1/*-
2 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD$");
35
36#include <netinet/sctp_os.h>
37#include <netinet/sctp_var.h>
38#include <netinet/sctp_sysctl.h>
39#include <netinet/sctp_pcb.h>
40#include <netinet/sctp_header.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp_output.h>
43#include <netinet/sctp_input.h>
44#include <netinet/sctp_indata.h>
45#include <netinet/sctp_uio.h>
46#include <netinet/sctp_timer.h>
47
48
49/*
50 * NOTES: On the outbound side of things I need to check the sack timer to
51 * see if I should generate a sack into the chunk queue (if I have data to
52 * send that is and will be sending it .. for bundling.
53 *
54 * The callback in sctp_usrreq.c will get called when the socket is read from.
55 * This will cause sctp_service_queues() to get called on the top entry in
56 * the list.
57 */
58
59void
60sctp_set_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc)
61{
62	asoc->my_rwnd = sctp_calc_rwnd(stcb, asoc);
63}
64
65/* Calculate what the rwnd would be */
66uint32_t
67sctp_calc_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc)
68{
69	uint32_t calc = 0;
70
71	/*
72	 * This is really set wrong with respect to a 1-2-m socket. Since
73	 * the sb_cc is the count that everyone as put up. When we re-write
74	 * sctp_soreceive then we will fix this so that ONLY this
75	 * associations data is taken into account.
76	 */
77	if (stcb->sctp_socket == NULL)
78		return (calc);
79
80	if (stcb->asoc.sb_cc == 0 &&
81	    asoc->size_on_reasm_queue == 0 &&
82	    asoc->size_on_all_streams == 0) {
83		/* Full rwnd granted */
84		calc = max(SCTP_SB_LIMIT_RCV(stcb->sctp_socket), SCTP_MINIMAL_RWND);
85		return (calc);
86	}
87	/* get actual space */
88	calc = (uint32_t) sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv);
89
90	/*
91	 * take out what has NOT been put on socket queue and we yet hold
92	 * for putting up.
93	 */
94	calc = sctp_sbspace_sub(calc, (uint32_t) (asoc->size_on_reasm_queue +
95	    asoc->cnt_on_reasm_queue * MSIZE));
96	calc = sctp_sbspace_sub(calc, (uint32_t) (asoc->size_on_all_streams +
97	    asoc->cnt_on_all_streams * MSIZE));
98
99	if (calc == 0) {
100		/* out of space */
101		return (calc);
102	}
103	/* what is the overhead of all these rwnd's */
104	calc = sctp_sbspace_sub(calc, stcb->asoc.my_rwnd_control_len);
105	/*
106	 * If the window gets too small due to ctrl-stuff, reduce it to 1,
107	 * even it is 0. SWS engaged
108	 */
109	if (calc < stcb->asoc.my_rwnd_control_len) {
110		calc = 1;
111	}
112	return (calc);
113}
114
115
116
117/*
118 * Build out our readq entry based on the incoming packet.
119 */
120struct sctp_queued_to_read *
121sctp_build_readq_entry(struct sctp_tcb *stcb,
122    struct sctp_nets *net,
123    uint32_t tsn, uint32_t ppid,
124    uint32_t context, uint16_t stream_no,
125    uint16_t stream_seq, uint8_t flags,
126    struct mbuf *dm)
127{
128	struct sctp_queued_to_read *read_queue_e = NULL;
129
130	sctp_alloc_a_readq(stcb, read_queue_e);
131	if (read_queue_e == NULL) {
132		goto failed_build;
133	}
134	read_queue_e->sinfo_stream = stream_no;
135	read_queue_e->sinfo_ssn = stream_seq;
136	read_queue_e->sinfo_flags = (flags << 8);
137	read_queue_e->sinfo_ppid = ppid;
138	read_queue_e->sinfo_context = context;
139	read_queue_e->sinfo_timetolive = 0;
140	read_queue_e->sinfo_tsn = tsn;
141	read_queue_e->sinfo_cumtsn = tsn;
142	read_queue_e->sinfo_assoc_id = sctp_get_associd(stcb);
143	read_queue_e->whoFrom = net;
144	read_queue_e->length = 0;
145	atomic_add_int(&net->ref_count, 1);
146	read_queue_e->data = dm;
147	read_queue_e->spec_flags = 0;
148	read_queue_e->tail_mbuf = NULL;
149	read_queue_e->aux_data = NULL;
150	read_queue_e->stcb = stcb;
151	read_queue_e->port_from = stcb->rport;
152	read_queue_e->do_not_ref_stcb = 0;
153	read_queue_e->end_added = 0;
154	read_queue_e->some_taken = 0;
155	read_queue_e->pdapi_aborted = 0;
156failed_build:
157	return (read_queue_e);
158}
159
160
161/*
162 * Build out our readq entry based on the incoming packet.
163 */
164static struct sctp_queued_to_read *
165sctp_build_readq_entry_chk(struct sctp_tcb *stcb,
166    struct sctp_tmit_chunk *chk)
167{
168	struct sctp_queued_to_read *read_queue_e = NULL;
169
170	sctp_alloc_a_readq(stcb, read_queue_e);
171	if (read_queue_e == NULL) {
172		goto failed_build;
173	}
174	read_queue_e->sinfo_stream = chk->rec.data.stream_number;
175	read_queue_e->sinfo_ssn = chk->rec.data.stream_seq;
176	read_queue_e->sinfo_flags = (chk->rec.data.rcv_flags << 8);
177	read_queue_e->sinfo_ppid = chk->rec.data.payloadtype;
178	read_queue_e->sinfo_context = stcb->asoc.context;
179	read_queue_e->sinfo_timetolive = 0;
180	read_queue_e->sinfo_tsn = chk->rec.data.TSN_seq;
181	read_queue_e->sinfo_cumtsn = chk->rec.data.TSN_seq;
182	read_queue_e->sinfo_assoc_id = sctp_get_associd(stcb);
183	read_queue_e->whoFrom = chk->whoTo;
184	read_queue_e->aux_data = NULL;
185	read_queue_e->length = 0;
186	atomic_add_int(&chk->whoTo->ref_count, 1);
187	read_queue_e->data = chk->data;
188	read_queue_e->tail_mbuf = NULL;
189	read_queue_e->stcb = stcb;
190	read_queue_e->port_from = stcb->rport;
191	read_queue_e->spec_flags = 0;
192	read_queue_e->do_not_ref_stcb = 0;
193	read_queue_e->end_added = 0;
194	read_queue_e->some_taken = 0;
195	read_queue_e->pdapi_aborted = 0;
196failed_build:
197	return (read_queue_e);
198}
199
200
201struct mbuf *
202sctp_build_ctl_nchunk(struct sctp_inpcb *inp, struct sctp_sndrcvinfo *sinfo)
203{
204	struct sctp_extrcvinfo *seinfo;
205	struct sctp_sndrcvinfo *outinfo;
206	struct sctp_rcvinfo *rcvinfo;
207	struct sctp_nxtinfo *nxtinfo;
208	struct cmsghdr *cmh;
209	struct mbuf *ret;
210	int len;
211	int use_extended;
212	int provide_nxt;
213
214	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT) &&
215	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVRCVINFO) &&
216	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVNXTINFO)) {
217		/* user does not want any ancillary data */
218		return (NULL);
219	}
220	len = 0;
221	if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO)) {
222		len += CMSG_SPACE(sizeof(struct sctp_rcvinfo));
223	}
224	seinfo = (struct sctp_extrcvinfo *)sinfo;
225	if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO) &&
226	    (seinfo->sreinfo_next_flags & SCTP_NEXT_MSG_AVAIL)) {
227		provide_nxt = 1;
228		len += CMSG_SPACE(sizeof(struct sctp_rcvinfo));
229	} else {
230		provide_nxt = 0;
231	}
232	if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) {
233		if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) {
234			use_extended = 1;
235			len += CMSG_SPACE(sizeof(struct sctp_extrcvinfo));
236		} else {
237			use_extended = 0;
238			len += CMSG_SPACE(sizeof(struct sctp_sndrcvinfo));
239		}
240	} else {
241		use_extended = 0;
242	}
243
244	ret = sctp_get_mbuf_for_msg(len, 0, M_NOWAIT, 1, MT_DATA);
245	if (ret == NULL) {
246		/* No space */
247		return (ret);
248	}
249	SCTP_BUF_LEN(ret) = 0;
250
251	/* We need a CMSG header followed by the struct */
252	cmh = mtod(ret, struct cmsghdr *);
253	/*
254	 * Make sure that there is no un-initialized padding between the
255	 * cmsg header and cmsg data and after the cmsg data.
256	 */
257	memset(cmh, 0, len);
258	if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO)) {
259		cmh->cmsg_level = IPPROTO_SCTP;
260		cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_rcvinfo));
261		cmh->cmsg_type = SCTP_RCVINFO;
262		rcvinfo = (struct sctp_rcvinfo *)CMSG_DATA(cmh);
263		rcvinfo->rcv_sid = sinfo->sinfo_stream;
264		rcvinfo->rcv_ssn = sinfo->sinfo_ssn;
265		rcvinfo->rcv_flags = sinfo->sinfo_flags;
266		rcvinfo->rcv_ppid = sinfo->sinfo_ppid;
267		rcvinfo->rcv_tsn = sinfo->sinfo_tsn;
268		rcvinfo->rcv_cumtsn = sinfo->sinfo_cumtsn;
269		rcvinfo->rcv_context = sinfo->sinfo_context;
270		rcvinfo->rcv_assoc_id = sinfo->sinfo_assoc_id;
271		cmh = (struct cmsghdr *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_rcvinfo)));
272		SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_rcvinfo));
273	}
274	if (provide_nxt) {
275		cmh->cmsg_level = IPPROTO_SCTP;
276		cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_nxtinfo));
277		cmh->cmsg_type = SCTP_NXTINFO;
278		nxtinfo = (struct sctp_nxtinfo *)CMSG_DATA(cmh);
279		nxtinfo->nxt_sid = seinfo->sreinfo_next_stream;
280		nxtinfo->nxt_flags = 0;
281		if (seinfo->sreinfo_next_flags & SCTP_NEXT_MSG_IS_UNORDERED) {
282			nxtinfo->nxt_flags |= SCTP_UNORDERED;
283		}
284		if (seinfo->sreinfo_next_flags & SCTP_NEXT_MSG_IS_NOTIFICATION) {
285			nxtinfo->nxt_flags |= SCTP_NOTIFICATION;
286		}
287		if (seinfo->sreinfo_next_flags & SCTP_NEXT_MSG_ISCOMPLETE) {
288			nxtinfo->nxt_flags |= SCTP_COMPLETE;
289		}
290		nxtinfo->nxt_ppid = seinfo->sreinfo_next_ppid;
291		nxtinfo->nxt_length = seinfo->sreinfo_next_length;
292		nxtinfo->nxt_assoc_id = seinfo->sreinfo_next_aid;
293		cmh = (struct cmsghdr *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_nxtinfo)));
294		SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_nxtinfo));
295	}
296	if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) {
297		cmh->cmsg_level = IPPROTO_SCTP;
298		outinfo = (struct sctp_sndrcvinfo *)CMSG_DATA(cmh);
299		if (use_extended) {
300			cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_extrcvinfo));
301			cmh->cmsg_type = SCTP_EXTRCV;
302			memcpy(outinfo, sinfo, sizeof(struct sctp_extrcvinfo));
303			SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_extrcvinfo));
304		} else {
305			cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo));
306			cmh->cmsg_type = SCTP_SNDRCV;
307			*outinfo = *sinfo;
308			SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_sndrcvinfo));
309		}
310	}
311	return (ret);
312}
313
314
315static void
316sctp_mark_non_revokable(struct sctp_association *asoc, uint32_t tsn)
317{
318	uint32_t gap, i, cumackp1;
319	int fnd = 0;
320
321	if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
322		return;
323	}
324	cumackp1 = asoc->cumulative_tsn + 1;
325	if (SCTP_TSN_GT(cumackp1, tsn)) {
326		/*
327		 * this tsn is behind the cum ack and thus we don't need to
328		 * worry about it being moved from one to the other.
329		 */
330		return;
331	}
332	SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn);
333	if (!SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
334		SCTP_PRINTF("gap:%x tsn:%x\n", gap, tsn);
335		sctp_print_mapping_array(asoc);
336#ifdef INVARIANTS
337		panic("Things are really messed up now!!");
338#endif
339	}
340	SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap);
341	SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
342	if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
343		asoc->highest_tsn_inside_nr_map = tsn;
344	}
345	if (tsn == asoc->highest_tsn_inside_map) {
346		/* We must back down to see what the new highest is */
347		for (i = tsn - 1; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
348			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
349			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
350				asoc->highest_tsn_inside_map = i;
351				fnd = 1;
352				break;
353			}
354		}
355		if (!fnd) {
356			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
357		}
358	}
359}
360
361
362/*
363 * We are delivering currently from the reassembly queue. We must continue to
364 * deliver until we either: 1) run out of space. 2) run out of sequential
365 * TSN's 3) hit the SCTP_DATA_LAST_FRAG flag.
366 */
367static void
368sctp_service_reassembly(struct sctp_tcb *stcb, struct sctp_association *asoc)
369{
370	struct sctp_tmit_chunk *chk, *nchk;
371	uint16_t nxt_todel;
372	uint16_t stream_no;
373	int end = 0;
374	int cntDel;
375	struct sctp_queued_to_read *control, *ctl, *nctl;
376
377	if (stcb == NULL)
378		return;
379
380	cntDel = stream_no = 0;
381	if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) ||
382	    (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) ||
383	    (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) {
384		/* socket above is long gone or going.. */
385abandon:
386		asoc->fragmented_delivery_inprogress = 0;
387		TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
388			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
389			asoc->size_on_reasm_queue -= chk->send_size;
390			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
391			/*
392			 * Lose the data pointer, since its in the socket
393			 * buffer
394			 */
395			if (chk->data) {
396				sctp_m_freem(chk->data);
397				chk->data = NULL;
398			}
399			/* Now free the address and data */
400			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
401			/* sa_ignore FREED_MEMORY */
402		}
403		return;
404	}
405	SCTP_TCB_LOCK_ASSERT(stcb);
406	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
407		if (chk->rec.data.TSN_seq != (asoc->tsn_last_delivered + 1)) {
408			/* Can't deliver more :< */
409			return;
410		}
411		stream_no = chk->rec.data.stream_number;
412		nxt_todel = asoc->strmin[stream_no].last_sequence_delivered + 1;
413		if (nxt_todel != chk->rec.data.stream_seq &&
414		    (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) {
415			/*
416			 * Not the next sequence to deliver in its stream OR
417			 * unordered
418			 */
419			return;
420		}
421		if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) {
422
423			control = sctp_build_readq_entry_chk(stcb, chk);
424			if (control == NULL) {
425				/* out of memory? */
426				return;
427			}
428			/* save it off for our future deliveries */
429			stcb->asoc.control_pdapi = control;
430			if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG)
431				end = 1;
432			else
433				end = 0;
434			sctp_mark_non_revokable(asoc, chk->rec.data.TSN_seq);
435			sctp_add_to_readq(stcb->sctp_ep,
436			    stcb, control, &stcb->sctp_socket->so_rcv, end,
437			    SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED);
438			cntDel++;
439		} else {
440			if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG)
441				end = 1;
442			else
443				end = 0;
444			sctp_mark_non_revokable(asoc, chk->rec.data.TSN_seq);
445			if (sctp_append_to_readq(stcb->sctp_ep, stcb,
446			    stcb->asoc.control_pdapi,
447			    chk->data, end, chk->rec.data.TSN_seq,
448			    &stcb->sctp_socket->so_rcv)) {
449				/*
450				 * something is very wrong, either
451				 * control_pdapi is NULL, or the tail_mbuf
452				 * is corrupt, or there is a EOM already on
453				 * the mbuf chain.
454				 */
455				if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
456					goto abandon;
457				} else {
458#ifdef INVARIANTS
459					if ((stcb->asoc.control_pdapi == NULL) || (stcb->asoc.control_pdapi->tail_mbuf == NULL)) {
460						panic("This should not happen control_pdapi NULL?");
461					}
462					/* if we did not panic, it was a EOM */
463					panic("Bad chunking ??");
464#else
465					if ((stcb->asoc.control_pdapi == NULL) || (stcb->asoc.control_pdapi->tail_mbuf == NULL)) {
466						SCTP_PRINTF("This should not happen control_pdapi NULL?\n");
467					}
468					SCTP_PRINTF("Bad chunking ??\n");
469					SCTP_PRINTF("Dumping re-assembly queue this will probably hose the association\n");
470
471#endif
472					goto abandon;
473				}
474			}
475			cntDel++;
476		}
477		/* pull it we did it */
478		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
479		if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) {
480			asoc->fragmented_delivery_inprogress = 0;
481			if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) {
482				asoc->strmin[stream_no].last_sequence_delivered++;
483			}
484			if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0) {
485				SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs);
486			}
487		} else if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) {
488			/*
489			 * turn the flag back on since we just  delivered
490			 * yet another one.
491			 */
492			asoc->fragmented_delivery_inprogress = 1;
493		}
494		asoc->tsn_of_pdapi_last_delivered = chk->rec.data.TSN_seq;
495		asoc->last_flags_delivered = chk->rec.data.rcv_flags;
496		asoc->last_strm_seq_delivered = chk->rec.data.stream_seq;
497		asoc->last_strm_no_delivered = chk->rec.data.stream_number;
498
499		asoc->tsn_last_delivered = chk->rec.data.TSN_seq;
500		asoc->size_on_reasm_queue -= chk->send_size;
501		sctp_ucount_decr(asoc->cnt_on_reasm_queue);
502		/* free up the chk */
503		chk->data = NULL;
504		sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
505
506		if (asoc->fragmented_delivery_inprogress == 0) {
507			/*
508			 * Now lets see if we can deliver the next one on
509			 * the stream
510			 */
511			struct sctp_stream_in *strm;
512
513			strm = &asoc->strmin[stream_no];
514			nxt_todel = strm->last_sequence_delivered + 1;
515			TAILQ_FOREACH_SAFE(ctl, &strm->inqueue, next, nctl) {
516				/* Deliver more if we can. */
517				if (nxt_todel == ctl->sinfo_ssn) {
518					TAILQ_REMOVE(&strm->inqueue, ctl, next);
519					asoc->size_on_all_streams -= ctl->length;
520					sctp_ucount_decr(asoc->cnt_on_all_streams);
521					strm->last_sequence_delivered++;
522					sctp_mark_non_revokable(asoc, ctl->sinfo_tsn);
523					sctp_add_to_readq(stcb->sctp_ep, stcb,
524					    ctl,
525					    &stcb->sctp_socket->so_rcv, 1,
526					    SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED);
527				} else {
528					break;
529				}
530				nxt_todel = strm->last_sequence_delivered + 1;
531			}
532			break;
533		}
534	}
535}
536
537/*
538 * Queue the chunk either right into the socket buffer if it is the next one
539 * to go OR put it in the correct place in the delivery queue.  If we do
540 * append to the so_buf, keep doing so until we are out of order. One big
541 * question still remains, what to do when the socket buffer is FULL??
542 */
543static void
544sctp_queue_data_to_stream(struct sctp_tcb *stcb, struct sctp_association *asoc,
545    struct sctp_queued_to_read *control, int *abort_flag)
546{
547	/*
548	 * FIX-ME maybe? What happens when the ssn wraps? If we are getting
549	 * all the data in one stream this could happen quite rapidly. One
550	 * could use the TSN to keep track of things, but this scheme breaks
551	 * down in the other type of stream useage that could occur. Send a
552	 * single msg to stream 0, send 4Billion messages to stream 1, now
553	 * send a message to stream 0. You have a situation where the TSN
554	 * has wrapped but not in the stream. Is this worth worrying about
555	 * or should we just change our queue sort at the bottom to be by
556	 * TSN.
557	 *
558	 * Could it also be legal for a peer to send ssn 1 with TSN 2 and ssn 2
559	 * with TSN 1? If the peer is doing some sort of funky TSN/SSN
560	 * assignment this could happen... and I don't see how this would be
561	 * a violation. So for now I am undecided an will leave the sort by
562	 * SSN alone. Maybe a hybred approach is the answer
563	 *
564	 */
565	struct sctp_stream_in *strm;
566	struct sctp_queued_to_read *at;
567	int queue_needed;
568	uint16_t nxt_todel;
569	struct mbuf *oper;
570
571	queue_needed = 1;
572	asoc->size_on_all_streams += control->length;
573	sctp_ucount_incr(asoc->cnt_on_all_streams);
574	strm = &asoc->strmin[control->sinfo_stream];
575	nxt_todel = strm->last_sequence_delivered + 1;
576	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
577		sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_INTO_STRD);
578	}
579	SCTPDBG(SCTP_DEBUG_INDATA1,
580	    "queue to stream called for ssn:%u lastdel:%u nxt:%u\n",
581	    (uint32_t) control->sinfo_stream,
582	    (uint32_t) strm->last_sequence_delivered,
583	    (uint32_t) nxt_todel);
584	if (SCTP_SSN_GE(strm->last_sequence_delivered, control->sinfo_ssn)) {
585		/* The incoming sseq is behind where we last delivered? */
586		SCTPDBG(SCTP_DEBUG_INDATA1, "Duplicate S-SEQ:%d delivered:%d from peer, Abort  association\n",
587		    control->sinfo_ssn, strm->last_sequence_delivered);
588protocol_error:
589		/*
590		 * throw it in the stream so it gets cleaned up in
591		 * association destruction
592		 */
593		TAILQ_INSERT_HEAD(&strm->inqueue, control, next);
594		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
595		    0, M_NOWAIT, 1, MT_DATA);
596		if (oper) {
597			struct sctp_paramhdr *ph;
598			uint32_t *ippp;
599
600			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
601			    (sizeof(uint32_t) * 3);
602			ph = mtod(oper, struct sctp_paramhdr *);
603			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
604			ph->param_length = htons(SCTP_BUF_LEN(oper));
605			ippp = (uint32_t *) (ph + 1);
606			*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_1);
607			ippp++;
608			*ippp = control->sinfo_tsn;
609			ippp++;
610			*ippp = ((control->sinfo_stream << 16) | control->sinfo_ssn);
611		}
612		stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_1;
613		sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
614		*abort_flag = 1;
615		return;
616
617	}
618	if (nxt_todel == control->sinfo_ssn) {
619		/* can be delivered right away? */
620		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
621			sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_IMMED_DEL);
622		}
623		/* EY it wont be queued if it could be delivered directly */
624		queue_needed = 0;
625		asoc->size_on_all_streams -= control->length;
626		sctp_ucount_decr(asoc->cnt_on_all_streams);
627		strm->last_sequence_delivered++;
628
629		sctp_mark_non_revokable(asoc, control->sinfo_tsn);
630		sctp_add_to_readq(stcb->sctp_ep, stcb,
631		    control,
632		    &stcb->sctp_socket->so_rcv, 1,
633		    SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED);
634		TAILQ_FOREACH_SAFE(control, &strm->inqueue, next, at) {
635			/* all delivered */
636			nxt_todel = strm->last_sequence_delivered + 1;
637			if (nxt_todel == control->sinfo_ssn) {
638				TAILQ_REMOVE(&strm->inqueue, control, next);
639				asoc->size_on_all_streams -= control->length;
640				sctp_ucount_decr(asoc->cnt_on_all_streams);
641				strm->last_sequence_delivered++;
642				/*
643				 * We ignore the return of deliver_data here
644				 * since we always can hold the chunk on the
645				 * d-queue. And we have a finite number that
646				 * can be delivered from the strq.
647				 */
648				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
649					sctp_log_strm_del(control, NULL,
650					    SCTP_STR_LOG_FROM_IMMED_DEL);
651				}
652				sctp_mark_non_revokable(asoc, control->sinfo_tsn);
653				sctp_add_to_readq(stcb->sctp_ep, stcb,
654				    control,
655				    &stcb->sctp_socket->so_rcv, 1,
656				    SCTP_READ_LOCK_NOT_HELD,
657				    SCTP_SO_NOT_LOCKED);
658				continue;
659			}
660			break;
661		}
662	}
663	if (queue_needed) {
664		/*
665		 * Ok, we did not deliver this guy, find the correct place
666		 * to put it on the queue.
667		 */
668		if (SCTP_TSN_GE(asoc->cumulative_tsn, control->sinfo_tsn)) {
669			goto protocol_error;
670		}
671		if (TAILQ_EMPTY(&strm->inqueue)) {
672			/* Empty queue */
673			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
674				sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_INSERT_HD);
675			}
676			TAILQ_INSERT_HEAD(&strm->inqueue, control, next);
677		} else {
678			TAILQ_FOREACH(at, &strm->inqueue, next) {
679				if (SCTP_SSN_GT(at->sinfo_ssn, control->sinfo_ssn)) {
680					/*
681					 * one in queue is bigger than the
682					 * new one, insert before this one
683					 */
684					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
685						sctp_log_strm_del(control, at,
686						    SCTP_STR_LOG_FROM_INSERT_MD);
687					}
688					TAILQ_INSERT_BEFORE(at, control, next);
689					break;
690				} else if (at->sinfo_ssn == control->sinfo_ssn) {
691					/*
692					 * Gak, He sent me a duplicate str
693					 * seq number
694					 */
695					/*
696					 * foo bar, I guess I will just free
697					 * this new guy, should we abort
698					 * too? FIX ME MAYBE? Or it COULD be
699					 * that the SSN's have wrapped.
700					 * Maybe I should compare to TSN
701					 * somehow... sigh for now just blow
702					 * away the chunk!
