end of frag, we are done */ notdone = 0; foundeom = 1; } else { /* * Its a begin or middle piece, we must mark all of * it */ notdone = 1; tp1 = TAILQ_NEXT(tp1, sctp_next); } } while (tp1 && notdone); if (foundeom == 0) { /* * The multi-part message was scattered across the send and * sent queue. */ TAILQ_FOREACH_SAFE(tp1, &stcb->asoc.send_queue, sctp_next, tp2) { if ((tp1->rec.data.stream_number != stream) || (tp1->rec.data.stream_seq != seq)) { break; } /* * save to chk in case we have some on stream out * queue. If so and we have an un-transmitted one we * don't have to fudge the TSN. */ chk = tp1; ret_sz += tp1->book_size; sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, reason, tp1, so_locked); if (tp1->data) { sctp_m_freem(tp1->data); tp1->data = NULL; } /* No flight involved here book the size to 0 */ tp1->book_size = 0; if (tp1->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { foundeom = 1; } do_wakeup_routine = 1; tp1->sent = SCTP_FORWARD_TSN_SKIP; TAILQ_REMOVE(&stcb->asoc.send_queue, tp1, sctp_next); /* * on to the sent queue so we can wait for it to be * passed by. */ TAILQ_INSERT_TAIL(&stcb->asoc.sent_queue, tp1, sctp_next); stcb->asoc.send_queue_cnt--; stcb->asoc.sent_queue_cnt++; } } if (foundeom == 0) { /* * Still no eom found. That means there is stuff left on the * stream out queue.. yuck. */ strq = &stcb->asoc.strmout[stream]; SCTP_TCB_SEND_LOCK(stcb); TAILQ_FOREACH(sp, &strq->outqueue, next) { /* FIXME: Shouldn't this be a serial number check? */ if (sp->strseq > seq) { break; } /* Check if its our SEQ */ if (sp->strseq == seq) { sp->discard_rest = 1; /* * We may need to put a chunk on the queue * that holds the TSN that would have been * sent with the LAST bit. */ if (chk == NULL) { /* Yep, we have to */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* * we are hosed. All we can * do is nothing.. which * will cause an abort if * the peer is paying * attention. */ goto oh_well; } memset(chk, 0, sizeof(*chk)); chk->rec.data.rcv_flags = SCTP_DATA_LAST_FRAG; chk->sent = SCTP_FORWARD_TSN_SKIP; chk->asoc = &stcb->asoc; chk->rec.data.stream_seq = sp->strseq; chk->rec.data.stream_number = sp->stream; chk->rec.data.payloadtype = sp->ppid; chk->rec.data.context = sp->context; chk->flags = sp->act_flags; if (sp->net) chk->whoTo = sp->net; else chk->whoTo = stcb->asoc.primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); chk->rec.data.TSN_seq = atomic_fetchadd_int(&stcb->asoc.sending_seq, 1); stcb->asoc.pr_sctp_cnt++; chk->pr_sctp_on = 1; TAILQ_INSERT_TAIL(&stcb->asoc.sent_queue, chk, sctp_next); stcb->asoc.sent_queue_cnt++; stcb->asoc.pr_sctp_cnt++; } else { chk->rec.data.rcv_flags |= SCTP_DATA_LAST_FRAG; } oh_well: if (sp->data) { /* * Pull any data to free up the SB * and allow sender to "add more" * whilc we will throw away :-) */ sctp_free_spbufspace(stcb, &stcb->asoc, sp); ret_sz += sp->length; do_wakeup_routine = 1; sp->some_taken = 1; sctp_m_freem(sp->data); sp->length = 0; sp->data = NULL; sp->tail_mbuf = NULL; } break; } } /* End tailq_foreach */ SCTP_TCB_SEND_UNLOCK(stcb); } if (do_wakeup_routine) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* assoc was freed while we were unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return (ret_sz); } } #endif sctp_sowwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } return (ret_sz); } /* * checks to see if the given address, sa, is one that is currently known by * the kernel note: can't distinguish the same address on multiple interfaces * and doesn't handle multiple addresses with different zone/scope id's note: * ifa_ifwithaddr() compares the entire sockaddr struct */ struct sctp_ifa * sctp_find_ifa_in_ep(struct sctp_inpcb *inp, struct sockaddr *addr, int holds_lock) { struct sctp_laddr *laddr; if (holds_lock == 0) { SCTP_INP_RLOCK(inp); } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) continue; if (addr->sa_family != laddr->ifa->address.sa.sa_family) continue; if (addr->sa_family == AF_INET) { if (((struct sockaddr_in *)addr)->sin_addr.s_addr == laddr->ifa->address.sin.sin_addr.s_addr) { /* found him. */ if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (laddr->ifa); break; } } #ifdef INET6 if (addr->sa_family == AF_INET6) { if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, &laddr->ifa->address.sin6)) { /* found him. */ if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (laddr->ifa); break; } } #endif } if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (NULL); } uint32_t sctp_get_ifa_hash_val(struct sockaddr *addr) { if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; return (sin->sin_addr.s_addr ^ (sin->sin_addr.s_addr >> 16)); } else if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; uint32_t hash_of_addr; sin6 = (struct sockaddr_in6 *)addr; hash_of_addr = (sin6->sin6_addr.s6_addr32[0] + sin6->sin6_addr.s6_addr32[1] + sin6->sin6_addr.s6_addr32[2] + sin6->sin6_addr.s6_addr32[3]); hash_of_addr = (hash_of_addr ^ (hash_of_addr >> 16)); return (hash_of_addr); } return (0); } struct sctp_ifa * sctp_find_ifa_by_addr(struct sockaddr *addr, uint32_t vrf_id, int holds_lock) { struct sctp_ifa *sctp_ifap; struct sctp_vrf *vrf; struct sctp_ifalist *hash_head; uint32_t hash_of_addr; if (holds_lock == 0) SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { stage_right: if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } hash_of_addr = sctp_get_ifa_hash_val(addr); hash_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; if (hash_head == NULL) { SCTP_PRINTF("hash_of_addr:%x mask:%x table:%x - ", hash_of_addr, (uint32_t) vrf->vrf_addr_hashmark, (uint32_t) (hash_of_addr & vrf->vrf_addr_hashmark)); sctp_print_address(addr); SCTP_PRINTF("No such bucket for address\n"); if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } LIST_FOREACH(sctp_ifap, hash_head, next_bucket) { if (sctp_ifap == NULL) { #ifdef INVARIANTS panic("Huh LIST_FOREACH corrupt"); goto stage_right; #else SCTP_PRINTF("LIST corrupt of sctp_ifap's?