diff --git a/sys/netinet/sctp_indata.c b/sys/netinet/sctp_indata.c index 055b12b68dcf..53dcf8f2a6f8 100644 --- a/sys/netinet/sctp_indata.c +++ b/sys/netinet/sctp_indata.c @@ -1,5838 +1,5837 @@ /*- * 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); calc = sctp_sbspace_sub(calc, (uint32_t) asoc->size_on_all_streams); 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; int fnd = 0; if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { 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 (compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { 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; (compare_with_wrap(i, asoc->mapping_array_base_tsn, MAX_TSN) || (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; uint16_t nxt_todel; uint16_t stream_no; int end = 0; int cntDel; struct sctp_queued_to_read *control, *ctl, *ctlat; 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; chk = TAILQ_FIRST(&asoc->reasmqueue); while (chk) { 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 */ chk = TAILQ_FIRST(&asoc->reasmqueue); } return; } SCTP_TCB_LOCK_ASSERT(stcb); do { chk = TAILQ_FIRST(&asoc->reasmqueue); if (chk == NULL) { return; } 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; ctl = TAILQ_FIRST(&strm->inqueue); if (ctl && (nxt_todel == ctl->sinfo_ssn)) { while (ctl != NULL) { /* Deliver more if we can. */ if (nxt_todel == ctl->sinfo_ssn) { ctlat = TAILQ_NEXT(ctl, next); 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); ctl = ctlat; } else { break; } nxt_todel = strm->last_sequence_delivered + 1; } } break; } /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->reasmqueue); } while (chk); } /* * 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 (compare_with_wrap(strm->last_sequence_delivered, control->sinfo_ssn, MAX_SEQ) || (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); control = TAILQ_FIRST(&strm->inqueue); while (control != NULL) { /* all delivered */ nxt_todel = strm->last_sequence_delivered + 1; if (nxt_todel == control->sinfo_ssn) { at = TAILQ_NEXT(control, next); 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); control = at; continue; } break; } } if (queue_needed) { /* * Ok, we did not deliver this guy, find the correct place * to put it on the queue. */ if ((compare_with_wrap(asoc->cumulative_tsn, control->sinfo_tsn, MAX_TSN)) || (control->sinfo_tsn == asoc->cumulative_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 (compare_with_wrap(at->sinfo_ssn, control->sinfo_ssn, MAX_SEQ)) { /* * 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; while (chk) { 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++; chk = TAILQ_NEXT(chk, sctp_next); } 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 (compare_with_wrap(at->rec.data.TSN_seq, chk->rec.data.TSN_seq, MAX_TSN)) { /* * 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 (compare_with_wrap(TSN_seq, at->rec.data.TSN_seq, MAX_TSN)) { /* 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 = ((ch->ch.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 (compare_with_wrap(asoc->cumulative_tsn, tsn, MAX_TSN) || 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 (compare_with_wrap(tsn, *high_tsn, MAX_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 (compare_with_wrap(tsn, asoc->highest_tsn_inside_map, MAX_TSN) && compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { /* 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 (compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { 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 && (compare_with_wrap(asoc->strmin[strmno].last_sequence_delivered, strmseq, MAX_SEQ) || 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) || ((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_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; SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); if (compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { asoc->highest_tsn_inside_nr_map = tsn; } 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 (compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { 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 (compare_with_wrap(tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { asoc->highest_tsn_inside_nr_map = tsn; } } else { SCTP_SET_TSN_PRESENT(asoc->mapping_array, gap); if (compare_with_wrap(tsn, asoc->highest_tsn_inside_map, MAX_TSN)) { 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) && ((compare_with_wrap(tsn, liste->tsn, MAX_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; unsigned