703					 */
704
705					if (control->data)
706						sctp_m_freem(control->data);
707					control->data = NULL;
708					asoc->size_on_all_streams -= control->length;
709					sctp_ucount_decr(asoc->cnt_on_all_streams);
710					if (control->whoFrom) {
711						sctp_free_remote_addr(control->whoFrom);
712						control->whoFrom = NULL;
713					}
714					sctp_free_a_readq(stcb, control);
715					return;
716				} else {
717					if (TAILQ_NEXT(at, next) == NULL) {
718						/*
719						 * We are at the end, insert
720						 * it after this one
721						 */
722						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
723							sctp_log_strm_del(control, at,
724							    SCTP_STR_LOG_FROM_INSERT_TL);
725						}
726						TAILQ_INSERT_AFTER(&strm->inqueue,
727						    at, control, next);
728						break;
729					}
730				}
731			}
732		}
733	}
734}
735
736/*
737 * Returns two things: You get the total size of the deliverable parts of the
738 * first fragmented message on the reassembly queue. And you get a 1 back if
739 * all of the message is ready or a 0 back if the message is still incomplete
740 */
741static int
742sctp_is_all_msg_on_reasm(struct sctp_association *asoc, uint32_t * t_size)
743{
744	struct sctp_tmit_chunk *chk;
745	uint32_t tsn;
746
747	*t_size = 0;
748	chk = TAILQ_FIRST(&asoc->reasmqueue);
749	if (chk == NULL) {
750		/* nothing on the queue */
751		return (0);
752	}
753	if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0) {
754		/* Not a first on the queue */
755		return (0);
756	}
757	tsn = chk->rec.data.TSN_seq;
758	TAILQ_FOREACH(chk, &asoc->reasmqueue, sctp_next) {
759		if (tsn != chk->rec.data.TSN_seq) {
760			return (0);
761		}
762		*t_size += chk->send_size;
763		if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) {
764			return (1);
765		}
766		tsn++;
767	}
768	return (0);
769}
770
771static void
772sctp_deliver_reasm_check(struct sctp_tcb *stcb, struct sctp_association *asoc)
773{
774	struct sctp_tmit_chunk *chk;
775	uint16_t nxt_todel;
776	uint32_t tsize, pd_point;
777
778doit_again:
779	chk = TAILQ_FIRST(&asoc->reasmqueue);
780	if (chk == NULL) {
781		/* Huh? */
782		asoc->size_on_reasm_queue = 0;
783		asoc->cnt_on_reasm_queue = 0;
784		return;
785	}
786	if (asoc->fragmented_delivery_inprogress == 0) {
787		nxt_todel =
788		    asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered + 1;
789		if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) &&
790		    (nxt_todel == chk->rec.data.stream_seq ||
791		    (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED))) {
792			/*
793			 * Yep the first one is here and its ok to deliver
794			 * but should we?
795			 */
796			if (stcb->sctp_socket) {
797				pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket) >> SCTP_PARTIAL_DELIVERY_SHIFT,
798				    stcb->sctp_ep->partial_delivery_point);
799			} else {
800				pd_point = stcb->sctp_ep->partial_delivery_point;
801			}
802			if (sctp_is_all_msg_on_reasm(asoc, &tsize) || (tsize >= pd_point)) {
803				/*
804				 * Yes, we setup to start reception, by
805				 * backing down the TSN just in case we
806				 * can't deliver. If we
807				 */
808				asoc->fragmented_delivery_inprogress = 1;
809				asoc->tsn_last_delivered =
810				    chk->rec.data.TSN_seq - 1;
811				asoc->str_of_pdapi =
812				    chk->rec.data.stream_number;
813				asoc->ssn_of_pdapi = chk->rec.data.stream_seq;
814				asoc->pdapi_ppid = chk->rec.data.payloadtype;
815				asoc->fragment_flags = chk->rec.data.rcv_flags;
816				sctp_service_reassembly(stcb, asoc);
817			}
818		}
819	} else {
820		/*
821		 * Service re-assembly will deliver stream data queued at
822		 * the end of fragmented delivery.. but it wont know to go
823		 * back and call itself again... we do that here with the
824		 * got doit_again
825		 */
826		sctp_service_reassembly(stcb, asoc);
827		if (asoc->fragmented_delivery_inprogress == 0) {
828			/*
829			 * finished our Fragmented delivery, could be more
830			 * waiting?
831			 */
832			goto doit_again;
833		}
834	}
835}
836
837/*
838 * Dump onto the re-assembly queue, in its proper place. After dumping on the
839 * queue, see if anthing can be delivered. If so pull it off (or as much as
840 * we can. If we run out of space then we must dump what we can and set the
841 * appropriate flag to say we queued what we could.
842 */
843static void
844sctp_queue_data_for_reasm(struct sctp_tcb *stcb, struct sctp_association *asoc,
845    struct sctp_tmit_chunk *chk, int *abort_flag)
846{
847	struct mbuf *oper;
848	uint32_t cum_ackp1, prev_tsn, post_tsn;
849	struct sctp_tmit_chunk *at, *prev, *next;
850
851	prev = next = NULL;
852	cum_ackp1 = asoc->tsn_last_delivered + 1;
853	if (TAILQ_EMPTY(&asoc->reasmqueue)) {
854		/* This is the first one on the queue */
855		TAILQ_INSERT_HEAD(&asoc->reasmqueue, chk, sctp_next);
856		/*
857		 * we do not check for delivery of anything when only one
858		 * fragment is here
859		 */
860		asoc->size_on_reasm_queue = chk->send_size;
861		sctp_ucount_incr(asoc->cnt_on_reasm_queue);
862		if (chk->rec.data.TSN_seq == cum_ackp1) {
863			if (asoc->fragmented_delivery_inprogress == 0 &&
864			    (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) !=
865			    SCTP_DATA_FIRST_FRAG) {
866				/*
867				 * An empty queue, no delivery inprogress,
868				 * we hit the next one and it does NOT have
869				 * a FIRST fragment mark.
870				 */
871				SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, its not first, no fragmented delivery in progress\n");
872				oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
873				    0, M_NOWAIT, 1, MT_DATA);
874
875				if (oper) {
876					struct sctp_paramhdr *ph;
877					uint32_t *ippp;
878
879					SCTP_BUF_LEN(oper) =
880					    sizeof(struct sctp_paramhdr) +
881					    (sizeof(uint32_t) * 3);
882					ph = mtod(oper, struct sctp_paramhdr *);
883					ph->param_type =
884					    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
885					ph->param_length = htons(SCTP_BUF_LEN(oper));
886					ippp = (uint32_t *) (ph + 1);
887					*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_2);
888					ippp++;
889					*ippp = chk->rec.data.TSN_seq;
890					ippp++;
891					*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
892
893				}
894				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_2;
895				sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
896				*abort_flag = 1;
897			} else if (asoc->fragmented_delivery_inprogress &&
898			    (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == SCTP_DATA_FIRST_FRAG) {
899				/*
900				 * We are doing a partial delivery and the
901				 * NEXT chunk MUST be either the LAST or
902				 * MIDDLE fragment NOT a FIRST
903				 */
904				SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS a first and fragmented delivery in progress\n");
905				oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
906				    0, M_NOWAIT, 1, MT_DATA);
907				if (oper) {
908					struct sctp_paramhdr *ph;
909					uint32_t *ippp;
910
911					SCTP_BUF_LEN(oper) =
912					    sizeof(struct sctp_paramhdr) +
913					    (3 * sizeof(uint32_t));
914					ph = mtod(oper, struct sctp_paramhdr *);
915					ph->param_type =
916					    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
917					ph->param_length = htons(SCTP_BUF_LEN(oper));
918					ippp = (uint32_t *) (ph + 1);
919					*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_3);
920					ippp++;
921					*ippp = chk->rec.data.TSN_seq;
922					ippp++;
923					*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
924				}
925				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_3;
926				sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
927				*abort_flag = 1;
928			} else if (asoc->fragmented_delivery_inprogress) {
929				/*
930				 * Here we are ok with a MIDDLE or LAST
931				 * piece
932				 */
933				if (chk->rec.data.stream_number !=
934				    asoc->str_of_pdapi) {
935					/* Got to be the right STR No */
936					SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS not same stream number %d vs %d\n",
937					    chk->rec.data.stream_number,
938					    asoc->str_of_pdapi);
939					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
940					    0, M_NOWAIT, 1, MT_DATA);
941					if (oper) {
942						struct sctp_paramhdr *ph;
943						uint32_t *ippp;
944
945						SCTP_BUF_LEN(oper) =
946						    sizeof(struct sctp_paramhdr) +
947						    (sizeof(uint32_t) * 3);
948						ph = mtod(oper,
949						    struct sctp_paramhdr *);
950						ph->param_type =
951						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
952						ph->param_length =
953						    htons(SCTP_BUF_LEN(oper));
954						ippp = (uint32_t *) (ph + 1);
955						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_4);
956						ippp++;
957						*ippp = chk->rec.data.TSN_seq;
958						ippp++;
959						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
960					}
961					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_4;
962					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
963					*abort_flag = 1;
964				} else if ((asoc->fragment_flags & SCTP_DATA_UNORDERED) !=
965					    SCTP_DATA_UNORDERED &&
966				    chk->rec.data.stream_seq != asoc->ssn_of_pdapi) {
967					/* Got to be the right STR Seq */
968					SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it IS not same stream seq %d vs %d\n",
969					    chk->rec.data.stream_seq,
970					    asoc->ssn_of_pdapi);
971					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
972					    0, M_NOWAIT, 1, MT_DATA);
973					if (oper) {
974						struct sctp_paramhdr *ph;
975						uint32_t *ippp;
976
977						SCTP_BUF_LEN(oper) =
978						    sizeof(struct sctp_paramhdr) +
979						    (3 * sizeof(uint32_t));
980						ph = mtod(oper,
981						    struct sctp_paramhdr *);
982						ph->param_type =
983						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
984						ph->param_length =
985						    htons(SCTP_BUF_LEN(oper));
986						ippp = (uint32_t *) (ph + 1);
987						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_5);
988						ippp++;
989						*ippp = chk->rec.data.TSN_seq;
990						ippp++;
991						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
992
993					}
994					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_5;
995					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
996					*abort_flag = 1;
997				}
998			}
999		}
1000		return;
1001	}
1002	/* Find its place */
1003	TAILQ_FOREACH(at, &asoc->reasmqueue, sctp_next) {
1004		if (SCTP_TSN_GT(at->rec.data.TSN_seq, chk->rec.data.TSN_seq)) {
1005			/*
1006			 * one in queue is bigger than the new one, insert
1007			 * before this one
1008			 */
1009			/* A check */
1010			asoc->size_on_reasm_queue += chk->send_size;
1011			sctp_ucount_incr(asoc->cnt_on_reasm_queue);
1012			next = at;
1013			TAILQ_INSERT_BEFORE(at, chk, sctp_next);
1014			break;
1015		} else if (at->rec.data.TSN_seq == chk->rec.data.TSN_seq) {
1016			/* Gak, He sent me a duplicate str seq number */
1017			/*
1018			 * foo bar, I guess I will just free this new guy,
1019			 * should we abort too? FIX ME MAYBE? Or it COULD be
1020			 * that the SSN's have wrapped. Maybe I should
1021			 * compare to TSN somehow... sigh for now just blow
1022			 * away the chunk!
1023			 */
1024			if (chk->data) {
1025				sctp_m_freem(chk->data);
1026				chk->data = NULL;
1027			}
1028			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
1029			return;
1030		} else {
1031			prev = at;
1032			if (TAILQ_NEXT(at, sctp_next) == NULL) {
1033				/*
1034				 * We are at the end, insert it after this
1035				 * one
1036				 */
1037				/* check it first */
1038				asoc->size_on_reasm_queue += chk->send_size;
1039				sctp_ucount_incr(asoc->cnt_on_reasm_queue);
1040				TAILQ_INSERT_AFTER(&asoc->reasmqueue, at, chk, sctp_next);
1041				break;
1042			}
1043		}
1044	}
1045	/* Now the audits */
1046	if (prev) {
1047		prev_tsn = chk->rec.data.TSN_seq - 1;
1048		if (prev_tsn == prev->rec.data.TSN_seq) {
1049			/*
1050			 * Ok the one I am dropping onto the end is the
1051			 * NEXT. A bit of valdiation here.
1052			 */
1053			if ((prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1054			    SCTP_DATA_FIRST_FRAG ||
1055			    (prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1056			    SCTP_DATA_MIDDLE_FRAG) {
1057				/*
1058				 * Insert chk MUST be a MIDDLE or LAST
1059				 * fragment
1060				 */
1061				if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1062				    SCTP_DATA_FIRST_FRAG) {
1063					SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - It can be a midlle or last but not a first\n");
1064					SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, it's a FIRST!\n");
1065					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1066					    0, M_NOWAIT, 1, MT_DATA);
1067					if (oper) {
1068						struct sctp_paramhdr *ph;
1069						uint32_t *ippp;
1070
1071						SCTP_BUF_LEN(oper) =
1072						    sizeof(struct sctp_paramhdr) +
1073						    (3 * sizeof(uint32_t));
1074						ph = mtod(oper,
1075						    struct sctp_paramhdr *);
1076						ph->param_type =
1077						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1078						ph->param_length =
1079						    htons(SCTP_BUF_LEN(oper));
1080						ippp = (uint32_t *) (ph + 1);
1081						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_6);
1082						ippp++;
1083						*ippp = chk->rec.data.TSN_seq;
1084						ippp++;
1085						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1086
1087					}
1088					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_6;
1089					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1090					*abort_flag = 1;
1091					return;
1092				}
1093				if (chk->rec.data.stream_number !=
1094				    prev->rec.data.stream_number) {
1095					/*
1096					 * Huh, need the correct STR here,
1097					 * they must be the same.
1098					 */
1099					SCTP_PRINTF("Prev check - Gak, Evil plot, ssn:%d not the same as at:%d\n",
1100					    chk->rec.data.stream_number,
1101					    prev->rec.data.stream_number);
1102					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1103					    0, M_NOWAIT, 1, MT_DATA);
1104					if (oper) {
1105						struct sctp_paramhdr *ph;
1106						uint32_t *ippp;
1107
1108						SCTP_BUF_LEN(oper) =
1109						    sizeof(struct sctp_paramhdr) +
1110						    (3 * sizeof(uint32_t));
1111						ph = mtod(oper,
1112						    struct sctp_paramhdr *);
1113						ph->param_type =
1114						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1115						ph->param_length =
1116						    htons(SCTP_BUF_LEN(oper));
1117						ippp = (uint32_t *) (ph + 1);
1118						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_7);
1119						ippp++;
1120						*ippp = chk->rec.data.TSN_seq;
1121						ippp++;
1122						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1123					}
1124					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_7;
1125					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1126					*abort_flag = 1;
1127					return;
1128				}
1129				if ((prev->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0 &&
1130				    chk->rec.data.stream_seq !=
1131				    prev->rec.data.stream_seq) {
1132					/*
1133					 * Huh, need the correct STR here,
1134					 * they must be the same.
1135					 */
1136					SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - Gak, Evil plot, sseq:%d not the same as at:%d\n",
1137					    chk->rec.data.stream_seq,
1138					    prev->rec.data.stream_seq);
1139					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1140					    0, M_NOWAIT, 1, MT_DATA);
1141					if (oper) {
1142						struct sctp_paramhdr *ph;
1143						uint32_t *ippp;
1144
1145						SCTP_BUF_LEN(oper) =
1146						    sizeof(struct sctp_paramhdr) +
1147						    (3 * sizeof(uint32_t));
1148						ph = mtod(oper,
1149						    struct sctp_paramhdr *);
1150						ph->param_type =
1151						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1152						ph->param_length =
1153						    htons(SCTP_BUF_LEN(oper));
1154						ippp = (uint32_t *) (ph + 1);
1155						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_8);
1156						ippp++;
1157						*ippp = chk->rec.data.TSN_seq;
1158						ippp++;
1159						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1160					}
1161					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_8;
1162					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1163					*abort_flag = 1;
1164					return;
1165				}
1166			} else if ((prev->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1167			    SCTP_DATA_LAST_FRAG) {
1168				/* Insert chk MUST be a FIRST */
1169				if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) !=
1170				    SCTP_DATA_FIRST_FRAG) {
1171					SCTPDBG(SCTP_DEBUG_INDATA1, "Prev check - Gak, evil plot, its not FIRST and it must be!\n");
1172					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1173					    0, M_NOWAIT, 1, MT_DATA);
1174					if (oper) {
1175						struct sctp_paramhdr *ph;
1176						uint32_t *ippp;
1177
1178						SCTP_BUF_LEN(oper) =
1179						    sizeof(struct sctp_paramhdr) +
1180						    (3 * sizeof(uint32_t));
1181						ph = mtod(oper,
1182						    struct sctp_paramhdr *);
1183						ph->param_type =
1184						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1185						ph->param_length =
1186						    htons(SCTP_BUF_LEN(oper));
1187						ippp = (uint32_t *) (ph + 1);
1188						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_9);
1189						ippp++;
1190						*ippp = chk->rec.data.TSN_seq;
1191						ippp++;
1192						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1193
1194					}
1195					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_9;
1196					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1197					*abort_flag = 1;
1198					return;
1199				}
1200			}
1201		}
1202	}
1203	if (next) {
1204		post_tsn = chk->rec.data.TSN_seq + 1;
1205		if (post_tsn == next->rec.data.TSN_seq) {
1206			/*
1207			 * Ok the one I am inserting ahead of is my NEXT
1208			 * one. A bit of valdiation here.
1209			 */
1210			if (next->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) {
1211				/* Insert chk MUST be a last fragment */
1212				if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK)
1213				    != SCTP_DATA_LAST_FRAG) {
1214					SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Next is FIRST, we must be LAST\n");
1215					SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, its not a last!\n");
1216					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1217					    0, M_NOWAIT, 1, MT_DATA);
1218					if (oper) {
1219						struct sctp_paramhdr *ph;
1220						uint32_t *ippp;
1221
1222						SCTP_BUF_LEN(oper) =
1223						    sizeof(struct sctp_paramhdr) +
1224						    (3 * sizeof(uint32_t));
1225						ph = mtod(oper,
1226						    struct sctp_paramhdr *);
1227						ph->param_type =
1228						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1229						ph->param_length =
1230						    htons(SCTP_BUF_LEN(oper));
1231						ippp = (uint32_t *) (ph + 1);
1232						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_10);
1233						ippp++;
1234						*ippp = chk->rec.data.TSN_seq;
1235						ippp++;
1236						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1237					}
1238					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_10;
1239					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1240					*abort_flag = 1;
1241					return;
1242				}
1243			} else if ((next->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1244				    SCTP_DATA_MIDDLE_FRAG ||
1245				    (next->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1246			    SCTP_DATA_LAST_FRAG) {
1247				/*
1248				 * Insert chk CAN be MIDDLE or FIRST NOT
1249				 * LAST
1250				 */
1251				if ((chk->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) ==
1252				    SCTP_DATA_LAST_FRAG) {
1253					SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Next is a MIDDLE/LAST\n");
1254					SCTPDBG(SCTP_DEBUG_INDATA1, "Gak, Evil plot, new prev chunk is a LAST\n");
1255					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1256					    0, M_NOWAIT, 1, MT_DATA);
1257					if (oper) {
1258						struct sctp_paramhdr *ph;
1259						uint32_t *ippp;
1260
1261						SCTP_BUF_LEN(oper) =
1262						    sizeof(struct sctp_paramhdr) +
1263						    (3 * sizeof(uint32_t));
1264						ph = mtod(oper,
1265						    struct sctp_paramhdr *);
1266						ph->param_type =
1267						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1268						ph->param_length =
1269						    htons(SCTP_BUF_LEN(oper));
1270						ippp = (uint32_t *) (ph + 1);
1271						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_11);
1272						ippp++;
1273						*ippp = chk->rec.data.TSN_seq;
1274						ippp++;
1275						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1276
1277					}
1278					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_11;
1279					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1280					*abort_flag = 1;
1281					return;
1282				}
1283				if (chk->rec.data.stream_number !=
1284				    next->rec.data.stream_number) {
1285					/*
1286					 * Huh, need the correct STR here,
1287					 * they must be the same.
1288					 */
1289					SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Gak, Evil plot, ssn:%d not the same as at:%d\n",
1290					    chk->rec.data.stream_number,
1291					    next->rec.data.stream_number);
1292					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1293					    0, M_NOWAIT, 1, MT_DATA);
1294					if (oper) {
1295						struct sctp_paramhdr *ph;
1296						uint32_t *ippp;
1297
1298						SCTP_BUF_LEN(oper) =
1299						    sizeof(struct sctp_paramhdr) +
1300						    (3 * sizeof(uint32_t));
1301						ph = mtod(oper,
1302						    struct sctp_paramhdr *);
1303						ph->param_type =
1304						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1305						ph->param_length =
1306						    htons(SCTP_BUF_LEN(oper));
1307						ippp = (uint32_t *) (ph + 1);
1308						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_12);
1309						ippp++;
1310						*ippp = chk->rec.data.TSN_seq;
1311						ippp++;
1312						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1313
1314					}
1315					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_12;
1316					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1317					*abort_flag = 1;
1318					return;
1319				}
1320				if ((next->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0 &&
1321				    chk->rec.data.stream_seq !=
1322				    next->rec.data.stream_seq) {
1323					/*
1324					 * Huh, need the correct STR here,
1325					 * they must be the same.
1326					 */
1327					SCTPDBG(SCTP_DEBUG_INDATA1, "Next chk - Gak, Evil plot, sseq:%d not the same as at:%d\n",
1328					    chk->rec.data.stream_seq,
1329					    next->rec.data.stream_seq);
1330					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1331					    0, M_NOWAIT, 1, MT_DATA);
1332					if (oper) {
1333						struct sctp_paramhdr *ph;
1334						uint32_t *ippp;
1335
1336						SCTP_BUF_LEN(oper) =
1337						    sizeof(struct sctp_paramhdr) +
1338						    (3 * sizeof(uint32_t));
1339						ph = mtod(oper,
1340						    struct sctp_paramhdr *);
1341						ph->param_type =
1342						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1343						ph->param_length =
1344						    htons(SCTP_BUF_LEN(oper));
1345						ippp = (uint32_t *) (ph + 1);
1346						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_13);
1347						ippp++;
1348						*ippp = chk->rec.data.TSN_seq;
1349						ippp++;
1350						*ippp = ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq);
1351					}
1352					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_13;
1353					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1354					*abort_flag = 1;
1355					return;
1356				}
1357			}
1358		}
1359	}
1360	/* Do we need to do some delivery? check */
1361	sctp_deliver_reasm_check(stcb, asoc);
1362}
1363
1364/*
1365 * This is an unfortunate routine. It checks to make sure a evil guy is not
1366 * stuffing us full of bad packet fragments. A broken peer could also do this
1367 * but this is doubtful. It is to bad I must worry about evil crackers sigh
1368 * :< more cycles.
1369 */
1370static int
1371sctp_does_tsn_belong_to_reasm(struct sctp_association *asoc,
1372    uint32_t TSN_seq)
1373{
1374	struct sctp_tmit_chunk *at;
1375	uint32_t tsn_est;
1376
1377	TAILQ_FOREACH(at, &asoc->reasmqueue, sctp_next) {
1378		if (SCTP_TSN_GT(TSN_seq, at->rec.data.TSN_seq)) {
1379			/* is it one bigger? */
1380			tsn_est = at->rec.data.TSN_seq + 1;
1381			if (tsn_est == TSN_seq) {
1382				/* yep. It better be a last then */
1383				if ((at->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) !=
1384				    SCTP_DATA_LAST_FRAG) {
1385					/*
1386					 * Ok this guy belongs next to a guy
1387					 * that is NOT last, it should be a
1388					 * middle/last, not a complete
1389					 * chunk.
1390					 */
1391					return (1);
1392				} else {
1393					/*
1394					 * This guy is ok since its a LAST
1395					 * and the new chunk is a fully
1396					 * self- contained one.
1397					 */
1398					return (0);
1399				}
1400			}
1401		} else if (TSN_seq == at->rec.data.TSN_seq) {
1402			/* Software error since I have a dup? */
1403			return (1);
1404		} else {
1405			/*
1406			 * Ok, 'at' is larger than new chunk but does it
1407			 * need to be right before it.
1408			 */
1409			tsn_est = TSN_seq + 1;
1410			if (tsn_est == at->rec.data.TSN_seq) {
1411				/* Yep, It better be a first */
1412				if ((at->rec.data.rcv_flags & SCTP_DATA_FRAG_MASK) !=
1413				    SCTP_DATA_FIRST_FRAG) {
1414					return (1);
1415				} else {
1416					return (0);
1417				}
1418			}
1419		}
1420	}
1421	return (0);
1422}
1423
1424static int
1425sctp_process_a_data_chunk(struct sctp_tcb *stcb, struct sctp_association *asoc,
1426    struct mbuf **m, int offset, struct sctp_data_chunk *ch, int chk_length,
1427    struct sctp_nets *net, uint32_t * high_tsn, int *abort_flag,
1428    int *break_flag, int last_chunk)
1429{
1430	/* Process a data chunk */
1431	/* struct sctp_tmit_chunk *chk; */
1432	struct sctp_tmit_chunk *chk;
1433	uint32_t tsn, gap;
1434	struct mbuf *dmbuf;
1435	int the_len;
1436	int need_reasm_check = 0;
1437	uint16_t strmno, strmseq;
1438	struct mbuf *oper;
1439	struct sctp_queued_to_read *control;
1440	int ordered;
1441	uint32_t protocol_id;
1442	uint8_t chunk_flags;
1443	struct sctp_stream_reset_list *liste;
1444
1445	chk = NULL;
1446	tsn = ntohl(ch->dp.tsn);
1447	chunk_flags = ch->ch.chunk_flags;
1448	if ((chunk_flags & SCTP_DATA_SACK_IMMEDIATELY) == SCTP_DATA_SACK_IMMEDIATELY) {
1449		asoc->send_sack = 1;
1450	}
1451	protocol_id = ch->dp.protocol_id;
1452	ordered = ((chunk_flags & SCTP_DATA_UNORDERED) == 0);
1453	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
1454		sctp_log_map(tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, SCTP_MAP_TSN_ENTERS);
1455	}
1456	if (stcb == NULL) {
1457		return (0);
1458	}
1459	SCTP_LTRACE_CHK(stcb->sctp_ep, stcb, ch->ch.chunk_type, tsn);
1460	if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) {
1461		/* It is a duplicate */
1462		SCTP_STAT_INCR(sctps_recvdupdata);
1463		if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) {
1464			/* Record a dup for the next outbound sack */
1465			asoc->dup_tsns[asoc->numduptsns] = tsn;
1466			asoc->numduptsns++;
1467		}
1468		asoc->send_sack = 1;
1469		return (0);
1470	}
1471	/* Calculate the number of TSN's between the base and this TSN */
1472	SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn);
1473	if (gap >= (SCTP_MAPPING_ARRAY << 3)) {
1474		/* Can't hold the bit in the mapping at max array, toss it */
1475		return (0);
1476	}
1477	if (gap >= (uint32_t) (asoc->mapping_array_size << 3)) {
1478		SCTP_TCB_LOCK_ASSERT(stcb);
1479		if (sctp_expand_mapping_array(asoc, gap)) {
1480			/* Can't expand, drop it */
1481			return (0);
1482		}
1483	}
1484	if (SCTP_TSN_GT(tsn, *high_tsn)) {
1485		*high_tsn = tsn;
1486	}
1487	/* See if we have received this one already */
1488	if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap) ||
1489	    SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, gap)) {
1490		SCTP_STAT_INCR(sctps_recvdupdata);
1491		if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) {
1492			/* Record a dup for the next outbound sack */
1493			asoc->dup_tsns[asoc->numduptsns] = tsn;
1494			asoc->numduptsns++;
1495		}
1496		asoc->send_sack = 1;
1497		return (0);
1498	}
1499	/*
1500	 * Check to see about the GONE flag, duplicates would cause a sack
1501	 * to be sent up above
1502	 */
1503	if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) ||
1504	    (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
1505	    (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET))
1506	    ) {
1507		/*
1508		 * wait a minute, this guy is gone, there is no longer a
1509		 * receiver. Send peer an ABORT!