\n"); goto stage_right; #endif } if (addr->sa_family != sctp_ifap->address.sa.sa_family) continue; if (addr->sa_family == AF_INET) { if (((struct sockaddr_in *)addr)->sin_addr.s_addr == sctp_ifap->address.sin.sin_addr.s_addr) { /* found him. */ if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (sctp_ifap); break; } } #ifdef INET6 if (addr->sa_family == AF_INET6) { if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, &sctp_ifap->address.sin6)) { /* found him. */ if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (sctp_ifap); break; } } #endif } if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } static void sctp_user_rcvd(struct sctp_tcb *stcb, uint32_t * freed_so_far, int hold_rlock, uint32_t rwnd_req) { /* User pulled some data, do we need a rwnd update? */ int r_unlocked = 0; uint32_t dif, rwnd; struct socket *so = NULL; if (stcb == NULL) return; atomic_add_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & (SCTP_STATE_ABOUT_TO_BE_FREED | SCTP_STATE_SHUTDOWN_RECEIVED | SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* Pre-check If we are freeing no update */ goto no_lock; } SCTP_INP_INCR_REF(stcb->sctp_ep); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { goto out; } so = stcb->sctp_socket; if (so == NULL) { goto out; } atomic_add_int(&stcb->freed_by_sorcv_sincelast, *freed_so_far); /* Have you have freed enough to look */ *freed_so_far = 0; /* Yep, its worth a look and the lock overhead */ /* Figure out what the rwnd would be */ rwnd = sctp_calc_rwnd(stcb, &stcb->asoc); if (rwnd >= stcb->asoc.my_last_reported_rwnd) { dif = rwnd - stcb->asoc.my_last_reported_rwnd; } else { dif = 0; } if (dif >= rwnd_req) { if (hold_rlock) { SCTP_INP_READ_UNLOCK(stcb->sctp_ep); r_unlocked = 1; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* * One last check before we allow the guy possibly * to get in. There is a race, where the guy has not * reached the gate. In that case */ goto out; } SCTP_TCB_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* No reports here */ SCTP_TCB_UNLOCK(stcb); goto out; } SCTP_STAT_INCR(sctps_wu_sacks_sent); sctp_send_sack(stcb); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_USR_RCVD, SCTP_SO_LOCKED); /* make sure no timer is running */ sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_6); SCTP_TCB_UNLOCK(stcb); } else { /* Update how much we have pending */ stcb->freed_by_sorcv_sincelast = dif; } out: if (so && r_unlocked && hold_rlock) { SCTP_INP_READ_LOCK(stcb->sctp_ep); } SCTP_INP_DECR_REF(stcb->sctp_ep); no_lock: atomic_add_int(&stcb->asoc.refcnt, -1); return; } int sctp_sorecvmsg(struct socket *so, struct uio *uio, struct mbuf **mp, struct sockaddr *from, int fromlen, int *msg_flags, struct sctp_sndrcvinfo *sinfo, int filling_sinfo) { /* * MSG flags we will look at MSG_DONTWAIT - non-blocking IO. * MSG_PEEK - Look don't touch :-D (only valid with OUT mbuf copy * mp=NULL thus uio is the copy method to userland) MSG_WAITALL - ?? * On the way out we may send out any combination of: * MSG_NOTIFICATION MSG_EOR * */ struct sctp_inpcb *inp = NULL; int my_len = 0; int cp_len = 0, error = 0; struct sctp_queued_to_read *control = NULL, *ctl = NULL, *nxt = NULL; struct mbuf *m = NULL; struct sctp_tcb *stcb = NULL; int wakeup_read_socket = 0; int freecnt_applied = 0; int out_flags = 0, in_flags = 0; int block_allowed = 1; uint32_t freed_so_far = 0; uint32_t copied_so_far = 0; int in_eeor_mode = 0; int no_rcv_needed = 0; uint32_t rwnd_req = 0; int hold_sblock = 0; int hold_rlock = 0; int slen = 0; uint32_t held_length = 0; int sockbuf_lock = 0; if (uio == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if (msg_flags) { in_flags = *msg_flags; if (in_flags & MSG_PEEK) SCTP_STAT_INCR(sctps_read_peeks); } else { in_flags = 0; } slen = uio->uio_resid; /* Pull in and set up our int flags */ if (in_flags & MSG_OOB) { /* Out of band's NOT supported */ return (EOPNOTSUPP); } if ((in_flags & MSG_PEEK) && (mp != NULL)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((in_flags & (MSG_DONTWAIT | MSG_NBIO )) || SCTP_SO_IS_NBIO(so)) { block_allowed = 0; } /* setup the endpoint */ inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EFAULT); return (EFAULT); } rwnd_req = (SCTP_SB_LIMIT_RCV(so) >> SCTP_RWND_HIWAT_SHIFT); /* Must be at least a MTU's worth */ if (rwnd_req < SCTP_MIN_RWND) rwnd_req = SCTP_MIN_RWND; in_eeor_mode = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SORECV_ENTER, rwnd_req, in_eeor_mode, so->so_rcv.sb_cc, uio->uio_resid); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SORECV_ENTERPL, rwnd_req, block_allowed, so->so_rcv.sb_cc, uio->uio_resid); } error = sblock(&so->so_rcv, (block_allowed ? SBL_WAIT : 0)); sockbuf_lock = 1; if (error) { goto release_unlocked; } restart: restart_nosblocks: if (hold_sblock == 0) { SOCKBUF_LOCK(&so->so_rcv); hold_sblock = 1; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { goto out; } if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { if (so->so_error) { error = so->so_error; if ((in_flags & MSG_PEEK) == 0) so->so_error = 0; goto out; } else { if (so->so_rcv.sb_cc == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); /* indicate EOF */ error = 0; goto out; } } } if ((so->so_rcv.sb_cc <= held_length) && block_allowed) { /* we need to wait for data */ if ((so->so_rcv.sb_cc == 0) && ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) { /* * For active open side clear flags for * re-use passive open is blocked by * connect. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED) { /* * You were aborted, passive side * always hits here */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); error = ECONNRESET; /* * You get this once if you are * active open side */ if (!