char inserted = 0; ctlOn = TAILQ_FIRST(&asoc->pending_reply_queue); while (ctlOn) { if (compare_with_wrap(control->sinfo_tsn, ctlOn->sinfo_tsn, MAX_TSN)) { ctlOn = TAILQ_NEXT(ctlOn, next); } 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 (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) && ((compare_with_wrap(asoc->cumulative_tsn, liste->tsn, MAX_TSN)) || (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; 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); ctl = TAILQ_FIRST(&asoc->pending_reply_queue); if (ctl && (liste == NULL)) { /* All can be removed */ while (ctl) { TAILQ_REMOVE(&asoc->pending_reply_queue, ctl, next); sctp_queue_data_to_stream(stcb, asoc, ctl, abort_flag); if (*abort_flag) { return (0); } ctl = TAILQ_FIRST(&asoc->pending_reply_queue); } } else if (ctl) { /* more than one in queue */ while (!compare_with_wrap(ctl->sinfo_tsn, liste->tsn, MAX_TSN)) { /* * 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); } ctl = TAILQ_FIRST(&asoc->pending_reply_queue); } } /* * 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? */ struct sctp_association *asoc; int at; int slide_from, slide_end, lgap, distance; /* EY nr_mapping array variables */ /* int nr_at; */ /* int nr_last_all_ones = 0; */ /* int nr_slide_from, nr_slide_end, nr_lgap, nr_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++) { if (asoc->nr_mapping_array[slide_from] == 0xff) { at += 8; } else { /* there is a 0 bit */ at += sctp_map_lookup_tab[asoc->nr_mapping_array[slide_from]]; break; } } asoc->cumulative_tsn = asoc->mapping_array_base_tsn + (at - 1); if (compare_with_wrap(asoc->cumulative_tsn, asoc->highest_tsn_inside_map, MAX_TSN) && - compare_with_wrap(asoc->cumulative_tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN) - ) { + compare_with_wrap(asoc->cumulative_tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { #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 (compare_with_wrap(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map, MAX_TSN)) { 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, i; /* 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); 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); } } 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++) { + for (ii = distance; ii < asoc->mapping_array_size; ii++) { asoc->mapping_array[ii] = 0; asoc->nr_mapping_array[ii] = 0; } 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 (compare_with_wrap(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map, MAX_TSN)) { 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 = compare_with_wrap(highest_tsn, stcb->asoc.cumulative_tsn, MAX_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 ((SCTP_BASE_SYSCTL(sctp_cmt_on_off)) && (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 = 0; struct mbuf *m; /* set the rwnd */ sctp_set_rwnd(stcb, &stcb->asoc); m = *mm; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; if (compare_with_wrap(stcb->asoc.highest_tsn_inside_map, stcb->asoc.cumulative_tsn, MAX_TSN)) { /* there was a gap before this data was processed */ was_a_gap = 1; } /* * 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 */ if ((stcb->asoc.cumulative_tsn == stcb->asoc.highest_tsn_inside_map) && (stcb->asoc.mapping_array[0] != 0xff)) { if ((stcb->asoc.data_pkts_seen >= stcb->asoc.sack_freq) || (stcb->asoc.delayed_ack == 0) || (stcb->asoc.numduptsns) || (stcb->asoc.send_sack == 1)) { if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { (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); } } } else { 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 (compare_with_wrap(tp1->rec.data.TSN_seq, *biggest_newly_acked_tsn, MAX_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 (compare_with_wrap(tp1->rec.data.TSN_seq, tp1->whoTo->this_sack_highest_newack, MAX_TSN)) { 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 (compare_with_wrap(tp1->rec.data.TSN_seq, stcb->asoc.this_sack_highest_gap, MAX_TSN)) { 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 (compare_with_wrap(tp1->rec.data.TSN_seq, theTSN, MAX_TSN)) 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; uint32_t last_frag_high; tp1 = NULL; last_frag_high = 0; chunk_freed = 0; for (i = 0; i < (num_seg + num_nr_seg); i++) { 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); /* some sanity checks on the fragment offsets */ if (frag_strt > frag_end) { /* this one is malformed, skip */ continue; } if (compare_with_wrap((frag_end + last_tsn), *biggest_tsn_acked, MAX_TSN)) *biggest_tsn_acked = frag_end + last_tsn; /* mark acked dgs and find out the highestTSN being acked */ if (tp1 == NULL) { tp1 = TAILQ_FIRST(&asoc->sent_queue); /* save the locations of the last frags */ last_frag_high = frag_end + last_tsn; } else { /* * now lets see if we need to reset the queue due to * a out-of-order SACK fragment */ if (compare_with_wrap(frag_strt + last_tsn, last_frag_high, MAX_TSN)) { /* * if the new frag starts after the last TSN * frag covered, we are ok and this one is * beyond the last one */ ; } else { /* * ok, they have reset us, so we need to * reset the queue this will cause extra * hunting but hey, they chose the * performance hit when they failed to order * their gaps */ tp1 = TAILQ_FIRST(&asoc->sent_queue); } last_frag_high = frag_end + last_tsn; } if (i < num_seg) { non_revocable = 0; } else { non_revocable = 1; } if (i == num_seg) { tp1 = NULL; } 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; } } 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; tp1 = TAILQ_FIRST(&asoc->sent_queue); while (tp1) { if (compare_with_wrap(tp1->rec.data.TSN_seq, cumack, MAX_TSN)) { /* * 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 (compare_with_wrap(tp1->rec.data.TSN_seq, biggest_tsn_acked, MAX_TSN)) 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; tp1 = TAILQ_NEXT(tp1, sctp_next); } 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && 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); } tp1 = TAILQ_FIRST(&asoc->sent_queue); while (tp1) { strike_flag = 0; if (tp1->no_fr_allowed) { /* this one had a timeout or something */ tp1 = TAILQ_NEXT(tp1, sctp_next); 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 (compare_with_wrap(tp1->rec.data.TSN_seq, biggest_tsn_acked, MAX_TSN) || 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 ( /* * TODO sctp_constants.h needs alternative * time macros when _KERNEL is undefined. */ (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); } tp1 = TAILQ_NEXT(tp1, sctp_next); continue; } } } if (compare_with_wrap(tp1->rec.data.TSN_seq, asoc->this_sack_highest_gap, MAX_TSN)) { /* we are beyond the tsn in the sack */ break; } if (tp1->sent >= SCTP_DATAGRAM_RESEND) { /* either a RESEND, ACKED, or MARKED */ /* skip */ tp1 = TAILQ_NEXT(tp1, sctp_next); 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. */ tp1 = TAILQ_NEXT(tp1, sctp_next); continue; } else if (tp1->whoTo && compare_with_wrap(tp1->rec.data.TSN_seq, tp1->whoTo->this_sack_highest_newack, MAX_TSN)) { /* * 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. */ tp1 = TAILQ_NEXT(tp1, sctp_next); 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && 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) && compare_with_wrap(this_sack_lowest_newack, tp1->rec.data.TSN_seq, MAX_TSN)) { 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) && (SCTP_BASE_SYSCTL(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 ((compare_with_wrap(biggest_tsn_newly_acked, tp1->rec.data.fast_retran_tsn, MAX_TSN)) || (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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && 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) && compare_with_wrap(this_sack_lowest_newack, tp1->rec.data.TSN_seq, MAX_TSN)) { 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 (compare_with_wrap(tp1->rec.data.TSN_seq, biggest_tsn_newly_acked, MAX_TSN)) { /* * 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && 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) && compare_with_wrap(this_sack_lowest_newack, tp1->rec.data.TSN_seq, MAX_TSN)) { 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++; tp1 = TAILQ_NEXT(tp1, sctp_next); 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { /* * 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) { /* * 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); } } tp1 = TAILQ_NEXT(tp1, sctp_next); } /* while (tp1) */ 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); } tp1 = TAILQ_FIRST(&asoc->sent_queue); while (tp1) { 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; } tp2 = TAILQ_NEXT(tp1, sctp_next); /* * 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 (compare_with_wrap(tp1->rec.data.TSN_seq, asoc->advanced_peer_ack_point, MAX_TSN)) { 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; } /* * If we hit here we just dumped tp1, move to next tsn on * sent queue. */ tp1 = tp2; } 