1510		 */
1511		struct mbuf *op_err;
1512
1513		op_err = sctp_generate_invmanparam(SCTP_CAUSE_OUT_OF_RESC);
1514		sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, SCTP_SO_NOT_LOCKED);
1515		*abort_flag = 1;
1516		return (0);
1517	}
1518	/*
1519	 * Now before going further we see if there is room. If NOT then we
1520	 * MAY let one through only IF this TSN is the one we are waiting
1521	 * for on a partial delivery API.
1522	 */
1523
1524	/* now do the tests */
1525	if (((asoc->cnt_on_all_streams +
1526	    asoc->cnt_on_reasm_queue +
1527	    asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) ||
1528	    (((int)asoc->my_rwnd) <= 0)) {
1529		/*
1530		 * When we have NO room in the rwnd we check to make sure
1531		 * the reader is doing its job...
1532		 */
1533		if (stcb->sctp_socket->so_rcv.sb_cc) {
1534			/* some to read, wake-up */
1535#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
1536			struct socket *so;
1537
1538			so = SCTP_INP_SO(stcb->sctp_ep);
1539			atomic_add_int(&stcb->asoc.refcnt, 1);
1540			SCTP_TCB_UNLOCK(stcb);
1541			SCTP_SOCKET_LOCK(so, 1);
1542			SCTP_TCB_LOCK(stcb);
1543			atomic_subtract_int(&stcb->asoc.refcnt, 1);
1544			if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) {
1545				/* assoc was freed while we were unlocked */
1546				SCTP_SOCKET_UNLOCK(so, 1);
1547				return (0);
1548			}
1549#endif
1550			sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket);
1551#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
1552			SCTP_SOCKET_UNLOCK(so, 1);
1553#endif
1554		}
1555		/* now is it in the mapping array of what we have accepted? */
1556		if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map) &&
1557		    SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
1558			/* Nope not in the valid range dump it */
1559			sctp_set_rwnd(stcb, asoc);
1560			if ((asoc->cnt_on_all_streams +
1561			    asoc->cnt_on_reasm_queue +
1562			    asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) {
1563				SCTP_STAT_INCR(sctps_datadropchklmt);
1564			} else {
1565				SCTP_STAT_INCR(sctps_datadroprwnd);
1566			}
1567			*break_flag = 1;
1568			return (0);
1569		}
1570	}
1571	strmno = ntohs(ch->dp.stream_id);
1572	if (strmno >= asoc->streamincnt) {
1573		struct sctp_paramhdr *phdr;
1574		struct mbuf *mb;
1575
1576		mb = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) * 2),
1577		    0, M_NOWAIT, 1, MT_DATA);
1578		if (mb != NULL) {
1579			/* add some space up front so prepend will work well */
1580			SCTP_BUF_RESV_UF(mb, sizeof(struct sctp_chunkhdr));
1581			phdr = mtod(mb, struct sctp_paramhdr *);
1582			/*
1583			 * Error causes are just param's and this one has
1584			 * two back to back phdr, one with the error type
1585			 * and size, the other with the streamid and a rsvd
1586			 */
1587			SCTP_BUF_LEN(mb) = (sizeof(struct sctp_paramhdr) * 2);
1588			phdr->param_type = htons(SCTP_CAUSE_INVALID_STREAM);
1589			phdr->param_length =
1590			    htons(sizeof(struct sctp_paramhdr) * 2);
1591			phdr++;
1592			/* We insert the stream in the type field */
1593			phdr->param_type = ch->dp.stream_id;
1594			/* And set the length to 0 for the rsvd field */
1595			phdr->param_length = 0;
1596			sctp_queue_op_err(stcb, mb);
1597		}
1598		SCTP_STAT_INCR(sctps_badsid);
1599		SCTP_TCB_LOCK_ASSERT(stcb);
1600		SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap);
1601		if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
1602			asoc->highest_tsn_inside_nr_map = tsn;
1603		}
1604		if (tsn == (asoc->cumulative_tsn + 1)) {
1605			/* Update cum-ack */
1606			asoc->cumulative_tsn = tsn;
1607		}
1608		return (0);
1609	}
1610	/*
1611	 * Before we continue lets validate that we are not being fooled by
1612	 * an evil attacker. We can only have 4k chunks based on our TSN
1613	 * spread allowed by the mapping array 512 * 8 bits, so there is no
1614	 * way our stream sequence numbers could have wrapped. We of course
1615	 * only validate the FIRST fragment so the bit must be set.
1616	 */
1617	strmseq = ntohs(ch->dp.stream_sequence);
1618#ifdef SCTP_ASOCLOG_OF_TSNS
1619	SCTP_TCB_LOCK_ASSERT(stcb);
1620	if (asoc->tsn_in_at >= SCTP_TSN_LOG_SIZE) {
1621		asoc->tsn_in_at = 0;
1622		asoc->tsn_in_wrapped = 1;
1623	}
1624	asoc->in_tsnlog[asoc->tsn_in_at].tsn = tsn;
1625	asoc->in_tsnlog[asoc->tsn_in_at].strm = strmno;
1626	asoc->in_tsnlog[asoc->tsn_in_at].seq = strmseq;
1627	asoc->in_tsnlog[asoc->tsn_in_at].sz = chk_length;
1628	asoc->in_tsnlog[asoc->tsn_in_at].flgs = chunk_flags;
1629	asoc->in_tsnlog[asoc->tsn_in_at].stcb = (void *)stcb;
1630	asoc->in_tsnlog[asoc->tsn_in_at].in_pos = asoc->tsn_in_at;
1631	asoc->in_tsnlog[asoc->tsn_in_at].in_out = 1;
1632	asoc->tsn_in_at++;
1633#endif
1634	if ((chunk_flags & SCTP_DATA_FIRST_FRAG) &&
1635	    (TAILQ_EMPTY(&asoc->resetHead)) &&
1636	    (chunk_flags & SCTP_DATA_UNORDERED) == 0 &&
1637	    SCTP_SSN_GE(asoc->strmin[strmno].last_sequence_delivered, strmseq)) {
1638		/* The incoming sseq is behind where we last delivered? */
1639		SCTPDBG(SCTP_DEBUG_INDATA1, "EVIL/Broken-Dup S-SEQ:%d delivered:%d from peer, Abort!\n",
1640		    strmseq, asoc->strmin[strmno].last_sequence_delivered);
1641		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1642		    0, M_NOWAIT, 1, MT_DATA);
1643		if (oper) {
1644			struct sctp_paramhdr *ph;
1645			uint32_t *ippp;
1646
1647			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1648			    (3 * sizeof(uint32_t));
1649			ph = mtod(oper, struct sctp_paramhdr *);
1650			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1651			ph->param_length = htons(SCTP_BUF_LEN(oper));
1652			ippp = (uint32_t *) (ph + 1);
1653			*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_14);
1654			ippp++;
1655			*ippp = tsn;
1656			ippp++;
1657			*ippp = ((strmno << 16) | strmseq);
1658
1659		}
1660		stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_14;
1661		sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1662		*abort_flag = 1;
1663		return (0);
1664	}
1665	/************************************
1666	 * From here down we may find ch-> invalid
1667	 * so its a good idea NOT to use it.
1668	 *************************************/
1669
1670	the_len = (chk_length - sizeof(struct sctp_data_chunk));
1671	if (last_chunk == 0) {
1672		dmbuf = SCTP_M_COPYM(*m,
1673		    (offset + sizeof(struct sctp_data_chunk)),
1674		    the_len, M_NOWAIT);
1675#ifdef SCTP_MBUF_LOGGING
1676		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) {
1677			struct mbuf *mat;
1678
1679			for (mat = dmbuf; mat; mat = SCTP_BUF_NEXT(mat)) {
1680				if (SCTP_BUF_IS_EXTENDED(mat)) {
1681					sctp_log_mb(mat, SCTP_MBUF_ICOPY);
1682				}
1683			}
1684		}
1685#endif
1686	} else {
1687		/* We can steal the last chunk */
1688		int l_len;
1689
1690		dmbuf = *m;
1691		/* lop off the top part */
1692		m_adj(dmbuf, (offset + sizeof(struct sctp_data_chunk)));
1693		if (SCTP_BUF_NEXT(dmbuf) == NULL) {
1694			l_len = SCTP_BUF_LEN(dmbuf);
1695		} else {
1696			/*
1697			 * need to count up the size hopefully does not hit
1698			 * this to often :-0
1699			 */
1700			struct mbuf *lat;
1701
1702			l_len = 0;
1703			for (lat = dmbuf; lat; lat = SCTP_BUF_NEXT(lat)) {
1704				l_len += SCTP_BUF_LEN(lat);
1705			}
1706		}
1707		if (l_len > the_len) {
1708			/* Trim the end round bytes off  too */
1709			m_adj(dmbuf, -(l_len - the_len));
1710		}
1711	}
1712	if (dmbuf == NULL) {
1713		SCTP_STAT_INCR(sctps_nomem);
1714		return (0);
1715	}
1716	if ((chunk_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG &&
1717	    asoc->fragmented_delivery_inprogress == 0 &&
1718	    TAILQ_EMPTY(&asoc->resetHead) &&
1719	    ((ordered == 0) ||
1720	    ((uint16_t) (asoc->strmin[strmno].last_sequence_delivered + 1) == strmseq &&
1721	    TAILQ_EMPTY(&asoc->strmin[strmno].inqueue)))) {
1722		/* Candidate for express delivery */
1723		/*
1724		 * Its not fragmented, No PD-API is up, Nothing in the
1725		 * delivery queue, Its un-ordered OR ordered and the next to
1726		 * deliver AND nothing else is stuck on the stream queue,
1727		 * And there is room for it in the socket buffer. Lets just
1728		 * stuff it up the buffer....
1729		 */
1730
1731		/* It would be nice to avoid this copy if we could :< */
1732		sctp_alloc_a_readq(stcb, control);
1733		sctp_build_readq_entry_mac(control, stcb, asoc->context, net, tsn,
1734		    protocol_id,
1735		    strmno, strmseq,
1736		    chunk_flags,
1737		    dmbuf);
1738		if (control == NULL) {
1739			goto failed_express_del;
1740		}
1741		SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap);
1742		if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
1743			asoc->highest_tsn_inside_nr_map = tsn;
1744		}
1745		sctp_add_to_readq(stcb->sctp_ep, stcb,
1746		    control, &stcb->sctp_socket->so_rcv,
1747		    1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED);
1748
1749		if ((chunk_flags & SCTP_DATA_UNORDERED) == 0) {
1750			/* for ordered, bump what we delivered */
1751			asoc->strmin[strmno].last_sequence_delivered++;
1752		}
1753		SCTP_STAT_INCR(sctps_recvexpress);
1754		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
1755			sctp_log_strm_del_alt(stcb, tsn, strmseq, strmno,
1756			    SCTP_STR_LOG_FROM_EXPRS_DEL);
1757		}
1758		control = NULL;
1759
1760		goto finish_express_del;
1761	}
1762failed_express_del:
1763	/* If we reach here this is a new chunk */
1764	chk = NULL;
1765	control = NULL;
1766	/* Express for fragmented delivery? */
1767	if ((asoc->fragmented_delivery_inprogress) &&
1768	    (stcb->asoc.control_pdapi) &&
1769	    (asoc->str_of_pdapi == strmno) &&
1770	    (asoc->ssn_of_pdapi == strmseq)
1771	    ) {
1772		control = stcb->asoc.control_pdapi;
1773		if ((chunk_flags & SCTP_DATA_FIRST_FRAG) == SCTP_DATA_FIRST_FRAG) {
1774			/* Can't be another first? */
1775			goto failed_pdapi_express_del;
1776		}
1777		if (tsn == (control->sinfo_tsn + 1)) {
1778			/* Yep, we can add it on */
1779			int end = 0;
1780
1781			if (chunk_flags & SCTP_DATA_LAST_FRAG) {
1782				end = 1;
1783			}
1784			if (sctp_append_to_readq(stcb->sctp_ep, stcb, control, dmbuf, end,
1785			    tsn,
1786			    &stcb->sctp_socket->so_rcv)) {
1787				SCTP_PRINTF("Append fails end:%d\n", end);
1788				goto failed_pdapi_express_del;
1789			}
1790			SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap);
1791			if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
1792				asoc->highest_tsn_inside_nr_map = tsn;
1793			}
1794			SCTP_STAT_INCR(sctps_recvexpressm);
1795			asoc->tsn_last_delivered = tsn;
1796			asoc->fragment_flags = chunk_flags;
1797			asoc->tsn_of_pdapi_last_delivered = tsn;
1798			asoc->last_flags_delivered = chunk_flags;
1799			asoc->last_strm_seq_delivered = strmseq;
1800			asoc->last_strm_no_delivered = strmno;
1801			if (end) {
1802				/* clean up the flags and such */
1803				asoc->fragmented_delivery_inprogress = 0;
1804				if ((chunk_flags & SCTP_DATA_UNORDERED) == 0) {
1805					asoc->strmin[strmno].last_sequence_delivered++;
1806				}
1807				stcb->asoc.control_pdapi = NULL;
1808				if (TAILQ_EMPTY(&asoc->reasmqueue) == 0) {
1809					/*
1810					 * There could be another message
1811					 * ready
1812					 */
1813					need_reasm_check = 1;
1814				}
1815			}
1816			control = NULL;
1817			goto finish_express_del;
1818		}
1819	}
1820failed_pdapi_express_del:
1821	control = NULL;
1822	if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
1823		SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap);
1824		if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) {
1825			asoc->highest_tsn_inside_nr_map = tsn;
1826		}
1827	} else {
1828		SCTP_SET_TSN_PRESENT(asoc->mapping_array, gap);
1829		if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map)) {
1830			asoc->highest_tsn_inside_map = tsn;
1831		}
1832	}
1833	if ((chunk_flags & SCTP_DATA_NOT_FRAG) != SCTP_DATA_NOT_FRAG) {
1834		sctp_alloc_a_chunk(stcb, chk);
1835		if (chk == NULL) {
1836			/* No memory so we drop the chunk */
1837			SCTP_STAT_INCR(sctps_nomem);
1838			if (last_chunk == 0) {
1839				/* we copied it, free the copy */
1840				sctp_m_freem(dmbuf);
1841			}
1842			return (0);
1843		}
1844		chk->rec.data.TSN_seq = tsn;
1845		chk->no_fr_allowed = 0;
1846		chk->rec.data.stream_seq = strmseq;
1847		chk->rec.data.stream_number = strmno;
1848		chk->rec.data.payloadtype = protocol_id;
1849		chk->rec.data.context = stcb->asoc.context;
1850		chk->rec.data.doing_fast_retransmit = 0;
1851		chk->rec.data.rcv_flags = chunk_flags;
1852		chk->asoc = asoc;
1853		chk->send_size = the_len;
1854		chk->whoTo = net;
1855		atomic_add_int(&net->ref_count, 1);
1856		chk->data = dmbuf;
1857	} else {
1858		sctp_alloc_a_readq(stcb, control);
1859		sctp_build_readq_entry_mac(control, stcb, asoc->context, net, tsn,
1860		    protocol_id,
1861		    strmno, strmseq,
1862		    chunk_flags,
1863		    dmbuf);
1864		if (control == NULL) {
1865			/* No memory so we drop the chunk */
1866			SCTP_STAT_INCR(sctps_nomem);
1867			if (last_chunk == 0) {
1868				/* we copied it, free the copy */
1869				sctp_m_freem(dmbuf);
1870			}
1871			return (0);
1872		}
1873		control->length = the_len;
1874	}
1875
1876	/* Mark it as received */
1877	/* Now queue it where it belongs */
1878	if (control != NULL) {
1879		/* First a sanity check */
1880		if (asoc->fragmented_delivery_inprogress) {
1881			/*
1882			 * Ok, we have a fragmented delivery in progress if
1883			 * this chunk is next to deliver OR belongs in our
1884			 * view to the reassembly, the peer is evil or
1885			 * broken.
1886			 */
1887			uint32_t estimate_tsn;
1888
1889			estimate_tsn = asoc->tsn_last_delivered + 1;
1890			if (TAILQ_EMPTY(&asoc->reasmqueue) &&
1891			    (estimate_tsn == control->sinfo_tsn)) {
1892				/* Evil/Broke peer */
1893				sctp_m_freem(control->data);
1894				control->data = NULL;
1895				if (control->whoFrom) {
1896					sctp_free_remote_addr(control->whoFrom);
1897					control->whoFrom = NULL;
1898				}
1899				sctp_free_a_readq(stcb, control);
1900				oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1901				    0, M_NOWAIT, 1, MT_DATA);
1902				if (oper) {
1903					struct sctp_paramhdr *ph;
1904					uint32_t *ippp;
1905
1906					SCTP_BUF_LEN(oper) =
1907					    sizeof(struct sctp_paramhdr) +
1908					    (3 * sizeof(uint32_t));
1909					ph = mtod(oper, struct sctp_paramhdr *);
1910					ph->param_type =
1911					    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1912					ph->param_length = htons(SCTP_BUF_LEN(oper));
1913					ippp = (uint32_t *) (ph + 1);
1914					*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_15);
1915					ippp++;
1916					*ippp = tsn;
1917					ippp++;
1918					*ippp = ((strmno << 16) | strmseq);
1919				}
1920				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_15;
1921				sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1922				*abort_flag = 1;
1923				return (0);
1924			} else {
1925				if (sctp_does_tsn_belong_to_reasm(asoc, control->sinfo_tsn)) {
1926					sctp_m_freem(control->data);
1927					control->data = NULL;
1928					if (control->whoFrom) {
1929						sctp_free_remote_addr(control->whoFrom);
1930						control->whoFrom = NULL;
1931					}
1932					sctp_free_a_readq(stcb, control);
1933
1934					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1935					    0, M_NOWAIT, 1, MT_DATA);
1936					if (oper) {
1937						struct sctp_paramhdr *ph;
1938						uint32_t *ippp;
1939
1940						SCTP_BUF_LEN(oper) =
1941						    sizeof(struct sctp_paramhdr) +
1942						    (3 * sizeof(uint32_t));
1943						ph = mtod(oper,
1944						    struct sctp_paramhdr *);
1945						ph->param_type =
1946						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1947						ph->param_length =
1948						    htons(SCTP_BUF_LEN(oper));
1949						ippp = (uint32_t *) (ph + 1);
1950						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_16);
1951						ippp++;
1952						*ippp = tsn;
1953						ippp++;
1954						*ippp = ((strmno << 16) | strmseq);
1955					}
1956					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_16;
1957					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
1958					*abort_flag = 1;
1959					return (0);
1960				}
1961			}
1962		} else {
1963			/* No PDAPI running */
1964			if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
1965				/*
1966				 * Reassembly queue is NOT empty validate
1967				 * that this tsn does not need to be in
1968				 * reasembly queue. If it does then our peer
1969				 * is broken or evil.
1970				 */
1971				if (sctp_does_tsn_belong_to_reasm(asoc, control->sinfo_tsn)) {
1972					sctp_m_freem(control->data);
1973					control->data = NULL;
1974					if (control->whoFrom) {
1975						sctp_free_remote_addr(control->whoFrom);
1976						control->whoFrom = NULL;
1977					}
1978					sctp_free_a_readq(stcb, control);
1979					oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
1980					    0, M_NOWAIT, 1, MT_DATA);
1981					if (oper) {
1982						struct sctp_paramhdr *ph;
1983						uint32_t *ippp;
1984
1985						SCTP_BUF_LEN(oper) =
1986						    sizeof(struct sctp_paramhdr) +
1987						    (3 * sizeof(uint32_t));
1988						ph = mtod(oper,
1989						    struct sctp_paramhdr *);
1990						ph->param_type =
1991						    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1992						ph->param_length =
1993						    htons(SCTP_BUF_LEN(oper));
1994						ippp = (uint32_t *) (ph + 1);
1995						*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_17);
1996						ippp++;
1997						*ippp = tsn;
1998						ippp++;
1999						*ippp = ((strmno << 16) | strmseq);
2000					}
2001					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_17;
2002					sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
2003					*abort_flag = 1;
2004					return (0);
2005				}
2006			}
2007		}
2008		/* ok, if we reach here we have passed the sanity checks */
2009		if (chunk_flags & SCTP_DATA_UNORDERED) {
2010			/* queue directly into socket buffer */
2011			sctp_mark_non_revokable(asoc, control->sinfo_tsn);
2012			sctp_add_to_readq(stcb->sctp_ep, stcb,
2013			    control,
2014			    &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED);
2015		} else {
2016			/*
2017			 * Special check for when streams are resetting. We
2018			 * could be more smart about this and check the
2019			 * actual stream to see if it is not being reset..
2020			 * that way we would not create a HOLB when amongst
2021			 * streams being reset and those not being reset.
2022			 *
2023			 * We take complete messages that have a stream reset
2024			 * intervening (aka the TSN is after where our
2025			 * cum-ack needs to be) off and put them on a
2026			 * pending_reply_queue. The reassembly ones we do
2027			 * not have to worry about since they are all sorted
2028			 * and proceessed by TSN order. It is only the
2029			 * singletons I must worry about.
2030			 */
2031			if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) &&
2032			    SCTP_TSN_GT(tsn, liste->tsn)) {
2033				/*
2034				 * yep its past where we need to reset... go
2035				 * ahead and queue it.
2036				 */
2037				if (TAILQ_EMPTY(&asoc->pending_reply_queue)) {
2038					/* first one on */
2039					TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next);
2040				} else {
2041					struct sctp_queued_to_read *ctlOn,
2042					                   *nctlOn;
2043					unsigned char inserted = 0;
2044
2045					TAILQ_FOREACH_SAFE(ctlOn, &asoc->pending_reply_queue, next, nctlOn) {
2046						if (SCTP_TSN_GT(control->sinfo_tsn, ctlOn->sinfo_tsn)) {
2047							continue;
2048						} else {
2049							/* found it */
2050							TAILQ_INSERT_BEFORE(ctlOn, control, next);
2051							inserted = 1;
2052							break;
2053						}
2054					}
2055					if (inserted == 0) {
2056						/*
2057						 * must be put at end, use
2058						 * prevP (all setup from
2059						 * loop) to setup nextP.
2060						 */
2061						TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next);
2062					}
2063				}
2064			} else {
2065				sctp_queue_data_to_stream(stcb, asoc, control, abort_flag);
2066				if (*abort_flag) {
2067					return (0);
2068				}
2069			}
2070		}
2071	} else {
2072		/* Into the re-assembly queue */
2073		sctp_queue_data_for_reasm(stcb, asoc, chk, abort_flag);
2074		if (*abort_flag) {
2075			/*
2076			 * the assoc is now gone and chk was put onto the
2077			 * reasm queue, which has all been freed.
2078			 */
2079			*m = NULL;
2080			return (0);
2081		}
2082	}
2083finish_express_del:
2084	if (tsn == (asoc->cumulative_tsn + 1)) {
2085		/* Update cum-ack */
2086		asoc->cumulative_tsn = tsn;
2087	}
2088	if (last_chunk) {
2089		*m = NULL;
2090	}
2091	if (ordered) {
2092		SCTP_STAT_INCR_COUNTER64(sctps_inorderchunks);
2093	} else {
2094		SCTP_STAT_INCR_COUNTER64(sctps_inunorderchunks);
2095	}
2096	SCTP_STAT_INCR(sctps_recvdata);
2097	/* Set it present please */
2098	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) {
2099		sctp_log_strm_del_alt(stcb, tsn, strmseq, strmno, SCTP_STR_LOG_FROM_MARK_TSN);
2100	}
2101	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
2102		sctp_log_map(asoc->mapping_array_base_tsn, asoc->cumulative_tsn,
2103		    asoc->highest_tsn_inside_map, SCTP_MAP_PREPARE_SLIDE);
2104	}
2105	/* check the special flag for stream resets */
2106	if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) &&
2107	    SCTP_TSN_GE(asoc->cumulative_tsn, liste->tsn)) {
2108		/*
2109		 * we have finished working through the backlogged TSN's now
2110		 * time to reset streams. 1: call reset function. 2: free
2111		 * pending_reply space 3: distribute any chunks in
2112		 * pending_reply_queue.
2113		 */
2114		struct sctp_queued_to_read *ctl, *nctl;
2115
2116		sctp_reset_in_stream(stcb, liste->number_entries, liste->list_of_streams);
2117		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
2118		SCTP_FREE(liste, SCTP_M_STRESET);
2119		/* sa_ignore FREED_MEMORY */
2120		liste = TAILQ_FIRST(&asoc->resetHead);
2121		if (TAILQ_EMPTY(&asoc->resetHead)) {
2122			/* All can be removed */
2123			TAILQ_FOREACH_SAFE(ctl, &asoc->pending_reply_queue, next, nctl) {
2124				TAILQ_REMOVE(&asoc->pending_reply_queue, ctl, next);
2125				sctp_queue_data_to_stream(stcb, asoc, ctl, abort_flag);
2126				if (*abort_flag) {
2127					return (0);
2128				}
2129			}
2130		} else {
2131			TAILQ_FOREACH_SAFE(ctl, &asoc->pending_reply_queue, next, nctl) {
2132				if (SCTP_TSN_GT(ctl->sinfo_tsn, liste->tsn)) {
2133					break;
2134				}
2135				/*
2136				 * if ctl->sinfo_tsn is <= liste->tsn we can
2137				 * process it which is the NOT of
2138				 * ctl->sinfo_tsn > liste->tsn
2139				 */
2140				TAILQ_REMOVE(&asoc->pending_reply_queue, ctl, next);
2141				sctp_queue_data_to_stream(stcb, asoc, ctl, abort_flag);
2142				if (*abort_flag) {
2143					return (0);
2144				}
2145			}
2146		}
2147		/*
2148		 * Now service re-assembly to pick up anything that has been
2149		 * held on reassembly queue?