(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * Remove flag if on the * active open side */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_ABORTED; } } so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); if (error == 0) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); error = ENOTCONN; } else { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_CONNECTED; } } goto out; } } error = sbwait(&so->so_rcv); if (error) { goto out; } held_length = 0; goto restart_nosblocks; } else if (so->so_rcv.sb_cc == 0) { if (so->so_error) { error = so->so_error; if ((in_flags & MSG_PEEK) == 0) so->so_error = 0; } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) { /* * For active open side clear flags * for re-use passive open is * blocked by connect. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED) { /* * You were aborted, passive * side always hits here */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); error = ECONNRESET; /* * You get this once if you * are active open side */ if (!(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * Remove flag if on * the active open * side */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_ABORTED; } } so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); if (error == 0) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); error = ENOTCONN; } else { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_CONNECTED; } } goto out; } } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EWOULDBLOCK); error = EWOULDBLOCK; } goto out; } if (hold_sblock == 1) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } /* we possibly have data we can read */ /* sa_ignore FREED_MEMORY */ control = TAILQ_FIRST(&inp->read_queue); if (control == NULL) { /* * This could be happening since the appender did the * increment but as not yet did the tailq insert onto the * read_queue */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } control = TAILQ_FIRST(&inp->read_queue); if ((control == NULL) && (so->so_rcv.sb_cc != 0)) { #ifdef INVARIANTS panic("Huh, its non zero and nothing on control?"); #endif so->so_rcv.sb_cc = 0; } SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; goto restart; } if ((control->length == 0) && (control->do_not_ref_stcb)) { /* * Clean up code for freeing assoc that left behind a * pdapi.. maybe a peer in EEOR that just closed after * sending and never indicated a EOR. */ if (hold_rlock == 0) { hold_rlock = 1; SCTP_INP_READ_LOCK(inp); } control->held_length = 0; if (control->data) { /* Hmm there is data here .. fix */ struct mbuf *m_tmp; int cnt = 0; m_tmp = control->data; while (m_tmp) { cnt += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { control->tail_mbuf = m_tmp; control->end_added = 1; } m_tmp = SCTP_BUF_NEXT(m_tmp); } control->length = cnt; } else { /* remove it */ TAILQ_REMOVE(&inp->read_queue, control, next); /* Add back any hiddend data */ sctp_free_remote_addr(control->whoFrom); sctp_free_a_readq(stcb, control); } if (hold_rlock) { hold_rlock = 0; SCTP_INP_READ_UNLOCK(inp); } goto restart; } if ((control->length == 0) && (control->end_added == 1)) { /* * Do we also need to check for (control->pdapi_aborted == * 1)? */ if (hold_rlock == 0) { hold_rlock = 1; SCTP_INP_READ_LOCK(inp); } TAILQ_REMOVE(&inp->read_queue, control, next); if (control->data) { #ifdef INVARIANTS panic("control->data not null but control->length == 0"); #else SCTP_PRINTF("Strange, data left in the control buffer. Cleaning up.\n"); sctp_m_freem(control->data); control->data = NULL; #endif } if (control->aux_data) { sctp_m_free(control->aux_data); control->aux_data = NULL; } sctp_free_remote_addr(control->whoFrom); sctp_free_a_readq(stcb, control); if (hold_rlock) { hold_rlock = 0; SCTP_INP_READ_UNLOCK(inp); } goto restart; } if (control->length == 0) { if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) && (filling_sinfo)) { /* find a more suitable one then this */ ctl = TAILQ_NEXT(control, next); while (ctl) { if ((ctl->stcb != control->stcb) && (ctl->length) && (ctl->some_taken || (ctl->spec_flags & M_NOTIFICATION) || ((ctl->do_not_ref_stcb == 0) && (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0))) ) { /*- * If we have a different TCB next, and there is data * present. If we have already taken some (pdapi), OR we can * ref the tcb and no delivery as started on this stream, we * take it. Note we allow a notification on a different * assoc to be delivered.. */ control = ctl; goto found_one; } else if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) && (ctl->length) && ((ctl->some_taken) || ((ctl->do_not_ref_stcb == 0) && ((ctl->spec_flags & M_NOTIFICATION) == 0) && (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0)))) { /*- * If we have the same tcb, and there is data present, and we * have the strm interleave feature present. Then if we have * taken some (pdapi) or we can refer to tht tcb AND we have * not started a delivery for this stream, we can take it. * Note we do NOT allow a notificaiton on the same assoc to * be delivered. */ control = ctl; goto found_one; } ctl = TAILQ_NEXT(ctl, next); } } /* * if we reach here, not suitable replacement is available * fragment interleave is NOT on. So stuff the sb_cc * into the our held count, and its time to sleep again. */ held_length = so->so_rcv.sb_cc; control->held_length = so->so_rcv.sb_cc; goto restart; } /* Clear the held length since there is something to read */ control->held_length = 0; if (hold_rlock) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } found_one: /* * If we reach here, control has a some data for us to read off. * Note that stcb COULD be NULL. */ control->some_taken++; if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } stcb = control->stcb; if (stcb) { if ((control->do_not_ref_stcb == 0) && (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED)) { if (freecnt_applied == 0) stcb = NULL; } else if (control->do_not_ref_stcb == 0) { /* you can't free it on me please */ /* * The lock on the socket buffer protects us so the * free code will stop. But since we used the * socketbuf lock and the sender uses the tcb_lock * to increment, we need to use the atomic add to * the refcnt */ if (freecnt_applied) { #ifdef INVARIANTS panic("refcnt already incremented"); #else printf("refcnt already incremented?