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; 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) { 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 (compare_with_wrap(asoc->last_acked_seq, cumack, MAX_TSN)) { /* 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->sent_queue_cnt * 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) { 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 ((cumack == send_s) || compare_with_wrap(cumack, send_s, MAX_TSN)) { #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 (compare_with_wrap(cumack, asoc->last_acked_seq, MAX_TSN)) { /* process the new consecutive TSN first */ tp1 = TAILQ_FIRST(&asoc->sent_queue); while (tp1) { tp2 = TAILQ_NEXT(tp1, sctp_next); if (compare_with_wrap(cumack, tp1->rec.data.TSN_seq, MAX_TSN) || 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); } 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); } tp1->data = NULL; asoc->sent_queue_cnt--; sctp_free_a_chunk(stcb, tp1); tp1 = tp2; } 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 (compare_with_wrap(asoc->last_acked_seq, asoc->nonce_wait_tsn, MAX_TSN) || (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 (compare_with_wrap(asoc->last_acked_seq, asoc->nonce_resync_tsn, MAX_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->sent_queue_cnt * 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) { 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_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 (compare_with_wrap(cumack, asoc->advanced_peer_ack_point, MAX_TSN)) { 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 (compare_with_wrap(asoc->advanced_peer_ack_point, cumack, MAX_TSN)) { /* * 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. */ if (compare_with_wrap(asoc->advanced_peer_ack_point, old_adv_peer_ack_point, MAX_TSN)) { 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 */ lchk->rec.data.fwd_tsn_cnt++; if (lchk->rec.data.fwd_tsn_cnt > 3) { send_forward_tsn(stcb, asoc); lchk->rec.data.fwd_tsn_cnt = 0; } } } 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 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 (cum_ack == send_s || compare_with_wrap(cum_ack, send_s, MAX_TSN)) { 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 (compare_with_wrap(asoc->last_acked_seq, last_tsn, MAX_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) { 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 */ tp1 = TAILQ_FIRST(&asoc->sent_queue); while (tp1) { if (compare_with_wrap(last_tsn, tp1->rec.data.TSN_seq, MAX_TSN) || 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; } tp1 = TAILQ_NEXT(tp1, sctp_next); } 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 ((biggest_tsn_acked == send_s) || (compare_with_wrap(biggest_tsn_acked, send_s, MAX_TSN))) { /* * 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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { 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 sendqueue */ /********************************************/ asoc->last_acked_seq = cum_ack; tp1 = TAILQ_FIRST(&asoc->sent_queue); if (tp1 == NULL) goto done_with_it; do { if (compare_with_wrap(tp1->rec.data.TSN_seq, cum_ack, MAX_TSN)) { break; } if (tp1->sent == SCTP_DATAGRAM_UNSENT) { /* no more sent on list */ printf("Warning, tp1->sent == %d and its now acked?\n", tp1->sent); } tp2 = TAILQ_NEXT(tp1, sctp_next); TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next); if (tp1->pr_sctp_on) { if (asoc->pr_sctp_cnt != 0) asoc->pr_sctp_cnt--; } if ((TAILQ_FIRST(&asoc->sent_queue) == NULL) && (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; } if (tp1->data) { /* sa_ignore NO_NULL_CHK */ sctp_free_bufspace(stcb, asoc, tp1, 1); sctp_m_freem(tp1->data); 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); } tp1->data = NULL; asoc->sent_queue_cnt--; sctp_free_a_chunk(stcb, tp1); wake_him++; tp1 = tp2; } while (tp1 != NULL); done_with_it: /* 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 (compare_with_wrap(asoc->last_acked_seq, asoc->fast_recovery_tsn, MAX_TSN) || 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); else if (asoc->saw_sack_with_frags) { int cnt_revoked = 0; tp1 = TAILQ_FIRST(&asoc->sent_queue); if (tp1 != NULL) { /* Peer revoked all dg's marked or acked */ TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { /* * EY- maybe check only if it is nr_acked * nr_marked may not be possible */ if ((tp1->sent == SCTP_DATAGRAM_NR_ACKED) || (tp1->sent == SCTP_DATAGRAM_NR_MARKED)) { continue; } 