2150		 */
2151		sctp_deliver_reasm_check(stcb, asoc);
2152		need_reasm_check = 0;
2153	}
2154	if (need_reasm_check) {
2155		/* Another one waits ? */
2156		sctp_deliver_reasm_check(stcb, asoc);
2157	}
2158	return (1);
2159}
2160
2161int8_t sctp_map_lookup_tab[256] = {
2162	0, 1, 0, 2, 0, 1, 0, 3,
2163	0, 1, 0, 2, 0, 1, 0, 4,
2164	0, 1, 0, 2, 0, 1, 0, 3,
2165	0, 1, 0, 2, 0, 1, 0, 5,
2166	0, 1, 0, 2, 0, 1, 0, 3,
2167	0, 1, 0, 2, 0, 1, 0, 4,
2168	0, 1, 0, 2, 0, 1, 0, 3,
2169	0, 1, 0, 2, 0, 1, 0, 6,
2170	0, 1, 0, 2, 0, 1, 0, 3,
2171	0, 1, 0, 2, 0, 1, 0, 4,
2172	0, 1, 0, 2, 0, 1, 0, 3,
2173	0, 1, 0, 2, 0, 1, 0, 5,
2174	0, 1, 0, 2, 0, 1, 0, 3,
2175	0, 1, 0, 2, 0, 1, 0, 4,
2176	0, 1, 0, 2, 0, 1, 0, 3,
2177	0, 1, 0, 2, 0, 1, 0, 7,
2178	0, 1, 0, 2, 0, 1, 0, 3,
2179	0, 1, 0, 2, 0, 1, 0, 4,
2180	0, 1, 0, 2, 0, 1, 0, 3,
2181	0, 1, 0, 2, 0, 1, 0, 5,
2182	0, 1, 0, 2, 0, 1, 0, 3,
2183	0, 1, 0, 2, 0, 1, 0, 4,
2184	0, 1, 0, 2, 0, 1, 0, 3,
2185	0, 1, 0, 2, 0, 1, 0, 6,
2186	0, 1, 0, 2, 0, 1, 0, 3,
2187	0, 1, 0, 2, 0, 1, 0, 4,
2188	0, 1, 0, 2, 0, 1, 0, 3,
2189	0, 1, 0, 2, 0, 1, 0, 5,
2190	0, 1, 0, 2, 0, 1, 0, 3,
2191	0, 1, 0, 2, 0, 1, 0, 4,
2192	0, 1, 0, 2, 0, 1, 0, 3,
2193	0, 1, 0, 2, 0, 1, 0, 8
2194};
2195
2196
2197void
2198sctp_slide_mapping_arrays(struct sctp_tcb *stcb)
2199{
2200	/*
2201	 * Now we also need to check the mapping array in a couple of ways.
2202	 * 1) Did we move the cum-ack point?
2203	 *
2204	 * When you first glance at this you might think that all entries that
2205	 * make up the postion of the cum-ack would be in the nr-mapping
2206	 * array only.. i.e. things up to the cum-ack are always
2207	 * deliverable. Thats true with one exception, when its a fragmented
2208	 * message we may not deliver the data until some threshold (or all
2209	 * of it) is in place. So we must OR the nr_mapping_array and
2210	 * mapping_array to get a true picture of the cum-ack.
2211	 */
2212	struct sctp_association *asoc;
2213	int at;
2214	uint8_t val;
2215	int slide_from, slide_end, lgap, distance;
2216	uint32_t old_cumack, old_base, old_highest, highest_tsn;
2217
2218	asoc = &stcb->asoc;
2219
2220	old_cumack = asoc->cumulative_tsn;
2221	old_base = asoc->mapping_array_base_tsn;
2222	old_highest = asoc->highest_tsn_inside_map;
2223	/*
2224	 * We could probably improve this a small bit by calculating the
2225	 * offset of the current cum-ack as the starting point.
2226	 */
2227	at = 0;
2228	for (slide_from = 0; slide_from < stcb->asoc.mapping_array_size; slide_from++) {
2229		val = asoc->nr_mapping_array[slide_from] | asoc->mapping_array[slide_from];
2230		if (val == 0xff) {
2231			at += 8;
2232		} else {
2233			/* there is a 0 bit */
2234			at += sctp_map_lookup_tab[val];
2235			break;
2236		}
2237	}
2238	asoc->cumulative_tsn = asoc->mapping_array_base_tsn + (at - 1);
2239
2240	if (SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_map) &&
2241	    SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_nr_map)) {
2242#ifdef INVARIANTS
2243		panic("huh, cumack 0x%x greater than high-tsn 0x%x in map",
2244		    asoc->cumulative_tsn, asoc->highest_tsn_inside_map);
2245#else
2246		SCTP_PRINTF("huh, cumack 0x%x greater than high-tsn 0x%x in map - should panic?\n",
2247		    asoc->cumulative_tsn, asoc->highest_tsn_inside_map);
2248		sctp_print_mapping_array(asoc);
2249		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
2250			sctp_log_map(0, 6, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT);
2251		}
2252		asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
2253		asoc->highest_tsn_inside_nr_map = asoc->cumulative_tsn;
2254#endif
2255	}
2256	if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) {
2257		highest_tsn = asoc->highest_tsn_inside_nr_map;
2258	} else {
2259		highest_tsn = asoc->highest_tsn_inside_map;
2260	}
2261	if ((asoc->cumulative_tsn == highest_tsn) && (at >= 8)) {
2262		/* The complete array was completed by a single FR */
2263		/* highest becomes the cum-ack */
2264		int clr;
2265
2266#ifdef INVARIANTS
2267		unsigned int i;
2268
2269#endif
2270
2271		/* clear the array */
2272		clr = ((at + 7) >> 3);
2273		if (clr > asoc->mapping_array_size) {
2274			clr = asoc->mapping_array_size;
2275		}
2276		memset(asoc->mapping_array, 0, clr);
2277		memset(asoc->nr_mapping_array, 0, clr);
2278#ifdef INVARIANTS
2279		for (i = 0; i < asoc->mapping_array_size; i++) {
2280			if ((asoc->mapping_array[i]) || (asoc->nr_mapping_array[i])) {
2281				SCTP_PRINTF("Error Mapping array's not clean at clear\n");
2282				sctp_print_mapping_array(asoc);
2283			}
2284		}
2285#endif
2286		asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
2287		asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
2288	} else if (at >= 8) {
2289		/* we can slide the mapping array down */
2290		/* slide_from holds where we hit the first NON 0xff byte */
2291
2292		/*
2293		 * now calculate the ceiling of the move using our highest
2294		 * TSN value
2295		 */
2296		SCTP_CALC_TSN_TO_GAP(lgap, highest_tsn, asoc->mapping_array_base_tsn);
2297		slide_end = (lgap >> 3);
2298		if (slide_end < slide_from) {
2299			sctp_print_mapping_array(asoc);
2300#ifdef INVARIANTS
2301			panic("impossible slide");
2302#else
2303			SCTP_PRINTF("impossible slide lgap:%x slide_end:%x slide_from:%x? at:%d\n",
2304			    lgap, slide_end, slide_from, at);
2305			return;
2306#endif
2307		}
2308		if (slide_end > asoc->mapping_array_size) {
2309#ifdef INVARIANTS
2310			panic("would overrun buffer");
2311#else
2312			SCTP_PRINTF("Gak, would have overrun map end:%d slide_end:%d\n",
2313			    asoc->mapping_array_size, slide_end);
2314			slide_end = asoc->mapping_array_size;
2315#endif
2316		}
2317		distance = (slide_end - slide_from) + 1;
2318		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
2319			sctp_log_map(old_base, old_cumack, old_highest,
2320			    SCTP_MAP_PREPARE_SLIDE);
2321			sctp_log_map((uint32_t) slide_from, (uint32_t) slide_end,
2322			    (uint32_t) lgap, SCTP_MAP_SLIDE_FROM);
2323		}
2324		if (distance + slide_from > asoc->mapping_array_size ||
2325		    distance < 0) {
2326			/*
2327			 * Here we do NOT slide forward the array so that
2328			 * hopefully when more data comes in to fill it up
2329			 * we will be able to slide it forward. Really I
2330			 * don't think this should happen :-0
2331			 */
2332
2333			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
2334				sctp_log_map((uint32_t) distance, (uint32_t) slide_from,
2335				    (uint32_t) asoc->mapping_array_size,
2336				    SCTP_MAP_SLIDE_NONE);
2337			}
2338		} else {
2339			int ii;
2340
2341			for (ii = 0; ii < distance; ii++) {
2342				asoc->mapping_array[ii] = asoc->mapping_array[slide_from + ii];
2343				asoc->nr_mapping_array[ii] = asoc->nr_mapping_array[slide_from + ii];
2344
2345			}
2346			for (ii = distance; ii < asoc->mapping_array_size; ii++) {
2347				asoc->mapping_array[ii] = 0;
2348				asoc->nr_mapping_array[ii] = 0;
2349			}
2350			if (asoc->highest_tsn_inside_map + 1 == asoc->mapping_array_base_tsn) {
2351				asoc->highest_tsn_inside_map += (slide_from << 3);
2352			}
2353			if (asoc->highest_tsn_inside_nr_map + 1 == asoc->mapping_array_base_tsn) {
2354				asoc->highest_tsn_inside_nr_map += (slide_from << 3);
2355			}
2356			asoc->mapping_array_base_tsn += (slide_from << 3);
2357			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
2358				sctp_log_map(asoc->mapping_array_base_tsn,
2359				    asoc->cumulative_tsn, asoc->highest_tsn_inside_map,
2360				    SCTP_MAP_SLIDE_RESULT);
2361			}
2362		}
2363	}
2364}
2365
2366void
2367sctp_sack_check(struct sctp_tcb *stcb, int was_a_gap)
2368{
2369	struct sctp_association *asoc;
2370	uint32_t highest_tsn;
2371
2372	asoc = &stcb->asoc;
2373	if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) {
2374		highest_tsn = asoc->highest_tsn_inside_nr_map;
2375	} else {
2376		highest_tsn = asoc->highest_tsn_inside_map;
2377	}
2378
2379	/*
2380	 * Now we need to see if we need to queue a sack or just start the
2381	 * timer (if allowed).
2382	 */
2383	if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) {
2384		/*
2385		 * Ok special case, in SHUTDOWN-SENT case. here we maker
2386		 * sure SACK timer is off and instead send a SHUTDOWN and a
2387		 * SACK
2388		 */
2389		if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
2390			sctp_timer_stop(SCTP_TIMER_TYPE_RECV,
2391			    stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_INDATA + SCTP_LOC_18);
2392		}
2393		sctp_send_shutdown(stcb,
2394		    ((stcb->asoc.alternate) ? stcb->asoc.alternate : stcb->asoc.primary_destination));
2395		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
2396	} else {
2397		int is_a_gap;
2398
2399		/* is there a gap now ? */
2400		is_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn);
2401
2402		/*
2403		 * CMT DAC algorithm: increase number of packets received
2404		 * since last ack
2405		 */
2406		stcb->asoc.cmt_dac_pkts_rcvd++;
2407
2408		if ((stcb->asoc.send_sack == 1) ||	/* We need to send a
2409							 * SACK */
2410		    ((was_a_gap) && (is_a_gap == 0)) ||	/* was a gap, but no
2411							 * longer is one */
2412		    (stcb->asoc.numduptsns) ||	/* we have dup's */
2413		    (is_a_gap) ||	/* is still a gap */
2414		    (stcb->asoc.delayed_ack == 0) ||	/* Delayed sack disabled */
2415		    (stcb->asoc.data_pkts_seen >= stcb->asoc.sack_freq)	/* hit limit of pkts */
2416		    ) {
2417
2418			if ((stcb->asoc.sctp_cmt_on_off > 0) &&
2419			    (SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) &&
2420			    (stcb->asoc.send_sack == 0) &&
2421			    (stcb->asoc.numduptsns == 0) &&
2422			    (stcb->asoc.delayed_ack) &&
2423			    (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer))) {
2424
2425				/*
2426				 * CMT DAC algorithm: With CMT, delay acks
2427				 * even in the face of
2428				 *
2429				 * reordering. Therefore, if acks that do not
2430				 * have to be sent because of the above
2431				 * reasons, will be delayed. That is, acks
2432				 * that would have been sent due to gap
2433				 * reports will be delayed with DAC. Start
2434				 * the delayed ack timer.
2435				 */
2436				sctp_timer_start(SCTP_TIMER_TYPE_RECV,
2437				    stcb->sctp_ep, stcb, NULL);
2438			} else {
2439				/*
2440				 * Ok we must build a SACK since the timer
2441				 * is pending, we got our first packet OR
2442				 * there are gaps or duplicates.
2443				 */
2444				(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
2445				sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
2446			}
2447		} else {
2448			if (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
2449				sctp_timer_start(SCTP_TIMER_TYPE_RECV,
2450				    stcb->sctp_ep, stcb, NULL);
2451			}
2452		}
2453	}
2454}
2455
2456void
2457sctp_service_queues(struct sctp_tcb *stcb, struct sctp_association *asoc)
2458{
2459	struct sctp_tmit_chunk *chk;
2460	uint32_t tsize, pd_point;
2461	uint16_t nxt_todel;
2462
2463	if (asoc->fragmented_delivery_inprogress) {
2464		sctp_service_reassembly(stcb, asoc);
2465	}
2466	/* Can we proceed further, i.e. the PD-API is complete */
2467	if (asoc->fragmented_delivery_inprogress) {
2468		/* no */
2469		return;
2470	}
2471	/*
2472	 * Now is there some other chunk I can deliver from the reassembly
2473	 * queue.
2474	 */
2475doit_again:
2476	chk = TAILQ_FIRST(&asoc->reasmqueue);
2477	if (chk == NULL) {
2478		asoc->size_on_reasm_queue = 0;
2479		asoc->cnt_on_reasm_queue = 0;
2480		return;
2481	}
2482	nxt_todel = asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered + 1;
2483	if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) &&
2484	    ((nxt_todel == chk->rec.data.stream_seq) ||
2485	    (chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED))) {
2486		/*
2487		 * Yep the first one is here. We setup to start reception,
2488		 * by backing down the TSN just in case we can't deliver.
2489		 */
2490
2491		/*
2492		 * Before we start though either all of the message should
2493		 * be here or the socket buffer max or nothing on the
2494		 * delivery queue and something can be delivered.
2495		 */
2496		if (stcb->sctp_socket) {
2497			pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket) >> SCTP_PARTIAL_DELIVERY_SHIFT,
2498			    stcb->sctp_ep->partial_delivery_point);
2499		} else {
2500			pd_point = stcb->sctp_ep->partial_delivery_point;
2501		}
2502		if (sctp_is_all_msg_on_reasm(asoc, &tsize) || (tsize >= pd_point)) {
2503			asoc->fragmented_delivery_inprogress = 1;
2504			asoc->tsn_last_delivered = chk->rec.data.TSN_seq - 1;
2505			asoc->str_of_pdapi = chk->rec.data.stream_number;
2506			asoc->ssn_of_pdapi = chk->rec.data.stream_seq;
2507			asoc->pdapi_ppid = chk->rec.data.payloadtype;
2508			asoc->fragment_flags = chk->rec.data.rcv_flags;
2509			sctp_service_reassembly(stcb, asoc);
2510			if (asoc->fragmented_delivery_inprogress == 0) {
2511				goto doit_again;
2512			}
2513		}
2514	}
2515}
2516
2517int
2518sctp_process_data(struct mbuf **mm, int iphlen, int *offset, int length,
2519    struct sockaddr *src, struct sockaddr *dst,
2520    struct sctphdr *sh, struct sctp_inpcb *inp,
2521    struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t * high_tsn,
2522    uint8_t use_mflowid, uint32_t mflowid,
2523    uint32_t vrf_id, uint16_t port)
2524{
2525	struct sctp_data_chunk *ch, chunk_buf;
2526	struct sctp_association *asoc;
2527	int num_chunks = 0;	/* number of control chunks processed */
2528	int stop_proc = 0;
2529	int chk_length, break_flag, last_chunk;
2530	int abort_flag = 0, was_a_gap;
2531	struct mbuf *m;
2532	uint32_t highest_tsn;
2533
2534	/* set the rwnd */
2535	sctp_set_rwnd(stcb, &stcb->asoc);
2536
2537	m = *mm;
2538	SCTP_TCB_LOCK_ASSERT(stcb);
2539	asoc = &stcb->asoc;
2540	if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) {
2541		highest_tsn = asoc->highest_tsn_inside_nr_map;
2542	} else {
2543		highest_tsn = asoc->highest_tsn_inside_map;
2544	}
2545	was_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn);
2546	/*
2547	 * setup where we got the last DATA packet from for any SACK that
2548	 * may need to go out. Don't bump the net. This is done ONLY when a
2549	 * chunk is assigned.
2550	 */
2551	asoc->last_data_chunk_from = net;
2552
2553	/*-
2554	 * Now before we proceed we must figure out if this is a wasted
2555	 * cluster... i.e. it is a small packet sent in and yet the driver
2556	 * underneath allocated a full cluster for it. If so we must copy it
2557	 * to a smaller mbuf and free up the cluster mbuf. This will help
2558	 * with cluster starvation. Note for __Panda__ we don't do this
2559	 * since it has clusters all the way down to 64 bytes.
2560	 */
2561	if (SCTP_BUF_LEN(m) < (long)MLEN && SCTP_BUF_NEXT(m) == NULL) {
2562		/* we only handle mbufs that are singletons.. not chains */
2563		m = sctp_get_mbuf_for_msg(SCTP_BUF_LEN(m), 0, M_NOWAIT, 1, MT_DATA);
2564		if (m) {
2565			/* ok lets see if we can copy the data up */
2566			caddr_t *from, *to;
2567
2568			/* get the pointers and copy */
2569			to = mtod(m, caddr_t *);
2570			from = mtod((*mm), caddr_t *);
2571			memcpy(to, from, SCTP_BUF_LEN((*mm)));
2572			/* copy the length and free up the old */
2573			SCTP_BUF_LEN(m) = SCTP_BUF_LEN((*mm));
2574			sctp_m_freem(*mm);
2575			/* sucess, back copy */
2576			*mm = m;
2577		} else {
2578			/* We are in trouble in the mbuf world .. yikes */
2579			m = *mm;
2580		}
2581	}
2582	/* get pointer to the first chunk header */
2583	ch = (struct sctp_data_chunk *)sctp_m_getptr(m, *offset,
2584	    sizeof(struct sctp_data_chunk), (uint8_t *) & chunk_buf);
2585	if (ch == NULL) {
2586		return (1);
2587	}
2588	/*
2589	 * process all DATA chunks...
2590	 */
2591	*high_tsn = asoc->cumulative_tsn;
2592	break_flag = 0;
2593	asoc->data_pkts_seen++;
2594	while (stop_proc == 0) {
2595		/* validate chunk length */
2596		chk_length = ntohs(ch->ch.chunk_length);
2597		if (length - *offset < chk_length) {
2598			/* all done, mutulated chunk */
2599			stop_proc = 1;
2600			continue;
2601		}
2602		if (ch->ch.chunk_type == SCTP_DATA) {
2603			if ((size_t)chk_length < sizeof(struct sctp_data_chunk) + 1) {
2604				/*
2605				 * Need to send an abort since we had a
2606				 * invalid data chunk.
2607				 */
2608				struct mbuf *op_err;
2609
2610				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 2 * sizeof(uint32_t)),
2611				    0, M_NOWAIT, 1, MT_DATA);
2612
2613				if (op_err) {
2614					struct sctp_paramhdr *ph;
2615					uint32_t *ippp;
2616
2617					SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr) +
2618					    (2 * sizeof(uint32_t));
2619					ph = mtod(op_err, struct sctp_paramhdr *);
2620					ph->param_type =
2621					    htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
2622					ph->param_length = htons(SCTP_BUF_LEN(op_err));
2623					ippp = (uint32_t *) (ph + 1);
2624					*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_19);
2625					ippp++;
2626					*ippp = asoc->cumulative_tsn;
2627
2628				}
2629				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_19;
2630				sctp_abort_association(inp, stcb, m, iphlen,
2631				    src, dst, sh, op_err,
2632				    use_mflowid, mflowid,
2633				    vrf_id, port);
2634				return (2);
2635			}
2636#ifdef SCTP_AUDITING_ENABLED
2637			sctp_audit_log(0xB1, 0);
2638#endif
2639			if (SCTP_SIZE32(chk_length) == (length - *offset)) {
2640				last_chunk = 1;
2641			} else {
2642				last_chunk = 0;
2643			}
2644			if (sctp_process_a_data_chunk(stcb, asoc, mm, *offset, ch,
2645			    chk_length, net, high_tsn, &abort_flag, &break_flag,
2646			    last_chunk)) {
2647				num_chunks++;
2648			}
2649			if (abort_flag)
2650				return (2);
2651
2652			if (break_flag) {
2653				/*
2654				 * Set because of out of rwnd space and no
2655				 * drop rep space left.
2656				 */
2657				stop_proc = 1;
2658				continue;
2659			}
2660		} else {
2661			/* not a data chunk in the data region */
2662			switch (ch->ch.chunk_type) {
2663			case SCTP_INITIATION:
2664			case SCTP_INITIATION_ACK:
2665			case SCTP_SELECTIVE_ACK:
2666			case SCTP_NR_SELECTIVE_ACK:
2667			case SCTP_HEARTBEAT_REQUEST:
2668			case SCTP_HEARTBEAT_ACK:
2669			case SCTP_ABORT_ASSOCIATION:
2670			case SCTP_SHUTDOWN:
2671			case SCTP_SHUTDOWN_ACK:
2672			case SCTP_OPERATION_ERROR:
2673			case SCTP_COOKIE_ECHO:
2674			case SCTP_COOKIE_ACK:
2675			case SCTP_ECN_ECHO:
2676			case SCTP_ECN_CWR:
2677			case SCTP_SHUTDOWN_COMPLETE:
2678			case SCTP_AUTHENTICATION:
2679			case SCTP_ASCONF_ACK:
2680			case SCTP_PACKET_DROPPED:
2681			case SCTP_STREAM_RESET:
2682			case SCTP_FORWARD_CUM_TSN:
2683			case SCTP_ASCONF:
2684				/*
2685				 * Now, what do we do with KNOWN chunks that
2686				 * are NOT in the right place?
2687				 *
2688				 * For now, I do nothing but ignore them. We
2689				 * may later want to add sysctl stuff to
2690				 * switch out and do either an ABORT() or
2691				 * possibly process them.
2692				 */
2693				if (SCTP_BASE_SYSCTL(sctp_strict_data_order)) {
2694					struct mbuf *op_err;
2695
2696					op_err = sctp_generate_invmanparam(SCTP_CAUSE_PROTOCOL_VIOLATION);
2697					sctp_abort_association(inp, stcb,
2698					    m, iphlen,
2699					    src, dst,
2700					    sh, op_err,
2701					    use_mflowid, mflowid,
2702					    vrf_id, port);
2703					return (2);
2704				}
2705				break;
2706			default:
2707				/* unknown chunk type, use bit rules */
2708				if (ch->ch.chunk_type & 0x40) {
2709					/* Add a error report to the queue */
2710					struct mbuf *merr;
2711					struct sctp_paramhdr *phd;
2712
2713					merr = sctp_get_mbuf_for_msg(sizeof(*phd), 0, M_NOWAIT, 1, MT_DATA);
2714					if (merr) {
2715						phd = mtod(merr, struct sctp_paramhdr *);
2716						/*
2717						 * We cheat and use param
2718						 * type since we did not
2719						 * bother to define a error
2720						 * cause struct. They are
2721						 * the same basic format
2722						 * with different names.
2723						 */
2724						phd->param_type =
2725						    htons(SCTP_CAUSE_UNRECOG_CHUNK);
2726						phd->param_length =
2727						    htons(chk_length + sizeof(*phd));
2728						SCTP_BUF_LEN(merr) = sizeof(*phd);
2729						SCTP_BUF_NEXT(merr) = SCTP_M_COPYM(m, *offset, chk_length, M_NOWAIT);
2730						if (SCTP_BUF_NEXT(merr)) {
2731							if (sctp_pad_lastmbuf(SCTP_BUF_NEXT(merr), SCTP_SIZE32(chk_length) - chk_length, NULL)) {
2732								sctp_m_freem(merr);
2733							} else {
2734								sctp_queue_op_err(stcb, merr);
2735							}
2736						} else {
2737							sctp_m_freem(merr);
2738						}
2739					}
2740				}
2741				if ((ch->ch.chunk_type & 0x80) == 0) {
2742					/* discard the rest of this packet */
2743					stop_proc = 1;
2744				}	/* else skip this bad chunk and
2745					 * continue... */
2746				break;
2747			}	/* switch of chunk type */
2748		}
2749		*offset += SCTP_SIZE32(chk_length);
2750		if ((*offset >= length) || stop_proc) {
2751			/* no more data left in the mbuf chain */
2752			stop_proc = 1;
2753			continue;
2754		}
2755		ch = (struct sctp_data_chunk *)sctp_m_getptr(m, *offset,
2756		    sizeof(struct sctp_data_chunk), (uint8_t *) & chunk_buf);
2757		if (ch == NULL) {
2758			*offset = length;
2759			stop_proc = 1;
2760			continue;
2761		}
2762	}
2763	if (break_flag) {
2764		/*
2765		 * we need to report rwnd overrun drops.
2766		 */
2767		sctp_send_packet_dropped(stcb, net, *mm, length, iphlen, 0);
2768	}
2769	if (num_chunks) {
2770		/*
2771		 * Did we get data, if so update the time for auto-close and
2772		 * give peer credit for being alive.