\n"); #endif } else { atomic_add_int(&stcb->asoc.refcnt, 1); freecnt_applied = 1; } /* * Setup to remember how much we have not yet told * the peer our rwnd has opened up. Note we grab the * value from the tcb from last time. Note too that * sack sending clears this when a sack is sent, * which is fine. Once we hit the rwnd_req, we then * will go to the sctp_user_rcvd() that will not * lock until it KNOWs it MUST send a WUP-SACK. */ freed_so_far = stcb->freed_by_sorcv_sincelast; stcb->freed_by_sorcv_sincelast = 0; } } if (stcb && ((control->spec_flags & M_NOTIFICATION) == 0) && control->do_not_ref_stcb == 0) { stcb->asoc.strmin[control->sinfo_stream].delivery_started = 1; } /* First lets get off the sinfo and sockaddr info */ if ((sinfo) && filling_sinfo) { memcpy(sinfo, control, sizeof(struct sctp_nonpad_sndrcvinfo)); nxt = TAILQ_NEXT(control, next); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) { struct sctp_extrcvinfo *s_extra; s_extra = (struct sctp_extrcvinfo *)sinfo; if ((nxt) && (nxt->length)) { s_extra->sreinfo_next_flags = SCTP_NEXT_MSG_AVAIL; if (nxt->sinfo_flags & SCTP_UNORDERED) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_IS_UNORDERED; } if (nxt->spec_flags & M_NOTIFICATION) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_IS_NOTIFICATION; } s_extra->sreinfo_next_aid = nxt->sinfo_assoc_id; s_extra->sreinfo_next_length = nxt->length; s_extra->sreinfo_next_ppid = nxt->sinfo_ppid; s_extra->sreinfo_next_stream = nxt->sinfo_stream; if (nxt->tail_mbuf != NULL) { if (nxt->end_added) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_ISCOMPLETE; } } } else { /* * we explicitly 0 this, since the memcpy * got some other things beyond the older * sinfo_ that is on the control's structure * :-D */ nxt = NULL; s_extra->sreinfo_next_flags = SCTP_NO_NEXT_MSG; s_extra->sreinfo_next_aid = 0; s_extra->sreinfo_next_length = 0; s_extra->sreinfo_next_ppid = 0; s_extra->sreinfo_next_stream = 0; } } /* * update off the real current cum-ack, if we have an stcb. */ if ((control->do_not_ref_stcb == 0) && stcb) sinfo->sinfo_cumtsn = stcb->asoc.cumulative_tsn; /* * mask off the high bits, we keep the actual chunk bits in * there. */ sinfo->sinfo_flags &= 0x00ff; if ((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED) { sinfo->sinfo_flags |= SCTP_UNORDERED; } } #ifdef SCTP_ASOCLOG_OF_TSNS { int index, newindex; struct sctp_pcbtsn_rlog *entry; do { index = inp->readlog_index; newindex = index + 1; if (newindex >= SCTP_READ_LOG_SIZE) { newindex = 0; } } while (atomic_cmpset_int(&inp->readlog_index, index, newindex) == 0); entry = &inp->readlog[index]; entry->vtag = control->sinfo_assoc_id; entry->strm = control->sinfo_stream; entry->seq = control->sinfo_ssn; entry->sz = control->length; entry->flgs = control->sinfo_flags; } #endif if (fromlen && from) { struct sockaddr *to; #ifdef INET cp_len = min((size_t)fromlen, (size_t)control->whoFrom->ro._l_addr.sin.sin_len); memcpy(from, &control->whoFrom->ro._l_addr, cp_len); ((struct sockaddr_in *)from)->sin_port = control->port_from; #else /* No AF_INET use AF_INET6 */ cp_len = min((size_t)fromlen, (size_t)control->whoFrom->ro._l_addr.sin6.sin6_len); memcpy(from, &control->whoFrom->ro._l_addr, cp_len); ((struct sockaddr_in6 *)from)->sin6_port = control->port_from; #endif to = from; #if defined(INET) && defined(INET6) if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) && (to->sa_family == AF_INET) && ((size_t)fromlen >= sizeof(struct sockaddr_in6))) { struct sockaddr_in *sin; struct sockaddr_in6 sin6; sin = (struct sockaddr_in *)to; bzero(&sin6, sizeof(sin6)); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_addr.s6_addr32[2] = htonl(0xffff); bcopy(&sin->sin_addr, &sin6.sin6_addr.s6_addr32[3], sizeof(sin6.sin6_addr.s6_addr32[3])); sin6.sin6_port = sin->sin_port; memcpy(from, (caddr_t)&sin6, sizeof(sin6)); } #endif #if defined(INET6) { struct sockaddr_in6 lsa6, *to6; to6 = (struct sockaddr_in6 *)to; sctp_recover_scope_mac(to6, (&lsa6)); } #endif } /* now copy out what data we can */ if (mp == NULL) { /* copy out each mbuf in the chain up to length */ get_more_data: m = control->data; while (m) { /* Move out all we can */ cp_len = (int)uio->uio_resid; my_len = (int)SCTP_BUF_LEN(m); if (cp_len > my_len) { /* not enough in this buf */ cp_len = my_len; } if (hold_rlock) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (cp_len > 0) error = uiomove(mtod(m, char *), cp_len, uio); /* re-read */ if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { goto release; } if ((control->do_not_ref_stcb == 0) && stcb && stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { no_rcv_needed = 1; } if (error) { /* error we are out of here */ goto release; } if ((SCTP_BUF_NEXT(m) == NULL) && (cp_len >= SCTP_BUF_LEN(m)) && ((control->end_added == 0) || (control->end_added && (TAILQ_NEXT(control, next) == NULL))) ) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } if (cp_len == SCTP_BUF_LEN(m)) { if ((SCTP_BUF_NEXT(m) == NULL) && (control->end_added)) { out_flags |= MSG_EOR; if ((control->do_not_ref_stcb == 0) && (control->stcb != NULL) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } /* we ate up the mbuf */ if (in_flags & MSG_PEEK) { /* just looking */ m = SCTP_BUF_NEXT(m); copied_so_far += cp_len; } else { /* dispose of the mbuf */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } copied_so_far += cp_len; freed_so_far += cp_len; freed_so_far += MSIZE; atomic_subtract_int(&control->length, cp_len); control->data = sctp_m_free(m); m = control->data; /* * been through it all, must hold sb * lock ok to null tail */ if (control->data == NULL) { #ifdef INVARIANTS if ((control->end_added == 0) || (TAILQ_NEXT(control, next) == NULL)) { /* * If the end is not * added, OR the * next is NOT null * we MUST have the * lock. */ if (mtx_owned(&inp->inp_rdata_mtx) == 0) { panic("Hmm we don't own the lock?"); } } #endif control->tail_mbuf = NULL; #ifdef INVARIANTS if ((control->end_added) && ((out_flags & MSG_EOR) == 0)) { panic("end_added, nothing left and no MSG_EOR"); } #endif } } } else { /* Do we need to trim the mbuf? */ if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } if ((in_flags & MSG_PEEK) == 0) { SCTP_BUF_RESV_UF(m, cp_len); SCTP_BUF_LEN(m) -= cp_len; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, cp_len); } atomic_subtract_int(&so->so_rcv.sb_cc, cp_len); if ((control->do_not_ref_stcb == 0) && stcb) { atomic_subtract_int(&stcb->asoc.sb_cc, cp_len); } copied_so_far += cp_len; freed_so_far += cp_len; freed_so_far += MSIZE; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_subtract_int(&control->length, cp_len); } else { copied_so_far += cp_len; } } if ((out_flags & MSG_EOR) || (uio->uio_resid == 0)) { break; } if (((stcb) && (in_flags & MSG_PEEK) == 0) && (control->do_not_ref_stcb == 0) && (freed_so_far >= rwnd_req)) { sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } } /* end while(m) */ /* * At this point we have looked at it all and we either have * a MSG_EOR/or read all the user wants... * control->length == 0. */ if ((out_flags & MSG_EOR) && ((in_flags & MSG_PEEK) == 0)) { /* we are done with this control */ if (control->length == 0) { if (control->data) { #ifdef INVARIANTS panic("control->data not null at read eor?"); #else SCTP_PRINTF("Strange, data left in the control buffer .. invarients would panic?\n"); sctp_m_freem(control->data); control->data = NULL; #endif } done_with_control: if (TAILQ_NEXT(control, next) == NULL) { /* * If we don't have a next we need a * lock, if there is a next * interrupt is filling ahead of us * and we don't need a lock to * remove this guy (which is the * head of the queue). */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } } TAILQ_REMOVE(&inp->read_queue, control, next); /* Add back any hiddend data */ if (control->held_length) { held_length = 0; control->held_length = 0; wakeup_read_socket = 1; } if (control->aux_data) { sctp_m_free(control->aux_data); control->aux_data = NULL; } no_rcv_needed = control->do_not_ref_stcb; sctp_free_remote_addr(control->whoFrom); control->data = NULL; sctp_free_a_readq(stcb, control); control = NULL; if ((freed_so_far >= rwnd_req) && (no_rcv_needed == 0)) sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } else { /* * The user did not read all of this * message, turn off the returned MSG_EOR * since we are leaving more behind on the * control to read. */ #ifdef INVARIANTS if (control->end_added && (control->data == NULL) && (control->tail_mbuf == NULL)) { panic("Gak, control->length is corrupt?"); } #endif no_rcv_needed = control->do_not_ref_stcb; out_flags &= ~MSG_EOR; } } if (out_flags & MSG_EOR) { goto release; } if ((uio->uio_resid == 0) || ((in_eeor_mode) && (copied_so_far >= max(so->so_rcv.sb_lowat, 1))) ) { goto release; } /* * If I hit here the receiver wants more and this message is * NOT done (pd-api). So two questions. Can we block? if not * we are done. Did the user NOT set MSG_WAITALL? */ if (block_allowed == 0) { goto release; } /* * We need to wait for more data a few things: - We don't * sbunlock() so we don't get someone else reading. - We * must be sure to account for the case where what is added * is NOT to our control when we wakeup. */ /* * Do we need to tell the transport a rwnd update might be * needed before we go to sleep? */ if (((stcb) && (in_flags & MSG_PEEK) == 0) && ((freed_so_far >= rwnd_req) && (control->do_not_ref_stcb == 0) && (no_rcv_needed == 0))) { sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } wait_some_more: if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { goto release; } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) goto release; if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock == 0) { SOCKBUF_LOCK(&so->so_rcv); hold_sblock = 1; } if ((copied_so_far) && (control->length == 0) && (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE))) { goto release; } if (so->so_rcv.sb_cc <= control->held_length) { error = sbwait(&so->so_rcv); if (error) { goto release; } control->held_length = 0; } if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if (control->length == 0) { /* still nothing here */ if (control->end_added == 1) { /* he aborted, or is done i.e.did a shutdown */ out_flags |= MSG_EOR; if (control->pdapi_aborted) { if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; out_flags |= MSG_TRUNC; } else { if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } goto done_with_control; } if (so->so_rcv.sb_cc > held_length) { control->held_length = so->so_rcv.sb_cc; held_length = 0; } goto wait_some_more; } else if (control->data == NULL) { /* * we must re-sync since data is probably being * added */ SCTP_INP_READ_LOCK(inp); if ((control->length > 0) && (control->data == NULL)) { /* * big trouble.. we have the lock and its * corrupt? */ #ifdef INVARIANTS panic("Impossible data==NULL length !