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_seg || num_nr_seg) asoc->saw_sack_with_frags = 1; else asoc->saw_sack_with_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_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 (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && 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 (compare_with_wrap(asoc->last_acked_seq, asoc->nonce_wait_tsn, MAX_TSN) || (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 (compare_with_wrap(asoc->last_acked_seq, asoc->nonce_resync_tsn, MAX_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) && ((compare_with_wrap(asoc->last_acked_seq, asoc->sat_t3_recovery_tsn, MAX_TSN) || (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->sent_queue_cnt * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)), a_rwnd); } asoc->peers_rwnd = sctp_sbspace_sub(a_rwnd, (uint32_t) (asoc->total_flight + (asoc->sent_queue_cnt * 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) { asoc->sent_queue_retran_cnt++; } } } done_once = 1; goto again; } /*********************************************/ /* Here we perform PR-SCTP procedures */ /* (section 4.2) */ /*********************************************/ /* C1. update advancedPeerAckPoint */ if (compare_with_wrap(cum_ack, asoc->advanced_peer_ack_point, MAX_TSN)) { 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 (compare_with_wrap(asoc->advanced_peer_ack_point, cum_ack, MAX_TSN)) { /* * 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. */ 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 (compare_with_wrap(asoc->advanced_peer_ack_point, old_adv_peer_ack_point, MAX_TSN)) { 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 */ lchk->rec.data.fwd_tsn_cnt++; if (lchk->rec.data.fwd_tsn_cnt > 3) { send_forward_tsn(stcb, asoc); lchk->rec.data.fwd_tsn_cnt = 0; } } } 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); } 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; int tt; asoc = &stcb->asoc; tt = strmin->last_sequence_delivered; /* * First deliver anything prior to and including the stream no that * came in */ ctl = TAILQ_FIRST(&strmin->inqueue); while (ctl) { nctl = TAILQ_NEXT(ctl, next); if (compare_with_wrap(tt, ctl->sinfo_ssn, MAX_SEQ) || (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; } ctl = nctl; } /* * now we must deliver things in queue the normal way if any are * now ready. */ tt = strmin->last_sequence_delivered + 1; ctl = TAILQ_FIRST(&strmin->inqueue); while (ctl) { nctl = TAILQ_NEXT(ctl, next); 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; } ctl = nctl; } } 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, *at; if (!TAILQ_EMPTY(&asoc->reasmqueue)) { /* 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. */ chk = TAILQ_FIRST(&asoc->reasmqueue); while (chk) { at = TAILQ_NEXT(chk, sctp_next); /* * 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)) { chk = at; continue; } if (chk->rec.data.stream_seq == seq) { /* It needs to be tossed */ TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (compare_with_wrap(chk->rec.data.TSN_seq, asoc->tsn_last_delivered, MAX_TSN)) { 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 && (compare_with_wrap(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered, MAX_SEQ))) { /* * 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 (compare_with_wrap(chk->rec.data.stream_seq, seq, MAX_SEQ)) { /* * If the stream_seq is > than the purging * one, we are done */ break; } chk = at; } } } 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, tsn, gap; unsigned int i, fwd_sz, cumack_set_flag, m_size, fnd = 0; uint32_t str_seq; struct sctp_stream_in *strm; struct sctp_tmit_chunk *chk, *at; 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 (compare_with_wrap(asoc->cumulative_tsn, new_cum_tsn, MAX_TSN) || asoc->cumulative_tsn == new_cum_tsn) { /* Already got there ... */ return; } if (compare_with_wrap(new_cum_tsn, asoc->highest_tsn_inside_map, MAX_TSN)) { asoc->highest_tsn_inside_map = new_cum_tsn; } if (compare_with_wrap(new_cum_tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { asoc->highest_tsn_inside_nr_map = new_cum_tsn; } /* * 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++) { SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, i); SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, i); /* FIX ME add something to set up highest TSN in map */ } if (compare_with_wrap(new_cum_tsn, asoc->highest_tsn_inside_nr_map, MAX_TSN)) { asoc->highest_tsn_inside_nr_map = new_cum_tsn; } if (compare_with_wrap(new_cum_tsn, asoc->highest_tsn_inside_map, MAX_TSN) || new_cum_tsn == asoc->highest_tsn_inside_map) { /* We must back down to see what the new highest is */ for (tsn = new_cum_tsn; (compare_with_wrap(tsn, asoc->mapping_array_base_tsn, MAX_TSN) || (tsn == asoc->mapping_array_base_tsn)); tsn--) { SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn); if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { asoc->highest_tsn_inside_map = tsn; fnd = 1; break; } } if (!fnd) { asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; } } /* * Now after marking all, slide thing forward but no sack * please. */ sctp_slide_mapping_arrays(stcb); } /*************************************************************/ /* 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); } if (!TAILQ_EMPTY(&asoc->reasmqueue)) { /* 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. */ chk = TAILQ_FIRST(&asoc->reasmqueue); while (chk) { at = TAILQ_NEXT(chk, sctp_next); if ((compare_with_wrap(new_cum_tsn, chk->rec.data.TSN_seq, MAX_TSN)) || (new_cum_tsn == chk->rec.data.TSN_seq)) { /* It needs to be tossed */ TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (compare_with_wrap(chk->rec.data.TSN_seq, asoc->tsn_last_delivered, MAX_TSN)) { 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 && (compare_with_wrap(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered, MAX_SEQ))) { /* * 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; } chk = at; } } /*******************************************************/ /* 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) && (compare_with_wrap(ctl->sinfo_ssn, stseq->sequence, MAX_SEQ))) { /* We are past our victim SSN */ break; } } strm = &asoc->strmin[stseq->stream]; if (compare_with_wrap(stseq->sequence, strm->last_sequence_delivered, MAX_SEQ)) { /* 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); } if (TAILQ_FIRST(&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/sctputil.c b/sys/netinet/sctputil.c index bcc253d9baf6..0f2805fb613d 100644 --- a/sys/netinet/sctputil.c +++ b/sys/netinet/sctputil.c @@ -1,6957 +1,6943 @@ /*- * 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 #define NUMBER_OF_MTU_SIZES 18 #if defined(__Windows__) && !defined(SCTP_LOCAL_TRACE_BUF) #include "eventrace_netinet.h" #include "sctputil.tmh" /* this is the file that will be auto * generated */ #else #ifndef KTR_SCTP #define KTR_SCTP KTR_SUBSYS #endif #endif 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->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); 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:%x flight was %d corrected to %d\n", (uint32_t) 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 /* * a list of sizes based on typical mtu's, used only if next hop size not * returned. */ static int sctp_mtu_sizes[] = { 68, 296, 508, 512, 544, 576, 1006, 1492, 1500, 1536, 2002, 2048, 4352, 4464, 8166, 17914, 32000, 65535 }; 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); } } int find_next_best_mtu(int totsz) { int i, perfer; /* * if we are in here we must find the next best fit based on the * size of the dg that failed to be sent. */ perfer = 0; for (i = 0; i < NUMBER_OF_MTU_SIZES; i++) { if (totsz < sctp_mtu_sizes[i]) { perfer = i - 1; if (perfer < 0) perfer = 0; break; } } return (sctp_mtu_sizes[perfer]); } 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 = (uint8_t) SCTP_BASE_SYSCTL(sctp_cmt_on_off); /* EY Init nr_sack variable */ asoc->sctp_nr_sack_on_off = (uint8_t) SCTP_BASE_SYSCTL(sctp_nr_sack_on_off); /* JRS 5/21/07 - Init CMT PF variables */ 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_cwr_tsn = asoc->init_seq_number - 1; 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; #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("smallest_mtu init'd with asoc to :%d\n", asoc->smallest_mtu); #endif 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; /* * JRS - Pick the default congestion control module based on the * sysctl. */ switch (m->sctp_ep.sctp_default_cc_module) { /* JRS - Standard TCP congestion control */ case SCTP_CC_RFC2581: { stcb->asoc.congestion_control_module = SCTP_CC_RFC2581; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } /* JRS - High Speed TCP congestion control (Floyd) */ case SCTP_CC_HSTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HSTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_hs_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_hs_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } /* JRS - HTCP congestion control */ case SCTP_CC_HTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_htcp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_htcp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_htcp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_htcp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_htcp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_htcp_cwnd_update_after_fr_timer; break; } /* JRS - By default, use RFC2581 */ default: { stcb->asoc.congestion_control_module = SCTP_CC_RFC2581; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } } /* * 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->strmout[i].next_spoke.tqe_next = 0; asoc->strmout[i].next_spoke.tqe_prev = 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->out_wheel); 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) { - int i, limit; + unsigned int i, limit; - printf("Mapping size:%d baseTSN:%8.8x cumAck:%8.8x highestTSN:%8.8x\n", + 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 - ); - limit = asoc->mapping_array_size; - for (i = asoc->mapping_array_size; i >= 0; i--) { - if (asoc->mapping_array[i]) { - limit = i; + 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; } } - if (limit == 0) - limit = 1; + printf("Renegable mapping array (last %d entries are zero):\n", asoc->mapping_array_size - limit); for (i = 0; i < limit; i++) { - printf("%2.2x ", asoc->mapping_array[i]); + printf("%2.2x%c", asoc->mapping_array[i], ((i + 1) % 16) ? ' ' : '\n'); if (((i + 1) % 16) == 0) printf("\n"); } - printf("\n"); - printf("NR Mapping size:%d baseTSN:%8.8x highestTSN:%8.8x\n", - asoc->mapping_array_size, - asoc->mapping_array_base_tsn, - asoc->highest_tsn_inside_nr_map - ); - limit = asoc->mapping_array_size; - for (i = asoc->mapping_array_size; i >= 0; i--) { - if (asoc->nr_mapping_array[i]) { - limit = i; + if (limit % 16) + printf("\n"); + for (limit = asoc->mapping_array_size; limit > 1; limit--) { + if (asoc->nr_mapping_array[limit - 1]) { break; } } - if (limit == 0) - limit = 1; - + 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 ", asoc->nr_mapping_array[i]); - if (((i + 1) % 16) == 0) - printf("\n"); + printf("%2.2x%c", asoc->mapping_array[i], ((i + 1) % 16) ? ' ' : '\n'); } - printf("\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); } #if defined(SCTP_USE_THREAD_BASED_ITERATOR) static void sctp_iterator_work(struct sctp_iterator *it) { int iteration_count = 0; int inp_skip = 0; SCTP_ITERATOR_LOCK(); if (it->inp) { SCTP_INP_DECR_REF(it->inp); } if (it->inp == NULL) { /* iterator is complete */ done_with_iterator: SCTP_ITERATOR_UNLOCK(); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); return; } select_a_new_ep: SCTP_INP_WLOCK(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_WUNLOCK(it->inp); goto done_with_iterator; } SCTP_INP_WUNLOCK(it->inp); it->inp = LIST_NEXT(it->inp, sctp_list); if (it->inp == NULL) { goto done_with_iterator; } SCTP_INP_WLOCK(it->inp); } SCTP_INP_WUNLOCK(it->inp); 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_ITERATOR_LOCK(); 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; SCTP_INP_WLOCK(it->inp); SCTP_INP_WUNLOCK(it->inp); if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; } else { SCTP_INP_INFO_RLOCK(); it->inp = LIST_NEXT(it->inp, sctp_list); SCTP_INP_INFO_RUNLOCK(); } if (it->inp == NULL) { goto done_with_iterator; } goto select_a_new_ep; } void sctp_iterator_worker(void) { struct sctp_iterator *it = NULL; /* This function is called with the WQ lock in place */ SCTP_BASE_INFO(iterator_running) = 1; again: it = TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead)); while (it) { /* now lets work on this one */ TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr); SCTP_IPI_ITERATOR_WQ_UNLOCK(); sctp_iterator_work(it); SCTP_IPI_ITERATOR_WQ_LOCK(); /* sa_ignore FREED_MEMORY */ it = TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead)); } if (TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead))) { goto again; } SCTP_BASE_INFO(iterator_running) = 0; return; } #endif static void sctp_handle_addr_wq(void) { /* deal with the ADDR wq from the rtsock calls */ struct sctp_laddr *wi; 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_IPI_ITERATOR_WQ_LOCK(); wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq)); while (wi != NULL) { LIST_REMOVE(wi, sctp_nxt_addr); LIST_INSERT_HEAD(&asc->list_of_work, wi, sctp_nxt_addr); asc->cnt++; wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq)); } SCTP_IPI_ITERATOR_WQ_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; struct sctp_iterator *it = NULL; 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; if (tmr->type == SCTP_TIMER_TYPE_ITERATOR) { it = (struct sctp_iterator *)inp; inp = NULL; } 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_ITERATOR: SCTP_STAT_INCR(sctps_timoiterator); sctp_iterator_timer(it); 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_DIRECTLY_NOCMPSET); 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_ITERATOR: { struct sctp_iterator *it; it = (struct sctp_iterator *)inp; tmr = &it->tmr; to_ticks = SCTP_ITERATOR_TICKS; } 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_ITERATOR: { struct sctp_iterator *it; it = (struct sctp_iterator *)inp; tmr = &it->tmr; } 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; #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_mtu_size_reset(%p, asoc:%p mtu:%d\n", inp, asoc, mtu); #endif 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); /* 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 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; } 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; struct sctp_stream_queue_pending *sp; int i; asoc = &stcb->asoc; 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; } /* now through all the gunk freeing chunks */ if (holds_lock == 0) { SCTP_TCB_SEND_LOCK(stcb); } /* sent queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->sent_queue)) { chk = TAILQ_FIRST(&asoc->sent_queue); while (chk) { 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 */ chk = TAILQ_FIRST(&asoc->sent_queue); } } /* pending send queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->send_queue)) { chk = TAILQ_FIRST(&asoc->send_queue); while (chk) { 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 */ chk = TAILQ_FIRST(&asoc->send_queue); } } for (i = 0; i < stcb->asoc.streamoutcnt; i++) { /* For each stream */ outs = &stcb->asoc.strmout[i]; /* clean up any sends there */ stcb->asoc.locked_on_sending = NULL; sp = TAILQ_FIRST(&outs->outqueue); while (sp) { stcb->asoc.stream_queue_cnt--; TAILQ_REMOVE(&outs->outqueue, sp, next); sctp_free_spbufspace(stcb, asoc, sp); 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 */ sp = TAILQ_FIRST(&outs->outqueue); } } 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 (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 } else { 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); } } } } #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); control = TAILQ_FIRST(&old_inp->read_queue); /* Pull off all for out target stcb */ while (control) { nctl = TAILQ_NEXT(control, next); 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); } } control = nctl; } 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 */ control = TAILQ_FIRST(&tmp_queue); SCTP_INP_READ_LOCK(new_inp); while (control) { nctl = TAILQ_NEXT(control, next); 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); } control = nctl; } 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 (!(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?? */ 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 (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; 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. */ next_on_sent: tp1 = TAILQ_FIRST(&stcb->asoc.send_queue); /* * recurse throught the send_queue too, starting at the * beginning. */ if ((tp1) && (tp1->rec.data.stream_number == stream) && (tp1->rec.data.stream_seq == seq)) { /* * 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++; goto next_on_sent; } } 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); sp = TAILQ_FIRST(&strq->outqueue); while (sp->strseq <= seq) { /* 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; chk->whoTo = sp->net; 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; } else { /* Next one please */ sp = TAILQ_NEXT(sp, next); } } /* End while */ 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, *embuf = 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->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); } embuf = m; 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; embuf = m; 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_IPI_ITERATOR_WQ_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_IPI_ITERATOR_WQ_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); }