2773		 */
2774		SCTP_STAT_INCR(sctps_recvpktwithdata);
2775		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
2776			sctp_misc_ints(SCTP_THRESHOLD_CLEAR,
2777			    stcb->asoc.overall_error_count,
2778			    0,
2779			    SCTP_FROM_SCTP_INDATA,
2780			    __LINE__);
2781		}
2782		stcb->asoc.overall_error_count = 0;
2783		(void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_last_rcvd);
2784	}
2785	/* now service all of the reassm queue if needed */
2786	if (!(TAILQ_EMPTY(&asoc->reasmqueue)))
2787		sctp_service_queues(stcb, asoc);
2788
2789	if (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) {
2790		/* Assure that we ack right away */
2791		stcb->asoc.send_sack = 1;
2792	}
2793	/* Start a sack timer or QUEUE a SACK for sending */
2794	sctp_sack_check(stcb, was_a_gap);
2795	return (0);
2796}
2797
2798static int
2799sctp_process_segment_range(struct sctp_tcb *stcb, struct sctp_tmit_chunk **p_tp1, uint32_t last_tsn,
2800    uint16_t frag_strt, uint16_t frag_end, int nr_sacking,
2801    int *num_frs,
2802    uint32_t * biggest_newly_acked_tsn,
2803    uint32_t * this_sack_lowest_newack,
2804    int *rto_ok)
2805{
2806	struct sctp_tmit_chunk *tp1;
2807	unsigned int theTSN;
2808	int j, wake_him = 0, circled = 0;
2809
2810	/* Recover the tp1 we last saw */
2811	tp1 = *p_tp1;
2812	if (tp1 == NULL) {
2813		tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue);
2814	}
2815	for (j = frag_strt; j <= frag_end; j++) {
2816		theTSN = j + last_tsn;
2817		while (tp1) {
2818			if (tp1->rec.data.doing_fast_retransmit)
2819				(*num_frs) += 1;
2820
2821			/*-
2822			 * CMT: CUCv2 algorithm. For each TSN being
2823			 * processed from the sent queue, track the
2824			 * next expected pseudo-cumack, or
2825			 * rtx_pseudo_cumack, if required. Separate
2826			 * cumack trackers for first transmissions,
2827			 * and retransmissions.
2828			 */
2829			if ((tp1->whoTo->find_pseudo_cumack == 1) && (tp1->sent < SCTP_DATAGRAM_RESEND) &&
2830			    (tp1->snd_count == 1)) {
2831				tp1->whoTo->pseudo_cumack = tp1->rec.data.TSN_seq;
2832				tp1->whoTo->find_pseudo_cumack = 0;
2833			}
2834			if ((tp1->whoTo->find_rtx_pseudo_cumack == 1) && (tp1->sent < SCTP_DATAGRAM_RESEND) &&
2835			    (tp1->snd_count > 1)) {
2836				tp1->whoTo->rtx_pseudo_cumack = tp1->rec.data.TSN_seq;
2837				tp1->whoTo->find_rtx_pseudo_cumack = 0;
2838			}
2839			if (tp1->rec.data.TSN_seq == theTSN) {
2840				if (tp1->sent != SCTP_DATAGRAM_UNSENT) {
2841					/*-
2842					 * must be held until
2843					 * cum-ack passes
2844					 */
2845					if (tp1->sent < SCTP_DATAGRAM_RESEND) {
2846						/*-
2847						 * If it is less than RESEND, it is
2848						 * now no-longer in flight.
2849						 * Higher values may already be set
2850						 * via previous Gap Ack Blocks...
2851						 * i.e. ACKED or RESEND.
2852						 */
2853						if (SCTP_TSN_GT(tp1->rec.data.TSN_seq,
2854						    *biggest_newly_acked_tsn)) {
2855							*biggest_newly_acked_tsn = tp1->rec.data.TSN_seq;
2856						}
2857						/*-
2858						 * CMT: SFR algo (and HTNA) - set
2859						 * saw_newack to 1 for dest being
2860						 * newly acked. update
2861						 * this_sack_highest_newack if
2862						 * appropriate.
2863						 */
2864						if (tp1->rec.data.chunk_was_revoked == 0)
2865							tp1->whoTo->saw_newack = 1;
2866
2867						if (SCTP_TSN_GT(tp1->rec.data.TSN_seq,
2868						    tp1->whoTo->this_sack_highest_newack)) {
2869							tp1->whoTo->this_sack_highest_newack =
2870							    tp1->rec.data.TSN_seq;
2871						}
2872						/*-
2873						 * CMT DAC algo: also update
2874						 * this_sack_lowest_newack
2875						 */
2876						if (*this_sack_lowest_newack == 0) {
2877							if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
2878								sctp_log_sack(*this_sack_lowest_newack,
2879								    last_tsn,
2880								    tp1->rec.data.TSN_seq,
2881								    0,
2882								    0,
2883								    SCTP_LOG_TSN_ACKED);
2884							}
2885							*this_sack_lowest_newack = tp1->rec.data.TSN_seq;
2886						}
2887						/*-
2888						 * CMT: CUCv2 algorithm. If (rtx-)pseudo-cumack for corresp
2889						 * dest is being acked, then we have a new (rtx-)pseudo-cumack. Set
2890						 * new_(rtx_)pseudo_cumack to TRUE so that the cwnd for this dest can be
2891						 * updated. Also trigger search for the next expected (rtx-)pseudo-cumack.
2892						 * Separate pseudo_cumack trackers for first transmissions and
2893						 * retransmissions.
2894						 */
2895						if (tp1->rec.data.TSN_seq == tp1->whoTo->pseudo_cumack) {
2896							if (tp1->rec.data.chunk_was_revoked == 0) {
2897								tp1->whoTo->new_pseudo_cumack = 1;
2898							}
2899							tp1->whoTo->find_pseudo_cumack = 1;
2900						}
2901						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
2902							sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK);
2903						}
2904						if (tp1->rec.data.TSN_seq == tp1->whoTo->rtx_pseudo_cumack) {
2905							if (tp1->rec.data.chunk_was_revoked == 0) {
2906								tp1->whoTo->new_pseudo_cumack = 1;
2907							}
2908							tp1->whoTo->find_rtx_pseudo_cumack = 1;
2909						}
2910						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
2911							sctp_log_sack(*biggest_newly_acked_tsn,
2912							    last_tsn,
2913							    tp1->rec.data.TSN_seq,
2914							    frag_strt,
2915							    frag_end,
2916							    SCTP_LOG_TSN_ACKED);
2917						}
2918						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
2919							sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_GAP,
2920							    tp1->whoTo->flight_size,
2921							    tp1->book_size,
2922							    (uintptr_t) tp1->whoTo,
2923							    tp1->rec.data.TSN_seq);
2924						}
2925						sctp_flight_size_decrease(tp1);
2926						if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) {
2927							(*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) (tp1->whoTo,
2928							    tp1);
2929						}
2930						sctp_total_flight_decrease(stcb, tp1);
2931
2932						tp1->whoTo->net_ack += tp1->send_size;
2933						if (tp1->snd_count < 2) {
2934							/*-
2935							 * True non-retransmited chunk
2936							 */
2937							tp1->whoTo->net_ack2 += tp1->send_size;
2938
2939							/*-
2940							 * update RTO too ?
2941							 */
2942							if (tp1->do_rtt) {
2943								if (*rto_ok) {
2944									tp1->whoTo->RTO =
2945									    sctp_calculate_rto(stcb,
2946									    &stcb->asoc,
2947									    tp1->whoTo,
2948									    &tp1->sent_rcv_time,
2949									    sctp_align_safe_nocopy,
2950									    SCTP_RTT_FROM_DATA);
2951									*rto_ok = 0;
2952								}
2953								if (tp1->whoTo->rto_needed == 0) {
2954									tp1->whoTo->rto_needed = 1;
2955								}
2956								tp1->do_rtt = 0;
2957							}
2958						}
2959					}
2960					if (tp1->sent <= SCTP_DATAGRAM_RESEND) {
2961						if (SCTP_TSN_GT(tp1->rec.data.TSN_seq,
2962						    stcb->asoc.this_sack_highest_gap)) {
2963							stcb->asoc.this_sack_highest_gap =
2964							    tp1->rec.data.TSN_seq;
2965						}
2966						if (tp1->sent == SCTP_DATAGRAM_RESEND) {
2967							sctp_ucount_decr(stcb->asoc.sent_queue_retran_cnt);
2968#ifdef SCTP_AUDITING_ENABLED
2969							sctp_audit_log(0xB2,
2970							    (stcb->asoc.sent_queue_retran_cnt & 0x000000ff));
2971#endif
2972						}
2973					}
2974					/*-
2975					 * All chunks NOT UNSENT fall through here and are marked
2976					 * (leave PR-SCTP ones that are to skip alone though)
2977					 */
2978					if ((tp1->sent != SCTP_FORWARD_TSN_SKIP) &&
2979					    (tp1->sent != SCTP_DATAGRAM_NR_ACKED)) {
2980						tp1->sent = SCTP_DATAGRAM_MARKED;
2981					}
2982					if (tp1->rec.data.chunk_was_revoked) {
2983						/* deflate the cwnd */
2984						tp1->whoTo->cwnd -= tp1->book_size;
2985						tp1->rec.data.chunk_was_revoked = 0;
2986					}
2987					/* NR Sack code here */
2988					if (nr_sacking &&
2989					    (tp1->sent != SCTP_DATAGRAM_NR_ACKED)) {
2990						if (stcb->asoc.strmout[tp1->rec.data.stream_number].chunks_on_queues > 0) {
2991							stcb->asoc.strmout[tp1->rec.data.stream_number].chunks_on_queues--;
2992#ifdef INVARIANTS
2993						} else {
2994							panic("No chunks on the queues for sid %u.", tp1->rec.data.stream_number);
2995#endif
2996						}
2997						tp1->sent = SCTP_DATAGRAM_NR_ACKED;
2998						if (tp1->data) {
2999							/*
3000							 * sa_ignore
3001							 * NO_NULL_CHK
3002							 */
3003							sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1);
3004							sctp_m_freem(tp1->data);
3005							tp1->data = NULL;
3006						}
3007						wake_him++;
3008					}
3009				}
3010				break;
3011			}	/* if (tp1->TSN_seq == theTSN) */
3012			if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, theTSN)) {
3013				break;
3014			}
3015			tp1 = TAILQ_NEXT(tp1, sctp_next);
3016			if ((tp1 == NULL) && (circled == 0)) {
3017				circled++;
3018				tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue);
3019			}
3020		}		/* end while (tp1) */
3021		if (tp1 == NULL) {
3022			circled = 0;
3023			tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue);
3024		}
3025		/* In case the fragments were not in order we must reset */
3026	}			/* end for (j = fragStart */
3027	*p_tp1 = tp1;
3028	return (wake_him);	/* Return value only used for nr-sack */
3029}
3030
3031
3032static int
3033sctp_handle_segments(struct mbuf *m, int *offset, struct sctp_tcb *stcb, struct sctp_association *asoc,
3034    uint32_t last_tsn, uint32_t * biggest_tsn_acked,
3035    uint32_t * biggest_newly_acked_tsn, uint32_t * this_sack_lowest_newack,
3036    int num_seg, int num_nr_seg, int *rto_ok)
3037{
3038	struct sctp_gap_ack_block *frag, block;
3039	struct sctp_tmit_chunk *tp1;
3040	int i;
3041	int num_frs = 0;
3042	int chunk_freed;
3043	int non_revocable;
3044	uint16_t frag_strt, frag_end, prev_frag_end;
3045
3046	tp1 = TAILQ_FIRST(&asoc->sent_queue);
3047	prev_frag_end = 0;
3048	chunk_freed = 0;
3049
3050	for (i = 0; i < (num_seg + num_nr_seg); i++) {
3051		if (i == num_seg) {
3052			prev_frag_end = 0;
3053			tp1 = TAILQ_FIRST(&asoc->sent_queue);
3054		}
3055		frag = (struct sctp_gap_ack_block *)sctp_m_getptr(m, *offset,
3056		    sizeof(struct sctp_gap_ack_block), (uint8_t *) & block);
3057		*offset += sizeof(block);
3058		if (frag == NULL) {
3059			return (chunk_freed);
3060		}
3061		frag_strt = ntohs(frag->start);
3062		frag_end = ntohs(frag->end);
3063
3064		if (frag_strt > frag_end) {
3065			/* This gap report is malformed, skip it. */
3066			continue;
3067		}
3068		if (frag_strt <= prev_frag_end) {
3069			/* This gap report is not in order, so restart. */
3070			tp1 = TAILQ_FIRST(&asoc->sent_queue);
3071		}
3072		if (SCTP_TSN_GT((last_tsn + frag_end), *biggest_tsn_acked)) {
3073			*biggest_tsn_acked = last_tsn + frag_end;
3074		}
3075		if (i < num_seg) {
3076			non_revocable = 0;
3077		} else {
3078			non_revocable = 1;
3079		}
3080		if (sctp_process_segment_range(stcb, &tp1, last_tsn, frag_strt, frag_end,
3081		    non_revocable, &num_frs, biggest_newly_acked_tsn,
3082		    this_sack_lowest_newack, rto_ok)) {
3083			chunk_freed = 1;
3084		}
3085		prev_frag_end = frag_end;
3086	}
3087	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3088		if (num_frs)
3089			sctp_log_fr(*biggest_tsn_acked,
3090			    *biggest_newly_acked_tsn,
3091			    last_tsn, SCTP_FR_LOG_BIGGEST_TSNS);
3092	}
3093	return (chunk_freed);
3094}
3095
3096static void
3097sctp_check_for_revoked(struct sctp_tcb *stcb,
3098    struct sctp_association *asoc, uint32_t cumack,
3099    uint32_t biggest_tsn_acked)
3100{
3101	struct sctp_tmit_chunk *tp1;
3102
3103	TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
3104		if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, cumack)) {
3105			/*
3106			 * ok this guy is either ACK or MARKED. If it is
3107			 * ACKED it has been previously acked but not this
3108			 * time i.e. revoked.  If it is MARKED it was ACK'ed
3109			 * again.
3110			 */
3111			if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, biggest_tsn_acked)) {
3112				break;
3113			}
3114			if (tp1->sent == SCTP_DATAGRAM_ACKED) {
3115				/* it has been revoked */
3116				tp1->sent = SCTP_DATAGRAM_SENT;
3117				tp1->rec.data.chunk_was_revoked = 1;
3118				/*
3119				 * We must add this stuff back in to assure
3120				 * timers and such get started.
3121				 */
3122				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
3123					sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE,
3124					    tp1->whoTo->flight_size,
3125					    tp1->book_size,
3126					    (uintptr_t) tp1->whoTo,
3127					    tp1->rec.data.TSN_seq);
3128				}
3129				sctp_flight_size_increase(tp1);
3130				sctp_total_flight_increase(stcb, tp1);
3131				/*
3132				 * We inflate the cwnd to compensate for our
3133				 * artificial inflation of the flight_size.
3134				 */
3135				tp1->whoTo->cwnd += tp1->book_size;
3136				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
3137					sctp_log_sack(asoc->last_acked_seq,
3138					    cumack,
3139					    tp1->rec.data.TSN_seq,
3140					    0,
3141					    0,
3142					    SCTP_LOG_TSN_REVOKED);
3143				}
3144			} else if (tp1->sent == SCTP_DATAGRAM_MARKED) {
3145				/* it has been re-acked in this SACK */
3146				tp1->sent = SCTP_DATAGRAM_ACKED;
3147			}
3148		}
3149		if (tp1->sent == SCTP_DATAGRAM_UNSENT)
3150			break;
3151	}
3152}
3153
3154
3155static void
3156sctp_strike_gap_ack_chunks(struct sctp_tcb *stcb, struct sctp_association *asoc,
3157    uint32_t biggest_tsn_acked, uint32_t biggest_tsn_newly_acked, uint32_t this_sack_lowest_newack, int accum_moved)
3158{
3159	struct sctp_tmit_chunk *tp1;
3160	int strike_flag = 0;
3161	struct timeval now;
3162	int tot_retrans = 0;
3163	uint32_t sending_seq;
3164	struct sctp_nets *net;
3165	int num_dests_sacked = 0;
3166
3167	/*
3168	 * select the sending_seq, this is either the next thing ready to be
3169	 * sent but not transmitted, OR, the next seq we assign.
3170	 */
3171	tp1 = TAILQ_FIRST(&stcb->asoc.send_queue);
3172	if (tp1 == NULL) {
3173		sending_seq = asoc->sending_seq;
3174	} else {
3175		sending_seq = tp1->rec.data.TSN_seq;
3176	}
3177
3178	/* CMT DAC algo: finding out if SACK is a mixed SACK */
3179	if ((asoc->sctp_cmt_on_off > 0) &&
3180	    SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) {
3181		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3182			if (net->saw_newack)
3183				num_dests_sacked++;
3184		}
3185	}
3186	if (stcb->asoc.peer_supports_prsctp) {
3187		(void)SCTP_GETTIME_TIMEVAL(&now);
3188	}
3189	TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
3190		strike_flag = 0;
3191		if (tp1->no_fr_allowed) {
3192			/* this one had a timeout or something */
3193			continue;
3194		}
3195		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3196			if (tp1->sent < SCTP_DATAGRAM_RESEND)
3197				sctp_log_fr(biggest_tsn_newly_acked,
3198				    tp1->rec.data.TSN_seq,
3199				    tp1->sent,
3200				    SCTP_FR_LOG_CHECK_STRIKE);
3201		}
3202		if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, biggest_tsn_acked) ||
3203		    tp1->sent == SCTP_DATAGRAM_UNSENT) {
3204			/* done */
3205			break;
3206		}
3207		if (stcb->asoc.peer_supports_prsctp) {
3208			if ((PR_SCTP_TTL_ENABLED(tp1->flags)) && tp1->sent < SCTP_DATAGRAM_ACKED) {
3209				/* Is it expired? */
3210				if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) {
3211					/* Yes so drop it */
3212					if (tp1->data != NULL) {
3213						(void)sctp_release_pr_sctp_chunk(stcb, tp1, 1,
3214						    SCTP_SO_NOT_LOCKED);
3215					}
3216					continue;
3217				}
3218			}
3219		}
3220		if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, asoc->this_sack_highest_gap)) {
3221			/* we are beyond the tsn in the sack  */
3222			break;
3223		}
3224		if (tp1->sent >= SCTP_DATAGRAM_RESEND) {
3225			/* either a RESEND, ACKED, or MARKED */
3226			/* skip */
3227			if (tp1->sent == SCTP_FORWARD_TSN_SKIP) {
3228				/* Continue strikin FWD-TSN chunks */
3229				tp1->rec.data.fwd_tsn_cnt++;
3230			}
3231			continue;
3232		}
3233		/*
3234		 * CMT : SFR algo (covers part of DAC and HTNA as well)
3235		 */
3236		if (tp1->whoTo && tp1->whoTo->saw_newack == 0) {
3237			/*
3238			 * No new acks were receieved for data sent to this
3239			 * dest. Therefore, according to the SFR algo for
3240			 * CMT, no data sent to this dest can be marked for
3241			 * FR using this SACK.
3242			 */
3243			continue;
3244		} else if (tp1->whoTo && SCTP_TSN_GT(tp1->rec.data.TSN_seq,
3245		    tp1->whoTo->this_sack_highest_newack)) {
3246			/*
3247			 * CMT: New acks were receieved for data sent to
3248			 * this dest. But no new acks were seen for data
3249			 * sent after tp1. Therefore, according to the SFR
3250			 * algo for CMT, tp1 cannot be marked for FR using
3251			 * this SACK. This step covers part of the DAC algo
3252			 * and the HTNA algo as well.
3253			 */
3254			continue;
3255		}
3256		/*
3257		 * Here we check to see if we were have already done a FR
3258		 * and if so we see if the biggest TSN we saw in the sack is
3259		 * smaller than the recovery point. If so we don't strike
3260		 * the tsn... otherwise we CAN strike the TSN.
3261		 */
3262		/*
3263		 * @@@ JRI: Check for CMT if (accum_moved &&
3264		 * asoc->fast_retran_loss_recovery && (sctp_cmt_on_off ==
3265		 * 0)) {
3266		 */
3267		if (accum_moved && asoc->fast_retran_loss_recovery) {
3268			/*
3269			 * Strike the TSN if in fast-recovery and cum-ack
3270			 * moved.
3271			 */
3272			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3273				sctp_log_fr(biggest_tsn_newly_acked,
3274				    tp1->rec.data.TSN_seq,
3275				    tp1->sent,
3276				    SCTP_FR_LOG_STRIKE_CHUNK);
3277			}
3278			if (tp1->sent < SCTP_DATAGRAM_RESEND) {
3279				tp1->sent++;
3280			}
3281			if ((asoc->sctp_cmt_on_off > 0) &&
3282			    SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) {
3283				/*
3284				 * CMT DAC algorithm: If SACK flag is set to
3285				 * 0, then lowest_newack test will not pass
3286				 * because it would have been set to the
3287				 * cumack earlier. If not already to be
3288				 * rtx'd, If not a mixed sack and if tp1 is
3289				 * not between two sacked TSNs, then mark by
3290				 * one more. NOTE that we are marking by one
3291				 * additional time since the SACK DAC flag
3292				 * indicates that two packets have been
3293				 * received after this missing TSN.
3294				 */
3295				if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) &&
3296				    SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.TSN_seq)) {
3297					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3298						sctp_log_fr(16 + num_dests_sacked,
3299						    tp1->rec.data.TSN_seq,
3300						    tp1->sent,
3301						    SCTP_FR_LOG_STRIKE_CHUNK);
3302					}
3303					tp1->sent++;
3304				}
3305			}
3306		} else if ((tp1->rec.data.doing_fast_retransmit) &&
3307		    (asoc->sctp_cmt_on_off == 0)) {
3308			/*
3309			 * For those that have done a FR we must take
3310			 * special consideration if we strike. I.e the
3311			 * biggest_newly_acked must be higher than the
3312			 * sending_seq at the time we did the FR.
3313			 */
3314			if (
3315#ifdef SCTP_FR_TO_ALTERNATE
3316			/*
3317			 * If FR's go to new networks, then we must only do
3318			 * this for singly homed asoc's. However if the FR's
3319			 * go to the same network (Armando's work) then its
3320			 * ok to FR multiple times.
3321			 */
3322			    (asoc->numnets < 2)
3323#else
3324			    (1)
3325#endif
3326			    ) {
3327
3328				if (SCTP_TSN_GE(biggest_tsn_newly_acked,
3329				    tp1->rec.data.fast_retran_tsn)) {
3330					/*
3331					 * Strike the TSN, since this ack is
3332					 * beyond where things were when we
3333					 * did a FR.
3334					 */
3335					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3336						sctp_log_fr(biggest_tsn_newly_acked,
3337						    tp1->rec.data.TSN_seq,
3338						    tp1->sent,
3339						    SCTP_FR_LOG_STRIKE_CHUNK);
3340					}
3341					if (tp1->sent < SCTP_DATAGRAM_RESEND) {
3342						tp1->sent++;
3343					}
3344					strike_flag = 1;
3345					if ((asoc->sctp_cmt_on_off > 0) &&
3346					    SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) {
3347						/*
3348						 * CMT DAC algorithm: If
3349						 * SACK flag is set to 0,
3350						 * then lowest_newack test
3351						 * will not pass because it
3352						 * would have been set to
3353						 * the cumack earlier. If
3354						 * not already to be rtx'd,
3355						 * If not a mixed sack and
3356						 * if tp1 is not between two
3357						 * sacked TSNs, then mark by
3358						 * one more. NOTE that we
3359						 * are marking by one
3360						 * additional time since the
3361						 * SACK DAC flag indicates
3362						 * that two packets have
3363						 * been received after this
3364						 * missing TSN.
3365						 */
3366						if ((tp1->sent < SCTP_DATAGRAM_RESEND) &&
3367						    (num_dests_sacked == 1) &&
3368						    SCTP_TSN_GT(this_sack_lowest_newack,
3369						    tp1->rec.data.TSN_seq)) {
3370							if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3371								sctp_log_fr(32 + num_dests_sacked,
3372								    tp1->rec.data.TSN_seq,
3373								    tp1->sent,
3374								    SCTP_FR_LOG_STRIKE_CHUNK);
3375							}
3376							if (tp1->sent < SCTP_DATAGRAM_RESEND) {
3377								tp1->sent++;
3378							}
3379						}
3380					}
3381				}
3382			}
3383			/*
3384			 * JRI: TODO: remove code for HTNA algo. CMT's SFR
3385			 * algo covers HTNA.
3386			 */
3387		} else if (SCTP_TSN_GT(tp1->rec.data.TSN_seq,
3388		    biggest_tsn_newly_acked)) {
3389			/*
3390			 * We don't strike these: This is the  HTNA
3391			 * algorithm i.e. we don't strike If our TSN is
3392			 * larger than the Highest TSN Newly Acked.
3393			 */
3394			;
3395		} else {
3396			/* Strike the TSN */
3397			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3398				sctp_log_fr(biggest_tsn_newly_acked,
3399				    tp1->rec.data.TSN_seq,
3400				    tp1->sent,
3401				    SCTP_FR_LOG_STRIKE_CHUNK);
3402			}
3403			if (tp1->sent < SCTP_DATAGRAM_RESEND) {
3404				tp1->sent++;
3405			}
3406			if ((asoc->sctp_cmt_on_off > 0) &&
3407			    SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) {
3408				/*
3409				 * CMT DAC algorithm: If SACK flag is set to
3410				 * 0, then lowest_newack test will not pass
3411				 * because it would have been set to the
3412				 * cumack earlier. If not already to be
3413				 * rtx'd, If not a mixed sack and if tp1 is
3414				 * not between two sacked TSNs, then mark by
3415				 * one more. NOTE that we are marking by one
3416				 * additional time since the SACK DAC flag
3417				 * indicates that two packets have been
3418				 * received after this missing TSN.
3419				 */
3420				if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) &&
3421				    SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.TSN_seq)) {
3422					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3423						sctp_log_fr(48 + num_dests_sacked,
3424						    tp1->rec.data.TSN_seq,
3425						    tp1->sent,
3426						    SCTP_FR_LOG_STRIKE_CHUNK);
3427					}
3428					tp1->sent++;
3429				}
3430			}
3431		}
3432		if (tp1->sent == SCTP_DATAGRAM_RESEND) {
3433			struct sctp_nets *alt;
3434
3435			/* fix counts and things */
3436			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
3437				sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND,
3438				    (tp1->whoTo ? (tp1->whoTo->flight_size) : 0),
3439				    tp1->book_size,
3440				    (uintptr_t) tp1->whoTo,
3441				    tp1->rec.data.TSN_seq);
3442			}
3443			if (tp1->whoTo) {
3444				tp1->whoTo->net_ack++;
3445				sctp_flight_size_decrease(tp1);
3446				if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) {
3447					(*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) (tp1->whoTo,
3448					    tp1);
3449				}
3450			}
3451			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) {
3452				sctp_log_rwnd(SCTP_INCREASE_PEER_RWND,
3453				    asoc->peers_rwnd, tp1->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh));
3454			}
3455			/* add back to the rwnd */
3456			asoc->peers_rwnd += (tp1->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh));
3457
3458			/* remove from the total flight */
3459			sctp_total_flight_decrease(stcb, tp1);
3460
3461			if ((stcb->asoc.peer_supports_prsctp) &&
3462			    (PR_SCTP_RTX_ENABLED(tp1->flags))) {
3463				/*
3464				 * Has it been retransmitted tv_sec times? -
3465				 * we store the retran count there.