=0"); #endif out_flags |= MSG_EOR; out_flags |= MSG_TRUNC; control->length = 0; SCTP_INP_READ_UNLOCK(inp); goto done_with_control; } SCTP_INP_READ_UNLOCK(inp); /* We will fall around to get more data */ } goto get_more_data; } else { /*- * Give caller back the mbuf chain, * store in uio_resid the length */ wakeup_read_socket = 0; if ((control->end_added == 0) || (TAILQ_NEXT(control, next) == NULL)) { /* Need to get rlock */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } } if (control->end_added) { out_flags |= MSG_EOR; if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } uio->uio_resid = control->length; *mp = control->data; m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &so->so_rcv, m); freed_so_far += SCTP_BUF_LEN(m); freed_so_far += MSIZE; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } control->data = control->tail_mbuf = NULL; control->length = 0; if (out_flags & MSG_EOR) { /* Done with this control */ goto done_with_control; } } release: if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock == 1) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } sbunlock(&so->so_rcv); sockbuf_lock = 0; release_unlocked: if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if ((stcb) && (in_flags & MSG_PEEK) == 0) { if ((freed_so_far >= rwnd_req) && (control && (control->do_not_ref_stcb == 0)) && (no_rcv_needed == 0)) sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } out: if (msg_flags) { *msg_flags = out_flags; } if (((out_flags & MSG_EOR) == 0) && ((in_flags & MSG_PEEK) == 0) && (sinfo) && (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO))) { struct sctp_extrcvinfo *s_extra; s_extra = (struct sctp_extrcvinfo *)sinfo; s_extra->sreinfo_next_flags = SCTP_NO_NEXT_MSG; } if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if (sockbuf_lock) { sbunlock(&so->so_rcv); } if (freecnt_applied) { /* * The lock on the socket buffer protects us so the free * code will stop. But since we used the socketbuf lock and * the sender uses the tcb_lock to increment, we need to use * the atomic add to the refcnt. */ if (stcb == NULL) { #ifdef INVARIANTS panic("stcb for refcnt has gone NULL?"); goto stage_left; #else goto stage_left; #endif } atomic_add_int(&stcb->asoc.refcnt, -1); freecnt_applied = 0; /* Save the value back for next time */ stcb->freed_by_sorcv_sincelast = freed_so_far; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { if (stcb) { sctp_misc_ints(SCTP_SORECV_DONE, freed_so_far, ((uio) ? (slen - uio->uio_resid) : slen), stcb->asoc.my_rwnd, so->so_rcv.sb_cc); } else { sctp_misc_ints(SCTP_SORECV_DONE, freed_so_far, ((uio) ? (slen - uio->uio_resid) : slen), 0, so->so_rcv.sb_cc); } } stage_left: if (wakeup_read_socket) { sctp_sorwakeup(inp, so); } return (error); } #ifdef SCTP_MBUF_LOGGING struct mbuf * sctp_m_free(struct mbuf *m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { if (SCTP_BUF_IS_EXTENDED(m)) { sctp_log_mb(m, SCTP_MBUF_IFREE); } } return (m_free(m)); } void sctp_m_freem(struct mbuf *mb) { while (mb != NULL) mb = sctp_m_free(mb); } #endif int sctp_dynamic_set_primary(struct sockaddr *sa, uint32_t vrf_id) { /* * Given a local address. For all associations that holds the * address, request a peer-set-primary. */ struct sctp_ifa *ifa; struct sctp_laddr *wi; ifa = sctp_find_ifa_by_addr(sa, vrf_id, 0); if (ifa == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EADDRNOTAVAIL); return (EADDRNOTAVAIL); } /* * Now that we have the ifa we must awaken the iterator with this * message. */ wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } /* Now incr the count and int wi structure */ SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = ifa; wi->action = SCTP_SET_PRIM_ADDR; atomic_add_int(&ifa->refcount, 1); /* Now add it to the work queue */ SCTP_WQ_ADDR_LOCK(); /* * Should this really be a tailq? As it is we will process the * newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_WQ_ADDR_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); return (0); } int sctp_soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { int error, fromlen; uint8_t sockbuf[256]; struct sockaddr *from; struct sctp_extrcvinfo sinfo; int filling_sinfo = 1; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; /* pickup the assoc we are reading from */ if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) || (controlp == NULL)) { /* user does not want the sndrcv ctl */ filling_sinfo = 0; } if (psa) { from = (struct sockaddr *)sockbuf; fromlen = sizeof(sockbuf); from->sa_len = 0; } else { from = NULL; fromlen = 0; } error = sctp_sorecvmsg(so, uio, mp0, from, fromlen, flagsp, (struct sctp_sndrcvinfo *)&sinfo, filling_sinfo); if ((controlp) && (filling_sinfo)) { /* copy back the sinfo in a CMSG format */ if (filling_sinfo) *controlp = sctp_build_ctl_nchunk(inp, (struct sctp_sndrcvinfo *)&sinfo); else *controlp = NULL; } if (psa) { /* copy back the address info */ if (from && from->sa_len) { *psa = sodupsockaddr(from, M_NOWAIT); } else { *psa = NULL; } } return (error); } int sctp_l_soreceive(struct socket *so, struct sockaddr **name, struct uio *uio, char **controlp, int *controllen, int *flag) { int error, fromlen; uint8_t sockbuf[256]; struct sockaddr *from; struct sctp_extrcvinfo sinfo; int filling_sinfo = 1; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; /* pickup the assoc we are reading from */ if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) || (controlp == NULL)) { /* user does not want the sndrcv ctl */ filling_sinfo = 0; } if (name) { from = (struct sockaddr *)sockbuf; fromlen = sizeof(sockbuf); from->sa_len = 0; } else { from = NULL; fromlen = 0; } error = sctp_sorecvmsg(so, uio, (struct mbuf **)NULL, from, fromlen, flag, (struct sctp_sndrcvinfo *)&sinfo, filling_sinfo); if ((controlp) && (filling_sinfo)) { /* * copy back the sinfo in a CMSG format note that the caller * has reponsibility for freeing the memory. */ if (filling_sinfo) *controlp = sctp_build_ctl_cchunk(inp, controllen, (struct sctp_sndrcvinfo *)&sinfo); } if (name) { /* copy back the address info */ if (from && from->sa_len) { *name = sodupsockaddr(from, M_WAIT); } else { *name = NULL; } } return (error); } int sctp_connectx_helper_add(struct sctp_tcb *stcb, struct sockaddr *addr, int totaddr, int *error) { int added = 0; int i; struct sctp_inpcb *inp; struct sockaddr *sa; size_t incr = 0; sa = addr; inp = stcb->sctp_ep; *error = 0; for (i = 0; i < totaddr; i++) { if (sa->sa_family == AF_INET) { incr = sizeof(struct sockaddr_in); if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { /* assoc gone no un-lock */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_7); *error = ENOBUFS; goto out_now; } added++; } else if (sa->sa_family == AF_INET6) { incr = sizeof(struct sockaddr_in6); if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { /* assoc gone no un-lock */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_8); *error = ENOBUFS; goto out_now; } added++; } sa = (struct sockaddr *)((caddr_t)sa + incr); } out_now: return (added); } struct sctp_tcb * sctp_connectx_helper_find(struct sctp_inpcb *inp, struct sockaddr *addr, int *totaddr, int *num_v4, int *num_v6, int *error, int limit, int *bad_addr) { struct sockaddr *sa; struct sctp_tcb *stcb = NULL; size_t incr, at, i; at = incr = 0; sa = addr; *error = *num_v6 = *num_v4 = 0; /* account and validate addresses */ for (i = 0; i < (size_t)*totaddr; i++) { if (sa->sa_family == AF_INET) { (*num_v4) += 1; incr = sizeof(struct sockaddr_in); if (sa->sa_len != incr) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } } else if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { /* Must be non-mapped for connectx */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } (*num_v6) += 1; incr = sizeof(struct sockaddr_in6); if (sa->sa_len != incr) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } } else { *totaddr = i; /* we are done */ break; } SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, sa, NULL, NULL, NULL); if (stcb != NULL) { /* Already have or am bring up an association */ return (stcb); } else { SCTP_INP_DECR_REF(inp); } if ((at + incr) > (size_t)limit) { *totaddr = i; break; } sa = (struct sockaddr *)((caddr_t)sa + incr); } return ((struct sctp_tcb *)NULL); } /* * sctp_bindx(ADD) for one address. * assumes all arguments are valid/checked by caller. */ void sctp_bindx_add_address(struct socket *so, struct sctp_inpcb *inp, struct sockaddr *sa, sctp_assoc_t assoc_id, uint32_t vrf_id, int *error, void *p) { struct sockaddr *addr_touse; #ifdef INET6 struct sockaddr_in sin; #endif /* see if we're bound all already! */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } addr_touse = sa; #if defined(INET6) && !defined(__Userspace__) /* TODO port in6_sin6_2_sin */ if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; if (sa->sa_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { /* can only bind v6 on PF_INET6 sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } sin6 = (struct sockaddr_in6 *)addr_touse; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4-mapped on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } in6_sin6_2_sin(&sin, sin6); addr_touse = (struct sockaddr *)&sin; } } #endif if (sa->sa_family == AF_INET) { if (sa->sa_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } } if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { if (p == NULL) { /* Can't get proc for Net/Open BSD */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } *error = sctp_inpcb_bind(so, addr_touse, NULL, p); return; } /* * No locks required here since bind and mgmt_ep_sa all do their own * locking. If we do something for the FIX: below we may need to * lock in that case. */ if (assoc_id == 0) { /* add the address */ struct sctp_inpcb *lep; struct sockaddr_in *lsin = (struct sockaddr_in *)addr_touse; /* validate the incoming port */ if ((lsin->sin_port != 0) && (lsin->sin_port != inp->sctp_lport)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } else { /* user specified 0 port, set it to existing port */ lsin->sin_port = inp->sctp_lport; } lep = sctp_pcb_findep(addr_touse, 1, 0, vrf_id); if (lep != NULL) { /* * We must decrement the refcount since we have the * ep already and are binding. No remove going on * here. */ SCTP_INP_DECR_REF(lep); } if (lep == inp) { /* already bound to it.. ok */ return; } else if (lep == NULL) { ((struct sockaddr_in *)addr_touse)->sin_port = 0; *error = sctp_addr_mgmt_ep_sa(inp, addr_touse, SCTP_ADD_IP_ADDRESS, vrf_id, NULL); } else { *error = EADDRINUSE; } if (*error) return; } else { /* * FIX: decide whether we allow assoc based bindx */ } } /* * sctp_bindx(DELETE) for one address. * assumes all arguments are valid/checked by caller. */ void sctp_bindx_delete_address(struct socket *so, struct sctp_inpcb *inp, struct sockaddr *sa, sctp_assoc_t assoc_id, uint32_t vrf_id, int *error) { struct sockaddr *addr_touse; #ifdef INET6 struct sockaddr_in sin; #endif /* see if we're bound all already! */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } addr_touse = sa; #if defined(INET6) && !defined(__Userspace__) /* TODO port in6_sin6_2_sin */ if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; if (sa->sa_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { /* can only bind v6 on PF_INET6 sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } sin6 = (struct sockaddr_in6 *)addr_touse; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind mapped-v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } in6_sin6_2_sin(&sin, sin6); addr_touse = (struct sockaddr *)&sin; } } #endif if (sa->sa_family == AF_INET) { if (sa->sa_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } } /* * No lock required mgmt_ep_sa does its own locking. If the FIX: * below is ever changed we may need to lock before calling * association level binding. */ if (assoc_id == 0) { /* delete the address */ *error = sctp_addr_mgmt_ep_sa(inp, addr_touse, SCTP_DEL_IP_ADDRESS, vrf_id, NULL); } else { /* * FIX: decide whether we allow assoc based bindx */ } } /* * returns the valid local address count for an assoc, taking into account * all scoping rules */ int sctp_local_addr_count(struct sctp_tcb *stcb) { int loopback_scope, ipv4_local_scope, local_scope, site_scope; int ipv4_addr_legal, ipv6_addr_legal; struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; int count = 0; /* Turn on all the appropriate scopes */ loopback_scope = stcb->asoc.loopback_scope; ipv4_local_scope = stcb->asoc.ipv4_local_scope; local_scope = stcb->asoc.local_scope; site_scope = stcb->asoc.site_scope; ipv4_addr_legal = ipv6_addr_legal = 0; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ipv6_addr_legal = 1; if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) { ipv4_addr_legal = 1; } } else { ipv4_addr_legal = 1; } SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(stcb->asoc.vrf_id); if (vrf == NULL) { /* no vrf, no addresses */ SCTP_IPI_ADDR_RUNLOCK(); return (0); } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* * bound all case: go through all ifns on the vrf */ LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (sctp_is_addr_restricted(stcb, sctp_ifa)) continue; switch (sctp_ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifa->address.sa; if (sin->sin_addr.s_addr == 0) { /* * skip unspecified * addrs */ continue; } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { continue; } /* count this one */ count++; } else { continue; } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifa->address.sa; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { continue; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) /* * * bad * * li * nk * * loc * al * * add * re * ss * */ continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } /* count this one */ count++; } break; #endif default: /* TSNH */ break; } } } } else { /* * subset bound case */ struct sctp_laddr *laddr; LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { if (sctp_is_addr_restricted(stcb, laddr->ifa)) { continue; } /* count this one */ count++; } } SCTP_IPI_ADDR_RUNLOCK(); return (count); } #if defined(SCTP_LOCAL_TRACE_BUF) void sctp_log_trace(uint32_t subsys, const char *str SCTP_UNUSED, uint32_t a, uint32_t b, uint32_t c, uint32_t d, uint32_t e, uint32_t f) { uint32_t saveindex, newindex; do { saveindex = SCTP_BASE_SYSCTL(sctp_log).index; if (saveindex >= SCTP_MAX_LOGGING_SIZE) { newindex = 1; } else { newindex = saveindex + 1; } } while (atomic_cmpset_int(&SCTP_BASE_SYSCTL(sctp_log).index, saveindex, newindex) == 0); if (saveindex >= SCTP_MAX_LOGGING_SIZE) { saveindex = 0; } SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].timestamp = SCTP_GET_CYCLECOUNT; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].subsys = subsys; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[0] = a; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[1] = b; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[2] = c; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[3] = d; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[4] = e; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[5] = f; } #endif /* We will need to add support * to bind the ports and such here * so we can do UDP tunneling. In * the mean-time, we return error */ #include #include #include #ifdef INET6 #include #endif static void sctp_recv_udp_tunneled_packet(struct mbuf *m, int off, struct inpcb *ignored) { struct ip *iph; struct mbuf *sp, *last; struct udphdr *uhdr; uint16_t port = 0, len; int header_size = sizeof(struct udphdr) + sizeof(struct sctphdr); /* * Split out the mbuf chain. Leave the IP header in m, place the * rest in the sp. */ if ((m->m_flags & M_PKTHDR) == 0) { /* Can't handle one that is not a pkt hdr */ goto out; } /* pull the src port */ iph = mtod(m, struct ip *); uhdr = (struct udphdr *)((caddr_t)iph + off); port = uhdr->uh_sport; sp = m_split(m, off, M_DONTWAIT); if (sp == NULL) { /* Gak, drop packet, we can't do a split */ goto out; } if (sp->m_pkthdr.len < header_size) { /* Gak, packet can't have an SCTP header in it - to small */ m_freem(sp); goto out; } /* ok now pull up the UDP header and SCTP header together */ sp = m_pullup(sp, header_size); if (sp == NULL) { /* Gak pullup failed */ goto out; } /* trim out the UDP header */ m_adj(sp, sizeof(struct udphdr)); /* Now reconstruct the mbuf chain */ /* 1) find last one */ last = m; while (last->m_next != NULL) { last = last->m_next; } last->m_next = sp; m->m_pkthdr.len += sp->m_pkthdr.len; last = m; while (last != NULL) { last = last->m_next; } /* Now its ready for sctp_input or sctp6_input */ iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ len = SCTP_GET_IPV4_LENGTH(iph); len -= sizeof(struct udphdr); SCTP_GET_IPV4_LENGTH(iph) = len; sctp_input_with_port(m, off, port); break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 - NOT supported */ goto out; break; } #endif default: { m_freem(m); break; } } return; out: m_freem(m); } void sctp_over_udp_stop(void) { struct socket *sop; /* * This function assumes sysctl caller holds sctp_sysctl_info_lock() * for writting! */ if (SCTP_BASE_INFO(udp_tun_socket) == NULL) { /* Nothing to do */ return; } sop = SCTP_BASE_INFO(udp_tun_socket); soclose(sop); SCTP_BASE_INFO(udp_tun_socket) = NULL; } int sctp_over_udp_start(void) { uint16_t port; int ret; struct sockaddr_in sin; struct socket *sop = NULL; struct thread *th; struct ucred *cred; /* * This function assumes sysctl caller holds sctp_sysctl_info_lock() * for writting! */ port = SCTP_BASE_SYSCTL(sctp_udp_tunneling_port); if (port == 0) { /* Must have a port set */ return (EINVAL); } if (SCTP_BASE_INFO(udp_tun_socket) != NULL) { /* Already running -- must stop first */ return (EALREADY); } th = curthread; cred = th->td_ucred; if ((ret = socreate(PF_INET, &sop, SOCK_DGRAM, IPPROTO_UDP, cred, th))) { return (ret); } SCTP_BASE_INFO(udp_tun_socket) = sop; /* call the special UDP hook */ ret = udp_set_kernel_tunneling(sop, sctp_recv_udp_tunneled_packet); if (ret) { goto exit_stage_left; } /* Ok we have a socket, bind it to the port */ memset(&sin, 0, sizeof(sin)); sin.sin_len = sizeof(sin); sin.sin_family = AF_INET; sin.sin_port = htons(port); ret = sobind(sop, (struct sockaddr *)&sin, th); if (ret) { /* Close up we cant get the port */ exit_stage_left: sctp_over_udp_stop(); return (ret); } /* * Ok we should now get UDP packets directly to our input routine * sctp_recv_upd_tunneled_packet(). */ return (0); }