3466				 */
3467				if (tp1->snd_count > tp1->rec.data.timetodrop.tv_sec) {
3468					/* Yes, so drop it */
3469					if (tp1->data != NULL) {
3470						(void)sctp_release_pr_sctp_chunk(stcb, tp1, 1,
3471						    SCTP_SO_NOT_LOCKED);
3472					}
3473					/* Make sure to flag we had a FR */
3474					tp1->whoTo->net_ack++;
3475					continue;
3476				}
3477			}
3478			/*
3479			 * SCTP_PRINTF("OK, we are now ready to FR this
3480			 * guy\n");
3481			 */
3482			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
3483				sctp_log_fr(tp1->rec.data.TSN_seq, tp1->snd_count,
3484				    0, SCTP_FR_MARKED);
3485			}
3486			if (strike_flag) {
3487				/* This is a subsequent FR */
3488				SCTP_STAT_INCR(sctps_sendmultfastretrans);
3489			}
3490			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
3491			if (asoc->sctp_cmt_on_off > 0) {
3492				/*
3493				 * CMT: Using RTX_SSTHRESH policy for CMT.
3494				 * If CMT is being used, then pick dest with
3495				 * largest ssthresh for any retransmission.
3496				 */
3497				tp1->no_fr_allowed = 1;
3498				alt = tp1->whoTo;
3499				/* sa_ignore NO_NULL_CHK */
3500				if (asoc->sctp_cmt_pf > 0) {
3501					/*
3502					 * JRS 5/18/07 - If CMT PF is on,
3503					 * use the PF version of
3504					 * find_alt_net()
3505					 */
3506					alt = sctp_find_alternate_net(stcb, alt, 2);
3507				} else {
3508					/*
3509					 * JRS 5/18/07 - If only CMT is on,
3510					 * use the CMT version of
3511					 * find_alt_net()
3512					 */
3513					/* sa_ignore NO_NULL_CHK */
3514					alt = sctp_find_alternate_net(stcb, alt, 1);
3515				}
3516				if (alt == NULL) {
3517					alt = tp1->whoTo;
3518				}
3519				/*
3520				 * CUCv2: If a different dest is picked for
3521				 * the retransmission, then new
3522				 * (rtx-)pseudo_cumack needs to be tracked
3523				 * for orig dest. Let CUCv2 track new (rtx-)
3524				 * pseudo-cumack always.
3525				 */
3526				if (tp1->whoTo) {
3527					tp1->whoTo->find_pseudo_cumack = 1;
3528					tp1->whoTo->find_rtx_pseudo_cumack = 1;
3529				}
3530			} else {/* CMT is OFF */
3531
3532#ifdef SCTP_FR_TO_ALTERNATE
3533				/* Can we find an alternate? */
3534				alt = sctp_find_alternate_net(stcb, tp1->whoTo, 0);
3535#else
3536				/*
3537				 * default behavior is to NOT retransmit
3538				 * FR's to an alternate. Armando Caro's
3539				 * paper details why.
3540				 */
3541				alt = tp1->whoTo;
3542#endif
3543			}
3544
3545			tp1->rec.data.doing_fast_retransmit = 1;
3546			tot_retrans++;
3547			/* mark the sending seq for possible subsequent FR's */
3548			/*
3549			 * SCTP_PRINTF("Marking TSN for FR new value %x\n",
3550			 * (uint32_t)tpi->rec.data.TSN_seq);
3551			 */
3552			if (TAILQ_EMPTY(&asoc->send_queue)) {
3553				/*
3554				 * If the queue of send is empty then its
3555				 * the next sequence number that will be
3556				 * assigned so we subtract one from this to
3557				 * get the one we last sent.
3558				 */
3559				tp1->rec.data.fast_retran_tsn = sending_seq;
3560			} else {
3561				/*
3562				 * If there are chunks on the send queue
3563				 * (unsent data that has made it from the
3564				 * stream queues but not out the door, we
3565				 * take the first one (which will have the
3566				 * lowest TSN) and subtract one to get the
3567				 * one we last sent.
3568				 */
3569				struct sctp_tmit_chunk *ttt;
3570
3571				ttt = TAILQ_FIRST(&asoc->send_queue);
3572				tp1->rec.data.fast_retran_tsn =
3573				    ttt->rec.data.TSN_seq;
3574			}
3575
3576			if (tp1->do_rtt) {
3577				/*
3578				 * this guy had a RTO calculation pending on
3579				 * it, cancel it
3580				 */
3581				if ((tp1->whoTo != NULL) &&
3582				    (tp1->whoTo->rto_needed == 0)) {
3583					tp1->whoTo->rto_needed = 1;
3584				}
3585				tp1->do_rtt = 0;
3586			}
3587			if (alt != tp1->whoTo) {
3588				/* yes, there is an alternate. */
3589				sctp_free_remote_addr(tp1->whoTo);
3590				/* sa_ignore FREED_MEMORY */
3591				tp1->whoTo = alt;
3592				atomic_add_int(&alt->ref_count, 1);
3593			}
3594		}
3595	}
3596}
3597
3598struct sctp_tmit_chunk *
3599sctp_try_advance_peer_ack_point(struct sctp_tcb *stcb,
3600    struct sctp_association *asoc)
3601{
3602	struct sctp_tmit_chunk *tp1, *tp2, *a_adv = NULL;
3603	struct timeval now;
3604	int now_filled = 0;
3605
3606	if (asoc->peer_supports_prsctp == 0) {
3607		return (NULL);
3608	}
3609	TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) {
3610		if (tp1->sent != SCTP_FORWARD_TSN_SKIP &&
3611		    tp1->sent != SCTP_DATAGRAM_RESEND &&
3612		    tp1->sent != SCTP_DATAGRAM_NR_ACKED) {
3613			/* no chance to advance, out of here */
3614			break;
3615		}
3616		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) {
3617			if ((tp1->sent == SCTP_FORWARD_TSN_SKIP) ||
3618			    (tp1->sent == SCTP_DATAGRAM_NR_ACKED)) {
3619				sctp_misc_ints(SCTP_FWD_TSN_CHECK,
3620				    asoc->advanced_peer_ack_point,
3621				    tp1->rec.data.TSN_seq, 0, 0);
3622			}
3623		}
3624		if (!PR_SCTP_ENABLED(tp1->flags)) {
3625			/*
3626			 * We can't fwd-tsn past any that are reliable aka
3627			 * retransmitted until the asoc fails.
3628			 */
3629			break;
3630		}
3631		if (!now_filled) {
3632			(void)SCTP_GETTIME_TIMEVAL(&now);
3633			now_filled = 1;
3634		}
3635		/*
3636		 * now we got a chunk which is marked for another
3637		 * retransmission to a PR-stream but has run out its chances
3638		 * already maybe OR has been marked to skip now. Can we skip
3639		 * it if its a resend?
3640		 */
3641		if (tp1->sent == SCTP_DATAGRAM_RESEND &&
3642		    (PR_SCTP_TTL_ENABLED(tp1->flags))) {
3643			/*
3644			 * Now is this one marked for resend and its time is
3645			 * now up?
3646			 */
3647			if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) {
3648				/* Yes so drop it */
3649				if (tp1->data) {
3650					(void)sctp_release_pr_sctp_chunk(stcb, tp1,
3651					    1, SCTP_SO_NOT_LOCKED);
3652				}
3653			} else {
3654				/*
3655				 * No, we are done when hit one for resend
3656				 * whos time as not expired.
3657				 */
3658				break;
3659			}
3660		}
3661		/*
3662		 * Ok now if this chunk is marked to drop it we can clean up
3663		 * the chunk, advance our peer ack point and we can check
3664		 * the next chunk.
3665		 */
3666		if ((tp1->sent == SCTP_FORWARD_TSN_SKIP) ||
3667		    (tp1->sent == SCTP_DATAGRAM_NR_ACKED)) {
3668			/* advance PeerAckPoint goes forward */
3669			if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, asoc->advanced_peer_ack_point)) {
3670				asoc->advanced_peer_ack_point = tp1->rec.data.TSN_seq;
3671				a_adv = tp1;
3672			} else if (tp1->rec.data.TSN_seq == asoc->advanced_peer_ack_point) {
3673				/* No update but we do save the chk */
3674				a_adv = tp1;
3675			}
3676		} else {
3677			/*
3678			 * If it is still in RESEND we can advance no
3679			 * further
3680			 */
3681			break;
3682		}
3683	}
3684	return (a_adv);
3685}
3686
3687static int
3688sctp_fs_audit(struct sctp_association *asoc)
3689{
3690	struct sctp_tmit_chunk *chk;
3691	int inflight = 0, resend = 0, inbetween = 0, acked = 0, above = 0;
3692	int entry_flight, entry_cnt, ret;
3693
3694	entry_flight = asoc->total_flight;
3695	entry_cnt = asoc->total_flight_count;
3696	ret = 0;
3697
3698	if (asoc->pr_sctp_cnt >= asoc->sent_queue_cnt)
3699		return (0);
3700
3701	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
3702		if (chk->sent < SCTP_DATAGRAM_RESEND) {
3703			SCTP_PRINTF("Chk TSN:%u size:%d inflight cnt:%d\n",
3704			    chk->rec.data.TSN_seq,
3705			    chk->send_size,
3706			    chk->snd_count);
3707			inflight++;
3708		} else if (chk->sent == SCTP_DATAGRAM_RESEND) {
3709			resend++;
3710		} else if (chk->sent < SCTP_DATAGRAM_ACKED) {
3711			inbetween++;
3712		} else if (chk->sent > SCTP_DATAGRAM_ACKED) {
3713			above++;
3714		} else {
3715			acked++;
3716		}
3717	}
3718
3719	if ((inflight > 0) || (inbetween > 0)) {
3720#ifdef INVARIANTS
3721		panic("Flight size-express incorrect? \n");
3722#else
3723		SCTP_PRINTF("asoc->total_flight:%d cnt:%d\n",
3724		    entry_flight, entry_cnt);
3725
3726		SCTP_PRINTF("Flight size-express incorrect F:%d I:%d R:%d Ab:%d ACK:%d\n",
3727		    inflight, inbetween, resend, above, acked);
3728		ret = 1;
3729#endif
3730	}
3731	return (ret);
3732}
3733
3734
3735static void
3736sctp_window_probe_recovery(struct sctp_tcb *stcb,
3737    struct sctp_association *asoc,
3738    struct sctp_tmit_chunk *tp1)
3739{
3740	tp1->window_probe = 0;
3741	if ((tp1->sent >= SCTP_DATAGRAM_ACKED) || (tp1->data == NULL)) {
3742		/* TSN's skipped we do NOT move back. */
3743		sctp_misc_ints(SCTP_FLIGHT_LOG_DWN_WP_FWD,
3744		    tp1->whoTo->flight_size,
3745		    tp1->book_size,
3746		    (uintptr_t) tp1->whoTo,
3747		    tp1->rec.data.TSN_seq);
3748		return;
3749	}
3750	/* First setup this by shrinking flight */
3751	if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) {
3752		(*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) (tp1->whoTo,
3753		    tp1);
3754	}
3755	sctp_flight_size_decrease(tp1);
3756	sctp_total_flight_decrease(stcb, tp1);
3757	/* Now mark for resend */
3758	tp1->sent = SCTP_DATAGRAM_RESEND;
3759	sctp_ucount_incr(asoc->sent_queue_retran_cnt);
3760
3761	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
3762		sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_WP,
3763		    tp1->whoTo->flight_size,
3764		    tp1->book_size,
3765		    (uintptr_t) tp1->whoTo,
3766		    tp1->rec.data.TSN_seq);
3767	}
3768}
3769
3770void
3771sctp_express_handle_sack(struct sctp_tcb *stcb, uint32_t cumack,
3772    uint32_t rwnd, int *abort_now, int ecne_seen)
3773{
3774	struct sctp_nets *net;
3775	struct sctp_association *asoc;
3776	struct sctp_tmit_chunk *tp1, *tp2;
3777	uint32_t old_rwnd;
3778	int win_probe_recovery = 0;
3779	int win_probe_recovered = 0;
3780	int j, done_once = 0;
3781	int rto_ok = 1;
3782
3783	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) {
3784		sctp_misc_ints(SCTP_SACK_LOG_EXPRESS, cumack,
3785		    rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd);
3786	}
3787	SCTP_TCB_LOCK_ASSERT(stcb);
3788#ifdef SCTP_ASOCLOG_OF_TSNS
3789	stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cumack;
3790	stcb->asoc.cumack_log_at++;
3791	if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) {
3792		stcb->asoc.cumack_log_at = 0;
3793	}
3794#endif
3795	asoc = &stcb->asoc;
3796	old_rwnd = asoc->peers_rwnd;
3797	if (SCTP_TSN_GT(asoc->last_acked_seq, cumack)) {
3798		/* old ack */
3799		return;
3800	} else if (asoc->last_acked_seq == cumack) {
3801		/* Window update sack */
3802		asoc->peers_rwnd = sctp_sbspace_sub(rwnd,
3803		    (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))));
3804		if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) {
3805			/* SWS sender side engages */
3806			asoc->peers_rwnd = 0;
3807		}
3808		if (asoc->peers_rwnd > old_rwnd) {
3809			goto again;
3810		}
3811		return;
3812	}
3813	/* First setup for CC stuff */
3814	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3815		if (SCTP_TSN_GT(cumack, net->cwr_window_tsn)) {
3816			/* Drag along the window_tsn for cwr's */
3817			net->cwr_window_tsn = cumack;
3818		}
3819		net->prev_cwnd = net->cwnd;
3820		net->net_ack = 0;
3821		net->net_ack2 = 0;
3822
3823		/*
3824		 * CMT: Reset CUC and Fast recovery algo variables before
3825		 * SACK processing
3826		 */
3827		net->new_pseudo_cumack = 0;
3828		net->will_exit_fast_recovery = 0;
3829		if (stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) {
3830			(*stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) (stcb, net);
3831		}
3832	}
3833	if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) {
3834		uint32_t send_s;
3835
3836		if (!TAILQ_EMPTY(&asoc->sent_queue)) {
3837			tp1 = TAILQ_LAST(&asoc->sent_queue,
3838			    sctpchunk_listhead);
3839			send_s = tp1->rec.data.TSN_seq + 1;
3840		} else {
3841			send_s = asoc->sending_seq;
3842		}
3843		if (SCTP_TSN_GE(cumack, send_s)) {
3844#ifndef INVARIANTS
3845			struct mbuf *oper;
3846
3847#endif
3848#ifdef INVARIANTS
3849			panic("Impossible sack 1");
3850#else
3851
3852			*abort_now = 1;
3853			/* XXX */
3854			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3855			    0, M_NOWAIT, 1, MT_DATA);
3856			if (oper) {
3857				struct sctp_paramhdr *ph;
3858				uint32_t *ippp;
3859
3860				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
3861				    sizeof(uint32_t);
3862				ph = mtod(oper, struct sctp_paramhdr *);
3863				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
3864				ph->param_length = htons(SCTP_BUF_LEN(oper));
3865				ippp = (uint32_t *) (ph + 1);
3866				*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_25);
3867			}
3868			stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_25;
3869			sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
3870			return;
3871#endif
3872		}
3873	}
3874	asoc->this_sack_highest_gap = cumack;
3875	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
3876		sctp_misc_ints(SCTP_THRESHOLD_CLEAR,
3877		    stcb->asoc.overall_error_count,
3878		    0,
3879		    SCTP_FROM_SCTP_INDATA,
3880		    __LINE__);
3881	}
3882	stcb->asoc.overall_error_count = 0;
3883	if (SCTP_TSN_GT(cumack, asoc->last_acked_seq)) {
3884		/* process the new consecutive TSN first */
3885		TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) {
3886			if (SCTP_TSN_GE(cumack, tp1->rec.data.TSN_seq)) {
3887				if (tp1->sent == SCTP_DATAGRAM_UNSENT) {
3888					SCTP_PRINTF("Warning, an unsent is now acked?\n");
3889				}
3890				if (tp1->sent < SCTP_DATAGRAM_ACKED) {
3891					/*
3892					 * If it is less than ACKED, it is
3893					 * now no-longer in flight. Higher
3894					 * values may occur during marking
3895					 */
3896					if (tp1->sent < SCTP_DATAGRAM_RESEND) {
3897						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
3898							sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA,
3899							    tp1->whoTo->flight_size,
3900							    tp1->book_size,
3901							    (uintptr_t) tp1->whoTo,
3902							    tp1->rec.data.TSN_seq);
3903						}
3904						sctp_flight_size_decrease(tp1);
3905						if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) {
3906							(*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) (tp1->whoTo,
3907							    tp1);
3908						}
3909						/* sa_ignore NO_NULL_CHK */
3910						sctp_total_flight_decrease(stcb, tp1);
3911					}
3912					tp1->whoTo->net_ack += tp1->send_size;
3913					if (tp1->snd_count < 2) {
3914						/*
3915						 * True non-retransmited
3916						 * chunk
3917						 */
3918						tp1->whoTo->net_ack2 +=
3919						    tp1->send_size;
3920
3921						/* update RTO too? */
3922						if (tp1->do_rtt) {
3923							if (rto_ok) {
3924								tp1->whoTo->RTO =
3925								/*
3926								 * sa_ignore
3927								 * NO_NULL_CH
3928								 * K
3929								 */
3930								    sctp_calculate_rto(stcb,
3931								    asoc, tp1->whoTo,
3932								    &tp1->sent_rcv_time,
3933								    sctp_align_safe_nocopy,
3934								    SCTP_RTT_FROM_DATA);
3935								rto_ok = 0;
3936							}
3937							if (tp1->whoTo->rto_needed == 0) {
3938								tp1->whoTo->rto_needed = 1;
3939							}
3940							tp1->do_rtt = 0;
3941						}
3942					}
3943					/*
3944					 * CMT: CUCv2 algorithm. From the
3945					 * cumack'd TSNs, for each TSN being
3946					 * acked for the first time, set the
3947					 * following variables for the
3948					 * corresp destination.
3949					 * new_pseudo_cumack will trigger a
3950					 * cwnd update.
3951					 * find_(rtx_)pseudo_cumack will
3952					 * trigger search for the next
3953					 * expected (rtx-)pseudo-cumack.
3954					 */
3955					tp1->whoTo->new_pseudo_cumack = 1;
3956					tp1->whoTo->find_pseudo_cumack = 1;
3957					tp1->whoTo->find_rtx_pseudo_cumack = 1;
3958
3959					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
3960						/* sa_ignore NO_NULL_CHK */
3961						sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK);
3962					}
3963				}
3964				if (tp1->sent == SCTP_DATAGRAM_RESEND) {
3965					sctp_ucount_decr(asoc->sent_queue_retran_cnt);
3966				}
3967				if (tp1->rec.data.chunk_was_revoked) {
3968					/* deflate the cwnd */
3969					tp1->whoTo->cwnd -= tp1->book_size;
3970					tp1->rec.data.chunk_was_revoked = 0;
3971				}
3972				if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) {
3973					if (asoc->strmout[tp1->rec.data.stream_number].chunks_on_queues > 0) {
3974						asoc->strmout[tp1->rec.data.stream_number].chunks_on_queues--;
3975#ifdef INVARIANTS
3976					} else {
3977						panic("No chunks on the queues for sid %u.", tp1->rec.data.stream_number);
3978#endif
3979					}
3980				}
3981				TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next);
3982				if (tp1->data) {
3983					/* sa_ignore NO_NULL_CHK */
3984					sctp_free_bufspace(stcb, asoc, tp1, 1);
3985					sctp_m_freem(tp1->data);
3986					tp1->data = NULL;
3987				}
3988				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
3989					sctp_log_sack(asoc->last_acked_seq,
3990					    cumack,
3991					    tp1->rec.data.TSN_seq,
3992					    0,
3993					    0,
3994					    SCTP_LOG_FREE_SENT);
3995				}
3996				asoc->sent_queue_cnt--;
3997				sctp_free_a_chunk(stcb, tp1, SCTP_SO_NOT_LOCKED);
3998			} else {
3999				break;
4000			}
4001		}
4002
4003	}
4004	/* sa_ignore NO_NULL_CHK */
4005	if (stcb->sctp_socket) {
4006#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4007		struct socket *so;
4008
4009#endif
4010		SOCKBUF_LOCK(&stcb->sctp_socket->so_snd);
4011		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) {
4012			/* sa_ignore NO_NULL_CHK */
4013			sctp_wakeup_log(stcb, 1, SCTP_WAKESND_FROM_SACK);
4014		}
4015#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4016		so = SCTP_INP_SO(stcb->sctp_ep);
4017		atomic_add_int(&stcb->asoc.refcnt, 1);
4018		SCTP_TCB_UNLOCK(stcb);
4019		SCTP_SOCKET_LOCK(so, 1);
4020		SCTP_TCB_LOCK(stcb);
4021		atomic_subtract_int(&stcb->asoc.refcnt, 1);
4022		if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) {
4023			/* assoc was freed while we were unlocked */
4024			SCTP_SOCKET_UNLOCK(so, 1);
4025			return;
4026		}
4027#endif
4028		sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket);
4029#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4030		SCTP_SOCKET_UNLOCK(so, 1);
4031#endif
4032	} else {
4033		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) {
4034			sctp_wakeup_log(stcb, 1, SCTP_NOWAKE_FROM_SACK);
4035		}
4036	}
4037
4038	/* JRS - Use the congestion control given in the CC module */
4039	if ((asoc->last_acked_seq != cumack) && (ecne_seen == 0)) {
4040		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4041			if (net->net_ack2 > 0) {
4042				/*
4043				 * Karn's rule applies to clearing error
4044				 * count, this is optional.
4045				 */
4046				net->error_count = 0;
4047				if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
4048					/* addr came good */
4049					net->dest_state |= SCTP_ADDR_REACHABLE;
4050					sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb,
4051					    0, (void *)net, SCTP_SO_NOT_LOCKED);
4052				}
4053				if (net == stcb->asoc.primary_destination) {
4054					if (stcb->asoc.alternate) {
4055						/*
4056						 * release the alternate,
4057						 * primary is good
4058						 */
4059						sctp_free_remote_addr(stcb->asoc.alternate);
4060						stcb->asoc.alternate = NULL;
4061					}
4062				}
4063				if (net->dest_state & SCTP_ADDR_PF) {
4064					net->dest_state &= ~SCTP_ADDR_PF;
4065					sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_3);
4066					sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
4067					asoc->cc_functions.sctp_cwnd_update_exit_pf(stcb, net);
4068					/* Done with this net */
4069					net->net_ack = 0;
4070				}
4071				/* restore any doubled timers */
4072				net->RTO = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
4073				if (net->RTO < stcb->asoc.minrto) {
4074					net->RTO = stcb->asoc.minrto;
4075				}
4076				if (net->RTO > stcb->asoc.maxrto) {
4077					net->RTO = stcb->asoc.maxrto;
4078				}
4079			}
4080		}
4081		asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, 1, 0, 0);
4082	}
4083	asoc->last_acked_seq = cumack;
4084
4085	if (TAILQ_EMPTY(&asoc->sent_queue)) {
4086		/* nothing left in-flight */
4087		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4088			net->flight_size = 0;
4089			net->partial_bytes_acked = 0;
4090		}
4091		asoc->total_flight = 0;
4092		asoc->total_flight_count = 0;
4093	}
4094	/* RWND update */
4095	asoc->peers_rwnd = sctp_sbspace_sub(rwnd,
4096	    (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))));
4097	if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) {
4098		/* SWS sender side engages */
4099		asoc->peers_rwnd = 0;
4100	}
4101	if (asoc->peers_rwnd > old_rwnd) {
4102		win_probe_recovery = 1;
4103	}
4104	/* Now assure a timer where data is queued at */
4105again:
4106	j = 0;
4107	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4108		int to_ticks;
4109
4110		if (win_probe_recovery && (net->window_probe)) {
4111			win_probe_recovered = 1;
4112			/*
4113			 * Find first chunk that was used with window probe
4114			 * and clear the sent
4115			 */
4116			/* sa_ignore FREED_MEMORY */
4117			TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
4118				if (tp1->window_probe) {
4119					/* move back to data send queue */
4120					sctp_window_probe_recovery(stcb, asoc, tp1);
4121					break;
4122				}
4123			}
4124		}
4125		if (net->RTO == 0) {
4126			to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto);
4127		} else {
4128			to_ticks = MSEC_TO_TICKS(net->RTO);
4129		}
4130		if (net->flight_size) {
4131			j++;
4132			(void)SCTP_OS_TIMER_START(&net->rxt_timer.timer, to_ticks,
4133			    sctp_timeout_handler, &net->rxt_timer);
4134			if (net->window_probe) {
4135				net->window_probe = 0;
4136			}
4137		} else {
4138			if (net->window_probe) {
4139				/*
4140				 * In window probes we must assure a timer
4141				 * is still running there
4142				 */
4143				net->window_probe = 0;
4144				if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) {
4145					SCTP_OS_TIMER_START(&net->rxt_timer.timer, to_ticks,
4146					    sctp_timeout_handler, &net->rxt_timer);
4147				}
4148			} else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) {
4149				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
4150				    stcb, net,
4151				    SCTP_FROM_SCTP_INDATA + SCTP_LOC_22);
4152			}
4153		}
4154	}
4155	if ((j == 0) &&
4156	    (!TAILQ_EMPTY(&asoc->sent_queue)) &&
4157	    (asoc->sent_queue_retran_cnt == 0) &&
4158	    (win_probe_recovered == 0) &&
4159	    (done_once == 0)) {
4160		/*
4161		 * huh, this should not happen unless all packets are
4162		 * PR-SCTP and marked to skip of course.
4163		 */
4164		if (sctp_fs_audit(asoc)) {
4165			TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4166				net->flight_size = 0;
4167			}
4168			asoc->total_flight = 0;
4169			asoc->total_flight_count = 0;
4170			asoc->sent_queue_retran_cnt = 0;
4171			TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
4172				if (tp1->sent < SCTP_DATAGRAM_RESEND) {
4173					sctp_flight_size_increase(tp1);
4174					sctp_total_flight_increase(stcb, tp1);
4175				} else if (tp1->sent == SCTP_DATAGRAM_RESEND) {
4176					sctp_ucount_incr(asoc->sent_queue_retran_cnt);
4177				}
4178			}
4179		}
4180		done_once = 1;
4181		goto again;
4182	}
4183	/**********************************/
4184	/* Now what about shutdown issues */
4185	/**********************************/
4186	if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) {
4187		/* nothing left on sendqueue.. consider done */
4188		/* clean up */
4189		if ((asoc->stream_queue_cnt == 1) &&
4190		    ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) ||
4191		    (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED)) &&
4192		    (asoc->locked_on_sending)
4193		    ) {
4194			struct sctp_stream_queue_pending *sp;
4195
4196			/*
4197			 * I may be in a state where we got all across.. but
4198			 * cannot write more due to a shutdown... we abort
4199			 * since the user did not indicate EOR in this case.
4200			 * The sp will be cleaned during free of the asoc.
4201			 */
4202			sp = TAILQ_LAST(&((asoc->locked_on_sending)->outqueue),
4203			    sctp_streamhead);
4204			if ((sp) && (sp->length == 0)) {
4205				/* Let cleanup code purge it */
4206				if (sp->msg_is_complete) {
4207					asoc->stream_queue_cnt--;
4208				} else {
4209					asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT;
4210					asoc->locked_on_sending = NULL;
4211					asoc->stream_queue_cnt--;
4212				}
4213			}
4214		}
4215		if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) &&
4216		    (asoc->stream_queue_cnt == 0)) {
4217			if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) {
4218				/* Need to abort here */
4219				struct mbuf *oper;
4220
4221		abort_out_now:
4222				*abort_now = 1;
4223				/* XXX */
4224				oper = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
4225				    0, M_NOWAIT, 1, MT_DATA);
4226				if (oper) {
4227					struct sctp_paramhdr *ph;
4228
4229					SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr);
4230					ph = mtod(oper, struct sctp_paramhdr *);
4231					ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
4232					ph->param_length = htons(SCTP_BUF_LEN(oper));
4233				}
4234				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_24;
4235				sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
4236			} else {
4237				struct sctp_nets *netp;
4238
4239				if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
4240				    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
4241					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
4242				}
4243				SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
4244				SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
4245				sctp_stop_timers_for_shutdown(stcb);
4246				if (asoc->alternate) {
4247					netp = asoc->alternate;
4248				} else {
4249					netp = asoc->primary_destination;
4250				}
4251				sctp_send_shutdown(stcb, netp);
4252				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
4253				    stcb->sctp_ep, stcb, netp);
4254				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
4255				    stcb->sctp_ep, stcb, netp);
4256			}
4257		} else if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) &&
4258		    (asoc->stream_queue_cnt == 0)) {
4259			struct sctp_nets *netp;
4260
4261			if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) {
4262				goto abort_out_now;
4263			}
4264			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
4265			SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_ACK_SENT);
4266			SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
4267			sctp_stop_timers_for_shutdown(stcb);
4268			if (asoc->alternate) {
4269				netp = asoc->alternate;
4270			} else {
4271				netp = asoc->primary_destination;
4272			}
4273			sctp_send_shutdown_ack(stcb, netp);
4274			sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK,
4275			    stcb->sctp_ep, stcb, netp);
4276		}
4277	}
4278	/*********************************************/
4279	/* Here we perform PR-SCTP procedures        */
4280	/* (section 4.2)                             */
4281	/*********************************************/
4282	/* C1. update advancedPeerAckPoint */
4283	if (SCTP_TSN_GT(cumack, asoc->advanced_peer_ack_point)) {
4284		asoc->advanced_peer_ack_point = cumack;
4285	}
4286	/* PR-Sctp issues need to be addressed too */
4287	if ((asoc->peer_supports_prsctp) && (asoc->pr_sctp_cnt > 0)) {
4288		struct sctp_tmit_chunk *lchk;
4289		uint32_t old_adv_peer_ack_point;
4290
4291		old_adv_peer_ack_point = asoc->advanced_peer_ack_point;
4292		lchk = sctp_try_advance_peer_ack_point(stcb, asoc);
4293		/* C3. See if we need to send a Fwd-TSN */
4294		if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cumack)) {
4295			/*
4296			 * ISSUE with ECN, see FWD-TSN processing.
4297			 */
4298			if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) {
4299				send_forward_tsn(stcb, asoc);
4300			} else if (lchk) {
4301				/* try to FR fwd-tsn's that get lost too */
4302				if (lchk->rec.data.fwd_tsn_cnt >= 3) {
4303					send_forward_tsn(stcb, asoc);
4304				}
4305			}
4306		}
4307		if (lchk) {
4308			/* Assure a timer is up */
4309			sctp_timer_start(SCTP_TIMER_TYPE_SEND,
4310			    stcb->sctp_ep, stcb, lchk->whoTo);
4311		}
4312	}
4313	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) {
4314		sctp_misc_ints(SCTP_SACK_RWND_UPDATE,
4315		    rwnd,
4316		    stcb->asoc.peers_rwnd,
4317		    stcb->asoc.total_flight,
4318		    stcb->asoc.total_output_queue_size);
4319	}
4320}
4321
4322void
4323sctp_handle_sack(struct mbuf *m, int offset_seg, int offset_dup,
4324    struct sctp_tcb *stcb,
4325    uint16_t num_seg, uint16_t num_nr_seg, uint16_t num_dup,
4326    int *abort_now, uint8_t flags,
4327    uint32_t cum_ack, uint32_t rwnd, int ecne_seen)
4328{
4329	struct sctp_association *asoc;
4330	struct sctp_tmit_chunk *tp1, *tp2;
4331	uint32_t last_tsn, biggest_tsn_acked, biggest_tsn_newly_acked, this_sack_lowest_newack;
4332	uint16_t wake_him = 0;
4333	uint32_t send_s = 0;
4334	long j;
4335	int accum_moved = 0;
4336	int will_exit_fast_recovery = 0;
4337	uint32_t a_rwnd, old_rwnd;
4338	int win_probe_recovery = 0;
4339	int win_probe_recovered = 0;
4340	struct sctp_nets *net = NULL;
4341	int done_once;
4342	int rto_ok = 1;
4343	uint8_t reneged_all = 0;
4344	uint8_t cmt_dac_flag;
4345
4346	/*
4347	 * we take any chance we can to service our queues since we cannot
4348	 * get awoken when the socket is read from :<
4349	 */
4350	/*
4351	 * Now perform the actual SACK handling: 1) Verify that it is not an
4352	 * old sack, if so discard. 2) If there is nothing left in the send
4353	 * queue (cum-ack is equal to last acked) then you have a duplicate
4354	 * too, update any rwnd change and verify no timers are running.
4355	 * then return. 3) Process any new consequtive data i.e. cum-ack
4356	 * moved process these first and note that it moved. 4) Process any
4357	 * sack blocks. 5) Drop any acked from the queue. 6) Check for any
4358	 * revoked blocks and mark. 7) Update the cwnd. 8) Nothing left,
4359	 * sync up flightsizes and things, stop all timers and also check
4360	 * for shutdown_pending state. If so then go ahead and send off the
4361	 * shutdown. If in shutdown recv, send off the shutdown-ack and
4362	 * start that timer, Ret. 9) Strike any non-acked things and do FR
4363	 * procedure if needed being sure to set the FR flag. 10) Do pr-sctp
4364	 * procedures. 11) Apply any FR penalties. 12) Assure we will SACK
4365	 * if in shutdown_recv state.
4366	 */
4367	SCTP_TCB_LOCK_ASSERT(stcb);
4368	/* CMT DAC algo */
4369	this_sack_lowest_newack = 0;
4370	SCTP_STAT_INCR(sctps_slowpath_sack);
4371	last_tsn = cum_ack;
4372	cmt_dac_flag = flags & SCTP_SACK_CMT_DAC;
4373#ifdef SCTP_ASOCLOG_OF_TSNS
4374	stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cum_ack;
4375	stcb->asoc.cumack_log_at++;
4376	if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) {
4377		stcb->asoc.cumack_log_at = 0;
4378	}
4379#endif
4380	a_rwnd = rwnd;
4381
4382	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) {
4383		sctp_misc_ints(SCTP_SACK_LOG_NORMAL, cum_ack,
4384		    rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd);
4385	}
4386	old_rwnd = stcb->asoc.peers_rwnd;
4387	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
4388		sctp_misc_ints(SCTP_THRESHOLD_CLEAR,
4389		    stcb->asoc.overall_error_count,
4390		    0,
4391		    SCTP_FROM_SCTP_INDATA,
4392		    __LINE__);
4393	}
4394	stcb->asoc.overall_error_count = 0;
4395	asoc = &stcb->asoc;
4396	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
4397		sctp_log_sack(asoc->last_acked_seq,
4398		    cum_ack,
4399		    0,
4400		    num_seg,
4401		    num_dup,
4402		    SCTP_LOG_NEW_SACK);
4403	}
4404	if ((num_dup) && (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE)) {
4405		uint16_t i;
4406		uint32_t *dupdata, dblock;
4407
4408		for (i = 0; i < num_dup; i++) {
4409			dupdata = (uint32_t *) sctp_m_getptr(m, offset_dup + i * sizeof(uint32_t),
4410			    sizeof(uint32_t), (uint8_t *) & dblock);
4411			if (dupdata == NULL) {
4412				break;
4413			}
4414			sctp_log_fr(*dupdata, 0, 0, SCTP_FR_DUPED);
4415		}
4416	}
4417	if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) {
4418		/* reality check */
4419		if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4420			tp1 = TAILQ_LAST(&asoc->sent_queue,
4421			    sctpchunk_listhead);
4422			send_s = tp1->rec.data.TSN_seq + 1;
4423		} else {
4424			tp1 = NULL;
4425			send_s = asoc->sending_seq;
4426		}
4427		if (SCTP_TSN_GE(cum_ack, send_s)) {
4428			struct mbuf *oper;
4429
4430			/*
4431			 * no way, we have not even sent this TSN out yet.
4432			 * Peer is hopelessly messed up with us.
4433			 */
4434			SCTP_PRINTF("NEW cum_ack:%x send_s:%x is smaller or equal\n",
4435			    cum_ack, send_s);
4436			if (tp1) {
4437				SCTP_PRINTF("Got send_s from tsn:%x + 1 of tp1:%p\n",
4438				    tp1->rec.data.TSN_seq, (void *)tp1);
4439			}
4440	hopeless_peer:
4441			*abort_now = 1;
4442			/* XXX */
4443			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
4444			    0, M_NOWAIT, 1, MT_DATA);
4445			if (oper) {
4446				struct sctp_paramhdr *ph;
4447				uint32_t *ippp;
4448
4449				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
4450				    sizeof(uint32_t);
4451				ph = mtod(oper, struct sctp_paramhdr *);
4452				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
4453				ph->param_length = htons(SCTP_BUF_LEN(oper));
4454				ippp = (uint32_t *) (ph + 1);
4455				*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_25);
4456			}
4457			stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_25;
4458			sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
4459			return;
4460		}
4461	}
4462	/**********************/
4463	/* 1) check the range */
4464	/**********************/
4465	if (SCTP_TSN_GT(asoc->last_acked_seq, last_tsn)) {
4466		/* acking something behind */
4467		return;
4468	}
4469	/* update the Rwnd of the peer */
4470	if (TAILQ_EMPTY(&asoc->sent_queue) &&
4471	    TAILQ_EMPTY(&asoc->send_queue) &&
4472	    (asoc->stream_queue_cnt == 0)) {
4473		/* nothing left on send/sent and strmq */
4474		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) {
4475			sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK,
4476			    asoc->peers_rwnd, 0, 0, a_rwnd);
4477		}
4478		asoc->peers_rwnd = a_rwnd;
4479		if (asoc->sent_queue_retran_cnt) {
4480			asoc->sent_queue_retran_cnt = 0;
4481		}
4482		if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) {
4483			/* SWS sender side engages */
4484			asoc->peers_rwnd = 0;
4485		}
4486		/* stop any timers */
4487		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4488			sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
4489			    stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_26);
4490			net->partial_bytes_acked = 0;
4491			net->flight_size = 0;
4492		}
4493		asoc->total_flight = 0;
4494		asoc->total_flight_count = 0;
4495		return;
4496	}
4497	/*
4498	 * We init netAckSz and netAckSz2 to 0. These are used to track 2
4499	 * things. The total byte count acked is tracked in netAckSz AND
4500	 * netAck2 is used to track the total bytes acked that are un-
4501	 * amibguious and were never retransmitted. We track these on a per
4502	 * destination address basis.
4503	 */
4504	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4505		if (SCTP_TSN_GT(cum_ack, net->cwr_window_tsn)) {
4506			/* Drag along the window_tsn for cwr's */
4507			net->cwr_window_tsn = cum_ack;
4508		}
4509		net->prev_cwnd = net->cwnd;
4510		net->net_ack = 0;
4511		net->net_ack2 = 0;
4512
4513		/*
4514		 * CMT: Reset CUC and Fast recovery algo variables before
4515		 * SACK processing
4516		 */
4517		net->new_pseudo_cumack = 0;
4518		net->will_exit_fast_recovery = 0;
4519		if (stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) {
4520			(*stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) (stcb, net);
4521		}
4522	}
4523	/* process the new consecutive TSN first */
4524	TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
4525		if (SCTP_TSN_GE(last_tsn, tp1->rec.data.TSN_seq)) {
4526			if (tp1->sent != SCTP_DATAGRAM_UNSENT) {
4527				accum_moved = 1;
4528				if (tp1->sent < SCTP_DATAGRAM_ACKED) {
4529					/*
4530					 * If it is less than ACKED, it is
4531					 * now no-longer in flight. Higher
4532					 * values may occur during marking
4533					 */
4534					if ((tp1->whoTo->dest_state &
4535					    SCTP_ADDR_UNCONFIRMED) &&
4536					    (tp1->snd_count < 2)) {
4537						/*
4538						 * If there was no retran
4539						 * and the address is
4540						 * un-confirmed and we sent
4541						 * there and are now
4542						 * sacked.. its confirmed,
4543						 * mark it so.
4544						 */
4545						tp1->whoTo->dest_state &=
4546						    ~SCTP_ADDR_UNCONFIRMED;
4547					}
4548					if (tp1->sent < SCTP_DATAGRAM_RESEND) {
4549						if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
4550							sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA,
4551							    tp1->whoTo->flight_size,
4552							    tp1->book_size,
4553							    (uintptr_t) tp1->whoTo,
4554							    tp1->rec.data.TSN_seq);
4555						}
4556						sctp_flight_size_decrease(tp1);
4557						sctp_total_flight_decrease(stcb, tp1);
4558						if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) {
4559							(*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) (tp1->whoTo,
4560							    tp1);
4561						}
4562					}
4563					tp1->whoTo->net_ack += tp1->send_size;
4564
4565					/* CMT SFR and DAC algos */
4566					this_sack_lowest_newack = tp1->rec.data.TSN_seq;
4567					tp1->whoTo->saw_newack = 1;
4568
4569					if (tp1->snd_count < 2) {
4570						/*
4571						 * True non-retransmited
4572						 * chunk
4573						 */
4574						tp1->whoTo->net_ack2 +=
4575						    tp1->send_size;
4576
4577						/* update RTO too? */
4578						if (tp1->do_rtt) {
4579							if (rto_ok) {
4580								tp1->whoTo->RTO =
4581								    sctp_calculate_rto(stcb,
4582								    asoc, tp1->whoTo,
4583								    &tp1->sent_rcv_time,
4584								    sctp_align_safe_nocopy,
4585								    SCTP_RTT_FROM_DATA);
4586								rto_ok = 0;
4587							}
4588							if (tp1->whoTo->rto_needed == 0) {
4589								tp1->whoTo->rto_needed = 1;
4590							}
4591							tp1->do_rtt = 0;
4592						}
4593					}
4594					/*
4595					 * CMT: CUCv2 algorithm. From the
4596					 * cumack'd TSNs, for each TSN being
4597					 * acked for the first time, set the
4598					 * following variables for the
4599					 * corresp destination.
4600					 * new_pseudo_cumack will trigger a
4601					 * cwnd update.
4602					 * find_(rtx_)pseudo_cumack will
4603					 * trigger search for the next
4604					 * expected (rtx-)pseudo-cumack.
4605					 */
4606					tp1->whoTo->new_pseudo_cumack = 1;
4607					tp1->whoTo->find_pseudo_cumack = 1;
4608					tp1->whoTo->find_rtx_pseudo_cumack = 1;
4609
4610
4611					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
4612						sctp_log_sack(asoc->last_acked_seq,
4613						    cum_ack,
4614						    tp1->rec.data.TSN_seq,
4615						    0,
4616						    0,
4617						    SCTP_LOG_TSN_ACKED);
4618					}
4619					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
4620						sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.TSN_seq, SCTP_CWND_LOG_FROM_SACK);
4621					}
4622				}
4623				if (tp1->sent == SCTP_DATAGRAM_RESEND) {
4624					sctp_ucount_decr(asoc->sent_queue_retran_cnt);
4625#ifdef SCTP_AUDITING_ENABLED
4626					sctp_audit_log(0xB3,
4627					    (asoc->sent_queue_retran_cnt & 0x000000ff));
4628#endif
4629				}
4630				if (tp1->rec.data.chunk_was_revoked) {
4631					/* deflate the cwnd */
4632					tp1->whoTo->cwnd -= tp1->book_size;
4633					tp1->rec.data.chunk_was_revoked = 0;
4634				}
4635				if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) {
4636					tp1->sent = SCTP_DATAGRAM_ACKED;
4637				}
4638			}
4639		} else {
4640			break;
4641		}
4642	}
4643	biggest_tsn_newly_acked = biggest_tsn_acked = last_tsn;
4644	/* always set this up to cum-ack */
4645	asoc->this_sack_highest_gap = last_tsn;
4646
4647	if ((num_seg > 0) || (num_nr_seg > 0)) {
4648
4649		/*
4650		 * CMT: SFR algo (and HTNA) - this_sack_highest_newack has
4651		 * to be greater than the cumack. Also reset saw_newack to 0
4652		 * for all dests.
4653		 */
4654		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4655			net->saw_newack = 0;
4656			net->this_sack_highest_newack = last_tsn;
4657		}
4658
4659		/*
4660		 * thisSackHighestGap will increase while handling NEW
4661		 * segments this_sack_highest_newack will increase while
4662		 * handling NEWLY ACKED chunks. this_sack_lowest_newack is
4663		 * used for CMT DAC algo. saw_newack will also change.
4664		 */
4665		if (sctp_handle_segments(m, &offset_seg, stcb, asoc, last_tsn, &biggest_tsn_acked,
4666		    &biggest_tsn_newly_acked, &this_sack_lowest_newack,
4667		    num_seg, num_nr_seg, &rto_ok)) {
4668			wake_him++;
4669		}
4670		if (SCTP_BASE_SYSCTL(sctp_strict_sacks)) {
4671			/*
4672			 * validate the biggest_tsn_acked in the gap acks if
4673			 * strict adherence is wanted.
4674			 */
4675			if (SCTP_TSN_GE(biggest_tsn_acked, send_s)) {
4676				/*
4677				 * peer is either confused or we are under
4678				 * attack. We must abort.
4679				 */
4680				SCTP_PRINTF("Hopeless peer! biggest_tsn_acked:%x largest seq:%x\n",
4681				    biggest_tsn_acked, send_s);
4682				goto hopeless_peer;
4683			}
4684		}
4685	}
4686	/*******************************************/
4687	/* cancel ALL T3-send timer if accum moved */
4688	/*******************************************/
4689	if (asoc->sctp_cmt_on_off > 0) {
4690		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4691			if (net->new_pseudo_cumack)
4692				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
4693				    stcb, net,
4694				    SCTP_FROM_SCTP_INDATA + SCTP_LOC_27);
4695
4696		}
4697	} else {
4698		if (accum_moved) {
4699			TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4700				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
4701				    stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_28);
4702			}
4703		}
4704	}
4705	/********************************************/
4706	/* drop the acked chunks from the sentqueue */
4707	/********************************************/
4708	asoc->last_acked_seq = cum_ack;
4709
4710	TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) {
4711		if (SCTP_TSN_GT(tp1->rec.data.TSN_seq, cum_ack)) {
4712			break;
4713		}
4714		if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) {
4715			if (asoc->strmout[tp1->rec.data.stream_number].chunks_on_queues > 0) {
4716				asoc->strmout[tp1->rec.data.stream_number].chunks_on_queues--;
4717#ifdef INVARIANTS
4718			} else {
4719				panic("No chunks on the queues for sid %u.", tp1->rec.data.stream_number);
4720#endif
4721			}
4722		}
4723		TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next);
4724		if (PR_SCTP_ENABLED(tp1->flags)) {
4725			if (asoc->pr_sctp_cnt != 0)
4726				asoc->pr_sctp_cnt--;
4727		}
4728		asoc->sent_queue_cnt--;
4729		if (tp1->data) {
4730			/* sa_ignore NO_NULL_CHK */
4731			sctp_free_bufspace(stcb, asoc, tp1, 1);
4732			sctp_m_freem(tp1->data);
4733			tp1->data = NULL;
4734			if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(tp1->flags)) {
4735				asoc->sent_queue_cnt_removeable--;
4736			}
4737		}
4738		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) {
4739			sctp_log_sack(asoc->last_acked_seq,
4740			    cum_ack,
4741			    tp1->rec.data.TSN_seq,
4742			    0,
4743			    0,
4744			    SCTP_LOG_FREE_SENT);
4745		}
4746		sctp_free_a_chunk(stcb, tp1, SCTP_SO_NOT_LOCKED);
4747		wake_him++;
4748	}
4749	if (TAILQ_EMPTY(&asoc->sent_queue) && (asoc->total_flight > 0)) {
4750#ifdef INVARIANTS
4751		panic("Warning flight size is postive and should be 0");
4752#else
4753		SCTP_PRINTF("Warning flight size incorrect should be 0 is %d\n",
4754		    asoc->total_flight);
4755#endif
4756		asoc->total_flight = 0;
4757	}
4758	/* sa_ignore NO_NULL_CHK */
4759	if ((wake_him) && (stcb->sctp_socket)) {
4760#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4761		struct socket *so;
4762
4763#endif
4764		SOCKBUF_LOCK(&stcb->sctp_socket->so_snd);
4765		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) {
4766			sctp_wakeup_log(stcb, wake_him, SCTP_WAKESND_FROM_SACK);
4767		}
4768#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4769		so = SCTP_INP_SO(stcb->sctp_ep);
4770		atomic_add_int(&stcb->asoc.refcnt, 1);
4771		SCTP_TCB_UNLOCK(stcb);
4772		SCTP_SOCKET_LOCK(so, 1);
4773		SCTP_TCB_LOCK(stcb);
4774		atomic_subtract_int(&stcb->asoc.refcnt, 1);
4775		if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) {
4776			/* assoc was freed while we were unlocked */
4777			SCTP_SOCKET_UNLOCK(so, 1);
4778			return;
4779		}
4780#endif
4781		sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket);
4782#if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
4783		SCTP_SOCKET_UNLOCK(so, 1);
4784#endif
4785	} else {
4786		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) {
4787			sctp_wakeup_log(stcb, wake_him, SCTP_NOWAKE_FROM_SACK);
4788		}
4789	}
4790
4791	if (asoc->fast_retran_loss_recovery && accum_moved) {
4792		if (SCTP_TSN_GE(asoc->last_acked_seq, asoc->fast_recovery_tsn)) {
4793			/* Setup so we will exit RFC2582 fast recovery */
4794			will_exit_fast_recovery = 1;
4795		}
4796	}
4797	/*
4798	 * Check for revoked fragments:
4799	 *
4800	 * if Previous sack - Had no frags then we can't have any revoked if
4801	 * Previous sack - Had frag's then - If we now have frags aka
4802	 * num_seg > 0 call sctp_check_for_revoked() to tell if peer revoked
4803	 * some of them. else - The peer revoked all ACKED fragments, since
4804	 * we had some before and now we have NONE.
4805	 */
4806
4807	if (num_seg) {
4808		sctp_check_for_revoked(stcb, asoc, cum_ack, biggest_tsn_acked);
4809		asoc->saw_sack_with_frags = 1;
4810	} else if (asoc->saw_sack_with_frags) {
4811		int cnt_revoked = 0;
4812
4813		/* Peer revoked all dg's marked or acked */
4814		TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
4815			if (tp1->sent == SCTP_DATAGRAM_ACKED) {
4816				tp1->sent = SCTP_DATAGRAM_SENT;
4817				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
4818					sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE,
4819					    tp1->whoTo->flight_size,
4820					    tp1->book_size,
4821					    (uintptr_t) tp1->whoTo,
4822					    tp1->rec.data.TSN_seq);
4823				}
4824				sctp_flight_size_increase(tp1);
4825				sctp_total_flight_increase(stcb, tp1);
4826				tp1->rec.data.chunk_was_revoked = 1;
4827				/*
4828				 * To ensure that this increase in
4829				 * flightsize, which is artificial, does not
4830				 * throttle the sender, we also increase the
4831				 * cwnd artificially.
4832				 */
4833				tp1->whoTo->cwnd += tp1->book_size;
4834				cnt_revoked++;
4835			}
4836		}
4837		if (cnt_revoked) {
4838			reneged_all = 1;
4839		}
4840		asoc->saw_sack_with_frags = 0;
4841	}
4842	if (num_nr_seg > 0)
4843		asoc->saw_sack_with_nr_frags = 1;
4844	else
4845		asoc->saw_sack_with_nr_frags = 0;
4846
4847	/* JRS - Use the congestion control given in the CC module */
4848	if (ecne_seen == 0) {
4849		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4850			if (net->net_ack2 > 0) {
4851				/*
4852				 * Karn's rule applies to clearing error
4853				 * count, this is optional.
4854				 */
4855				net->error_count = 0;
4856				if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
4857					/* addr came good */
4858					net->dest_state |= SCTP_ADDR_REACHABLE;
4859					sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb,
4860					    0, (void *)net, SCTP_SO_NOT_LOCKED);
4861				}
4862				if (net == stcb->asoc.primary_destination) {
4863					if (stcb->asoc.alternate) {
4864						/*
4865						 * release the alternate,
4866						 * primary is good
4867						 */
4868						sctp_free_remote_addr(stcb->asoc.alternate);
4869						stcb->asoc.alternate = NULL;
4870					}
4871				}
4872				if (net->dest_state & SCTP_ADDR_PF) {
4873					net->dest_state &= ~SCTP_ADDR_PF;
4874					sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_3);
4875					sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
4876					asoc->cc_functions.sctp_cwnd_update_exit_pf(stcb, net);
4877					/* Done with this net */
4878					net->net_ack = 0;
4879				}
4880				/* restore any doubled timers */
4881				net->RTO = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
4882				if (net->RTO < stcb->asoc.minrto) {
4883					net->RTO = stcb->asoc.minrto;
4884				}
4885				if (net->RTO > stcb->asoc.maxrto) {
4886					net->RTO = stcb->asoc.maxrto;
4887				}
4888			}
4889		}
4890		asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, accum_moved, reneged_all, will_exit_fast_recovery);
4891	}
4892	if (TAILQ_EMPTY(&asoc->sent_queue)) {
4893		/* nothing left in-flight */
4894		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4895			/* stop all timers */
4896			sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
4897			    stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_30);
4898			net->flight_size = 0;
4899			net->partial_bytes_acked = 0;
4900		}
4901		asoc->total_flight = 0;
4902		asoc->total_flight_count = 0;
4903	}
4904	/**********************************/
4905	/* Now what about shutdown issues */
4906	/**********************************/
4907	if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) {
4908		/* nothing left on sendqueue.. consider done */
4909		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) {
4910			sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK,
4911			    asoc->peers_rwnd, 0, 0, a_rwnd);
4912		}
4913		asoc->peers_rwnd = a_rwnd;
4914		if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) {
4915			/* SWS sender side engages */
4916			asoc->peers_rwnd = 0;
4917		}
4918		/* clean up */
4919		if ((asoc->stream_queue_cnt == 1) &&
4920		    ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) ||
4921		    (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED)) &&
4922		    (asoc->locked_on_sending)
4923		    ) {
4924			struct sctp_stream_queue_pending *sp;
4925
4926			/*
4927			 * I may be in a state where we got all across.. but
4928			 * cannot write more due to a shutdown... we abort
4929			 * since the user did not indicate EOR in this case.
4930			 */
4931			sp = TAILQ_LAST(&((asoc->locked_on_sending)->outqueue),
4932			    sctp_streamhead);
4933			if ((sp) && (sp->length == 0)) {
4934				asoc->locked_on_sending = NULL;
4935				if (sp->msg_is_complete) {
4936					asoc->stream_queue_cnt--;
4937				} else {
4938					asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT;
4939					asoc->stream_queue_cnt--;
4940				}
4941			}
4942		}
4943		if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) &&
4944		    (asoc->stream_queue_cnt == 0)) {
4945			if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) {
4946				/* Need to abort here */
4947				struct mbuf *oper;
4948
4949		abort_out_now:
4950				*abort_now = 1;
4951				/* XXX */
4952				oper = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
4953				    0, M_NOWAIT, 1, MT_DATA);
4954				if (oper) {
4955					struct sctp_paramhdr *ph;
4956
4957					SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr);
4958					ph = mtod(oper, struct sctp_paramhdr *);
4959					ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
4960					ph->param_length = htons(SCTP_BUF_LEN(oper));
4961				}
4962				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_31;
4963				sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
4964				return;
4965			} else {
4966				struct sctp_nets *netp;
4967
4968				if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
4969				    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
4970					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
4971				}
4972				SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
4973				SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
4974				sctp_stop_timers_for_shutdown(stcb);
4975				if (asoc->alternate) {
4976					netp = asoc->alternate;
4977				} else {
4978					netp = asoc->primary_destination;
4979				}
4980				sctp_send_shutdown(stcb, netp);
4981				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
4982				    stcb->sctp_ep, stcb, netp);
4983				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
4984				    stcb->sctp_ep, stcb, netp);
4985			}
4986			return;
4987		} else if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) &&
4988		    (asoc->stream_queue_cnt == 0)) {
4989			struct sctp_nets *netp;
4990
4991			if (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT) {
4992				goto abort_out_now;
4993			}
4994			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
4995			SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_ACK_SENT);
4996			SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
4997			sctp_stop_timers_for_shutdown(stcb);
4998			if (asoc->alternate) {
4999				netp = asoc->alternate;
5000			} else {
5001				netp = asoc->primary_destination;
5002			}
5003			sctp_send_shutdown_ack(stcb, netp);
5004			sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK,
5005			    stcb->sctp_ep, stcb, netp);
5006			return;
5007		}
5008	}
5009	/*
5010	 * Now here we are going to recycle net_ack for a different use...
5011	 * HEADS UP.
5012	 */
5013	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
5014		net->net_ack = 0;
5015	}
5016
5017	/*
5018	 * CMT DAC algorithm: If SACK DAC flag was 0, then no extra marking
5019	 * to be done. Setting this_sack_lowest_newack to the cum_ack will
5020	 * automatically ensure that.
5021	 */
5022	if ((asoc->sctp_cmt_on_off > 0) &&
5023	    SCTP_BASE_SYSCTL(sctp_cmt_use_dac) &&
5024	    (cmt_dac_flag == 0)) {
5025		this_sack_lowest_newack = cum_ack;
5026	}
5027	if ((num_seg > 0) || (num_nr_seg > 0)) {
5028		sctp_strike_gap_ack_chunks(stcb, asoc, biggest_tsn_acked,
5029		    biggest_tsn_newly_acked, this_sack_lowest_newack, accum_moved);
5030	}
5031	/* JRS - Use the congestion control given in the CC module */
5032	asoc->cc_functions.sctp_cwnd_update_after_fr(stcb, asoc);
5033
5034	/* Now are we exiting loss recovery ? */
5035	if (will_exit_fast_recovery) {
5036		/* Ok, we must exit fast recovery */
5037		asoc->fast_retran_loss_recovery = 0;
5038	}
5039	if ((asoc->sat_t3_loss_recovery) &&
5040	    SCTP_TSN_GE(asoc->last_acked_seq, asoc->sat_t3_recovery_tsn)) {
5041		/* end satellite t3 loss recovery */
5042		asoc->sat_t3_loss_recovery = 0;
5043	}
5044	/*
5045	 * CMT Fast recovery
5046	 */
5047	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
5048		if (net->will_exit_fast_recovery) {
5049			/* Ok, we must exit fast recovery */
5050			net->fast_retran_loss_recovery = 0;
5051		}
5052	}
5053
5054	/* Adjust and set the new rwnd value */
5055	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) {
5056		sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK,
5057		    asoc->peers_rwnd, asoc->total_flight, (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)), a_rwnd);
5058	}
5059	asoc->peers_rwnd = sctp_sbspace_sub(a_rwnd,
5060	    (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))));
5061	if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) {
5062		/* SWS sender side engages */
5063		asoc->peers_rwnd = 0;
5064	}
5065	if (asoc->peers_rwnd > old_rwnd) {
5066		win_probe_recovery = 1;
5067	}
5068	/*
5069	 * Now we must setup so we have a timer up for anyone with
5070	 * outstanding data.
5071	 */
5072	done_once = 0;
5073again:
5074	j = 0;
5075	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
5076		if (win_probe_recovery && (net->window_probe)) {
5077			win_probe_recovered = 1;
5078			/*-
5079			 * Find first chunk that was used with
5080			 * window probe and clear the event. Put
5081			 * it back into the send queue as if has
5082			 * not been sent.
5083			 */
5084			TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
5085				if (tp1->window_probe) {
5086					sctp_window_probe_recovery(stcb, asoc, tp1);
5087					break;
5088				}
5089			}
5090		}
5091		if (net->flight_size) {
5092			j++;
5093			if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) {
5094				sctp_timer_start(SCTP_TIMER_TYPE_SEND,
5095				    stcb->sctp_ep, stcb, net);
5096			}
5097			if (net->window_probe) {
5098				net->window_probe = 0;
5099			}
5100		} else {
5101			if (net->window_probe) {
5102				/*
5103				 * In window probes we must assure a timer
5104				 * is still running there
5105				 */
5106				if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) {
5107					sctp_timer_start(SCTP_TIMER_TYPE_SEND,
5108					    stcb->sctp_ep, stcb, net);
5109
5110				}
5111			} else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) {
5112				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
5113				    stcb, net,
5114				    SCTP_FROM_SCTP_INDATA + SCTP_LOC_22);
5115			}
5116		}
5117	}
5118	if ((j == 0) &&
5119	    (!TAILQ_EMPTY(&asoc->sent_queue)) &&
5120	    (asoc->sent_queue_retran_cnt == 0) &&
5121	    (win_probe_recovered == 0) &&
5122	    (done_once == 0)) {
5123		/*
5124		 * huh, this should not happen unless all packets are
5125		 * PR-SCTP and marked to skip of course.
5126		 */
5127		if (sctp_fs_audit(asoc)) {
5128			TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
5129				net->flight_size = 0;
5130			}
5131			asoc->total_flight = 0;
5132			asoc->total_flight_count = 0;
5133			asoc->sent_queue_retran_cnt = 0;
5134			TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) {
5135				if (tp1->sent < SCTP_DATAGRAM_RESEND) {
5136					sctp_flight_size_increase(tp1);
5137					sctp_total_flight_increase(stcb, tp1);
5138				} else if (tp1->sent == SCTP_DATAGRAM_RESEND) {
5139					sctp_ucount_incr(asoc->sent_queue_retran_cnt);
5140				}
5141			}
5142		}
5143		done_once = 1;
5144		goto again;
5145	}
5146	/*********************************************/
5147	/* Here we perform PR-SCTP procedures        */
5148	/* (section 4.2)                             */
5149	/*********************************************/
5150	/* C1. update advancedPeerAckPoint */
5151	if (SCTP_TSN_GT(cum_ack, asoc->advanced_peer_ack_point)) {
5152		asoc->advanced_peer_ack_point = cum_ack;
5153	}
5154	/* C2. try to further move advancedPeerAckPoint ahead */
5155	if ((asoc->peer_supports_prsctp) && (asoc->pr_sctp_cnt > 0)) {
5156		struct sctp_tmit_chunk *lchk;
5157		uint32_t old_adv_peer_ack_point;
5158
5159		old_adv_peer_ack_point = asoc->advanced_peer_ack_point;
5160		lchk = sctp_try_advance_peer_ack_point(stcb, asoc);
5161		/* C3. See if we need to send a Fwd-TSN */
5162		if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cum_ack)) {
5163			/*
5164			 * ISSUE with ECN, see FWD-TSN processing.
5165			 */
5166			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) {
5167				sctp_misc_ints(SCTP_FWD_TSN_CHECK,
5168				    0xee, cum_ack, asoc->advanced_peer_ack_point,
5169				    old_adv_peer_ack_point);
5170			}
5171			if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) {
5172				send_forward_tsn(stcb, asoc);
5173			} else if (lchk) {
5174				/* try to FR fwd-tsn's that get lost too */
5175				if (lchk->rec.data.fwd_tsn_cnt >= 3) {
5176					send_forward_tsn(stcb, asoc);
5177				}
5178			}
5179		}
5180		if (lchk) {
5181			/* Assure a timer is up */
5182			sctp_timer_start(SCTP_TIMER_TYPE_SEND,
5183			    stcb->sctp_ep, stcb, lchk->whoTo);
5184		}
5185	}
5186	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) {
5187		sctp_misc_ints(SCTP_SACK_RWND_UPDATE,
5188		    a_rwnd,
5189		    stcb->asoc.peers_rwnd,
5190		    stcb->asoc.total_flight,
5191		    stcb->asoc.total_output_queue_size);
5192	}
5193}
5194
5195void
5196sctp_update_acked(struct sctp_tcb *stcb, struct sctp_shutdown_chunk *cp, int *abort_flag)
5197{
5198	/* Copy cum-ack */
5199	uint32_t cum_ack, a_rwnd;
5200
5201	cum_ack = ntohl(cp->cumulative_tsn_ack);
5202	/* Arrange so a_rwnd does NOT change */
5203	a_rwnd = stcb->asoc.peers_rwnd + stcb->asoc.total_flight;
5204
5205	/* Now call the express sack handling */
5206	sctp_express_handle_sack(stcb, cum_ack, a_rwnd, abort_flag, 0);
5207}
5208
5209static void
5210sctp_kick_prsctp_reorder_queue(struct sctp_tcb *stcb,
5211    struct sctp_stream_in *strmin)
5212{
5213	struct sctp_queued_to_read *ctl, *nctl;
5214	struct sctp_association *asoc;
5215	uint16_t tt;
5216
5217	asoc = &stcb->asoc;
5218	tt = strmin->last_sequence_delivered;
5219	/*
5220	 * First deliver anything prior to and including the stream no that
5221	 * came in
5222	 */
5223	TAILQ_FOREACH_SAFE(ctl, &strmin->inqueue, next, nctl) {
5224		if (SCTP_SSN_GE(tt, ctl->sinfo_ssn)) {
5225			/* this is deliverable now */
5226			TAILQ_REMOVE(&strmin->inqueue, ctl, next);
5227			/* subtract pending on streams */
5228			asoc->size_on_all_streams -= ctl->length;
5229			sctp_ucount_decr(asoc->cnt_on_all_streams);
5230			/* deliver it to at least the delivery-q */
5231			if (stcb->sctp_socket) {
5232				sctp_mark_non_revokable(asoc, ctl->sinfo_tsn);
5233				sctp_add_to_readq(stcb->sctp_ep, stcb,
5234				    ctl,
5235				    &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_HELD, SCTP_SO_NOT_LOCKED);
5236			}
5237		} else {
5238			/* no more delivery now. */
5239			break;
5240		}
5241	}
5242	/*
5243	 * now we must deliver things in queue the normal way  if any are
5244	 * now ready.
5245	 */
5246	tt = strmin->last_sequence_delivered + 1;
5247	TAILQ_FOREACH_SAFE(ctl, &strmin->inqueue, next, nctl) {
5248		if (tt == ctl->sinfo_ssn) {
5249			/* this is deliverable now */
5250			TAILQ_REMOVE(&strmin->inqueue, ctl, next);
5251			/* subtract pending on streams */
5252			asoc->size_on_all_streams -= ctl->length;
5253			sctp_ucount_decr(asoc->cnt_on_all_streams);
5254			/* deliver it to at least the delivery-q */
5255			strmin->last_sequence_delivered = ctl->sinfo_ssn;
5256			if (stcb->sctp_socket) {
5257				sctp_mark_non_revokable(asoc, ctl->sinfo_tsn);
5258				sctp_add_to_readq(stcb->sctp_ep, stcb,
5259				    ctl,
5260				    &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_HELD, SCTP_SO_NOT_LOCKED);
5261
5262			}
5263			tt = strmin->last_sequence_delivered + 1;
5264		} else {
5265			break;
5266		}
5267	}
5268}
5269
5270static void
5271sctp_flush_reassm_for_str_seq(struct sctp_tcb *stcb,
5272    struct sctp_association *asoc,
5273    uint16_t stream, uint16_t seq)
5274{
5275	struct sctp_tmit_chunk *chk, *nchk;
5276
5277	/* For each one on here see if we need to toss it */
5278	/*
5279	 * For now large messages held on the reasmqueue that are complete
5280	 * will be tossed too. We could in theory do more work to spin
5281	 * through and stop after dumping one msg aka seeing the start of a
5282	 * new msg at the head, and call the delivery function... to see if
5283	 * it can be delivered... But for now we just dump everything on the
5284	 * queue.
5285	 */
5286	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5287		/*
5288		 * Do not toss it if on a different stream or marked for
5289		 * unordered delivery in which case the stream sequence
5290		 * number has no meaning.
5291		 */
5292		if ((chk->rec.data.stream_number != stream) ||
5293		    ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == SCTP_DATA_UNORDERED)) {
5294			continue;
5295		}
5296		if (chk->rec.data.stream_seq == seq) {
5297			/* It needs to be tossed */
5298			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5299			if (SCTP_TSN_GT(chk->rec.data.TSN_seq, asoc->tsn_last_delivered)) {
5300				asoc->tsn_last_delivered = chk->rec.data.TSN_seq;
5301				asoc->str_of_pdapi = chk->rec.data.stream_number;
5302				asoc->ssn_of_pdapi = chk->rec.data.stream_seq;
5303				asoc->fragment_flags = chk->rec.data.rcv_flags;
5304			}
5305			asoc->size_on_reasm_queue -= chk->send_size;
5306			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5307
5308			/* Clear up any stream problem */
5309			if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) != SCTP_DATA_UNORDERED &&
5310			    SCTP_SSN_GT(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered)) {
5311				/*
5312				 * We must dump forward this streams
5313				 * sequence number if the chunk is not
5314				 * unordered that is being skipped. There is
5315				 * a chance that if the peer does not
5316				 * include the last fragment in its FWD-TSN
5317				 * we WILL have a problem here since you
5318				 * would have a partial chunk in queue that
5319				 * may not be deliverable. Also if a Partial
5320				 * delivery API as started the user may get
5321				 * a partial chunk. The next read returning
5322				 * a new chunk... really ugly but I see no
5323				 * way around it! Maybe a notify??
5324				 */
5325				asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered = chk->rec.data.stream_seq;
5326			}
5327			if (chk->data) {
5328				sctp_m_freem(chk->data);
5329				chk->data = NULL;
5330			}
5331			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
5332		} else if (SCTP_SSN_GT(chk->rec.data.stream_seq, seq)) {
5333			/*
5334			 * If the stream_seq is > than the purging one, we
5335			 * are done
5336			 */
5337			break;
5338		}
5339	}
5340}
5341
5342
5343void
5344sctp_handle_forward_tsn(struct sctp_tcb *stcb,
5345    struct sctp_forward_tsn_chunk *fwd,
5346    int *abort_flag, struct mbuf *m, int offset)
5347{
5348	/* The pr-sctp fwd tsn */
5349	/*
5350	 * here we will perform all the data receiver side steps for
5351	 * processing FwdTSN, as required in by pr-sctp draft:
5352	 *
5353	 * Assume we get FwdTSN(x):
5354	 *
5355	 * 1) update local cumTSN to x 2) try to further advance cumTSN to x +
5356	 * others we have 3) examine and update re-ordering queue on
5357	 * pr-in-streams 4) clean up re-assembly queue 5) Send a sack to
5358	 * report where we are.
5359	 */
5360	struct sctp_association *asoc;
5361	uint32_t new_cum_tsn, gap;
5362	unsigned int i, fwd_sz, m_size;
5363	uint32_t str_seq;
5364	struct sctp_stream_in *strm;
5365	struct sctp_tmit_chunk *chk, *nchk;
5366	struct sctp_queued_to_read *ctl, *sv;
5367
5368	asoc = &stcb->asoc;
5369	if ((fwd_sz = ntohs(fwd->ch.chunk_length)) < sizeof(struct sctp_forward_tsn_chunk)) {
5370		SCTPDBG(SCTP_DEBUG_INDATA1,
5371		    "Bad size too small/big fwd-tsn\n");
5372		return;
5373	}
5374	m_size = (stcb->asoc.mapping_array_size << 3);
5375	/*************************************************************/
5376	/* 1. Here we update local cumTSN and shift the bitmap array */
5377	/*************************************************************/
5378	new_cum_tsn = ntohl(fwd->new_cumulative_tsn);
5379
5380	if (SCTP_TSN_GE(asoc->cumulative_tsn, new_cum_tsn)) {
5381		/* Already got there ... */
5382		return;
5383	}
5384	/*
5385	 * now we know the new TSN is more advanced, let's find the actual
5386	 * gap
5387	 */
5388	SCTP_CALC_TSN_TO_GAP(gap, new_cum_tsn, asoc->mapping_array_base_tsn);
5389	asoc->cumulative_tsn = new_cum_tsn;
5390	if (gap >= m_size) {
5391		if ((long)gap > sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv)) {
5392			struct mbuf *oper;
5393
5394			/*
5395			 * out of range (of single byte chunks in the rwnd I
5396			 * give out). This must be an attacker.
5397			 */
5398			*abort_flag = 1;
5399			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + 3 * sizeof(uint32_t)),
5400			    0, M_NOWAIT, 1, MT_DATA);
5401			if (oper) {
5402				struct sctp_paramhdr *ph;
5403				uint32_t *ippp;
5404
5405				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
5406				    (sizeof(uint32_t) * 3);
5407				ph = mtod(oper, struct sctp_paramhdr *);
5408				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
5409				ph->param_length = htons(SCTP_BUF_LEN(oper));
5410				ippp = (uint32_t *) (ph + 1);
5411				*ippp = htonl(SCTP_FROM_SCTP_INDATA + SCTP_LOC_33);
5412				ippp++;
5413				*ippp = asoc->highest_tsn_inside_map;
5414				ippp++;
5415				*ippp = new_cum_tsn;
5416			}
5417			stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_33;
5418			sctp_abort_an_association(stcb->sctp_ep, stcb, oper, SCTP_SO_NOT_LOCKED);
5419			return;
5420		}
5421		SCTP_STAT_INCR(sctps_fwdtsn_map_over);
5422
5423		memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size);
5424		asoc->mapping_array_base_tsn = new_cum_tsn + 1;
5425		asoc->highest_tsn_inside_map = new_cum_tsn;
5426
5427		memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size);
5428		asoc->highest_tsn_inside_nr_map = new_cum_tsn;
5429
5430		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
5431			sctp_log_map(0, 3, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT);
5432		}
5433	} else {
5434		SCTP_TCB_LOCK_ASSERT(stcb);
5435		for (i = 0; i <= gap; i++) {
5436			if (!SCTP_IS_TSN_PRESENT(asoc->mapping_array, i) &&
5437			    !SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, i)) {
5438				SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, i);
5439				if (SCTP_TSN_GT(asoc->mapping_array_base_tsn + i, asoc->highest_tsn_inside_nr_map)) {
5440					asoc->highest_tsn_inside_nr_map = asoc->mapping_array_base_tsn + i;
5441				}
5442			}
5443		}
5444	}
5445	/*************************************************************/
5446	/* 2. Clear up re-assembly queue                             */
5447	/*************************************************************/
5448	/*
5449	 * First service it if pd-api is up, just in case we can progress it
5450	 * forward
5451	 */
5452	if (asoc->fragmented_delivery_inprogress) {
5453		sctp_service_reassembly(stcb, asoc);
5454	}
5455	/* For each one on here see if we need to toss it */
5456	/*
5457	 * For now large messages held on the reasmqueue that are complete
5458	 * will be tossed too. We could in theory do more work to spin
5459	 * through and stop after dumping one msg aka seeing the start of a
5460	 * new msg at the head, and call the delivery function... to see if
5461	 * it can be delivered... But for now we just dump everything on the
5462	 * queue.
5463	 */
5464	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5465		if (SCTP_TSN_GE(new_cum_tsn, chk->rec.data.TSN_seq)) {
5466			/* It needs to be tossed */
5467			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5468			if (SCTP_TSN_GT(chk->rec.data.TSN_seq, asoc->tsn_last_delivered)) {
5469				asoc->tsn_last_delivered = chk->rec.data.TSN_seq;
5470				asoc->str_of_pdapi = chk->rec.data.stream_number;
5471				asoc->ssn_of_pdapi = chk->rec.data.stream_seq;
5472				asoc->fragment_flags = chk->rec.data.rcv_flags;
5473			}
5474			asoc->size_on_reasm_queue -= chk->send_size;
5475			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5476
5477			/* Clear up any stream problem */
5478			if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) != SCTP_DATA_UNORDERED &&
5479			    SCTP_SSN_GT(chk->rec.data.stream_seq, asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered)) {
5480				/*
5481				 * We must dump forward this streams
5482				 * sequence number if the chunk is not
5483				 * unordered that is being skipped. There is
5484				 * a chance that if the peer does not
5485				 * include the last fragment in its FWD-TSN
5486				 * we WILL have a problem here since you
5487				 * would have a partial chunk in queue that
5488				 * may not be deliverable. Also if a Partial
5489				 * delivery API as started the user may get
5490				 * a partial chunk. The next read returning
5491				 * a new chunk... really ugly but I see no
5492				 * way around it! Maybe a notify??
5493				 */
5494				asoc->strmin[chk->rec.data.stream_number].last_sequence_delivered = chk->rec.data.stream_seq;
5495			}
5496			if (chk->data) {
5497				sctp_m_freem(chk->data);
5498				chk->data = NULL;
5499			}
5500			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
5501		} else {
5502			/*
5503			 * Ok we have gone beyond the end of the fwd-tsn's
5504			 * mark.
5505			 */
5506			break;
5507		}
5508	}
5509	/*******************************************************/
5510	/* 3. Update the PR-stream re-ordering queues and fix  */
5511	/* delivery issues as needed.                       */
5512	/*******************************************************/
5513	fwd_sz -= sizeof(*fwd);
5514	if (m && fwd_sz) {
5515		/* New method. */
5516		unsigned int num_str;
5517		struct sctp_strseq *stseq, strseqbuf;
5518
5519		offset += sizeof(*fwd);
5520
5521		SCTP_INP_READ_LOCK(stcb->sctp_ep);
5522		num_str = fwd_sz / sizeof(struct sctp_strseq);
5523		for (i = 0; i < num_str; i++) {
5524			uint16_t st;
5525
5526			stseq = (struct sctp_strseq *)sctp_m_getptr(m, offset,
5527			    sizeof(struct sctp_strseq),
5528			    (uint8_t *) & strseqbuf);
5529			offset += sizeof(struct sctp_strseq);
5530			if (stseq == NULL) {
5531				break;
5532			}
5533			/* Convert */
5534			st = ntohs(stseq->stream);
5535			stseq->stream = st;
5536			st = ntohs(stseq->sequence);
5537			stseq->sequence = st;
5538
5539			/* now process */
5540
5541			/*
5542			 * Ok we now look for the stream/seq on the read
5543			 * queue where its not all delivered. If we find it
5544			 * we transmute the read entry into a PDI_ABORTED.
5545			 */
5546			if (stseq->stream >= asoc->streamincnt) {
5547				/* screwed up streams, stop!  */
5548				break;
5549			}
5550			if ((asoc->str_of_pdapi == stseq->stream) &&
5551			    (asoc->ssn_of_pdapi == stseq->sequence)) {
5552				/*
5553				 * If this is the one we were partially
5554				 * delivering now then we no longer are.
5555				 * Note this will change with the reassembly
5556				 * re-write.
5557				 */
5558				asoc->fragmented_delivery_inprogress = 0;
5559			}
5560			sctp_flush_reassm_for_str_seq(stcb, asoc, stseq->stream, stseq->sequence);
5561			TAILQ_FOREACH(ctl, &stcb->sctp_ep->read_queue, next) {
5562				if ((ctl->sinfo_stream == stseq->stream) &&
5563				    (ctl->sinfo_ssn == stseq->sequence)) {
5564					str_seq = (stseq->stream << 16) | stseq->sequence;
5565					ctl->end_added = 1;
5566					ctl->pdapi_aborted = 1;
5567					sv = stcb->asoc.control_pdapi;
5568					stcb->asoc.control_pdapi = ctl;
5569					sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
5570					    stcb,
5571					    SCTP_PARTIAL_DELIVERY_ABORTED,
5572					    (void *)&str_seq,
5573					    SCTP_SO_NOT_LOCKED);
5574					stcb->asoc.control_pdapi = sv;
5575					break;
5576				} else if ((ctl->sinfo_stream == stseq->stream) &&
5577				    SCTP_SSN_GT(ctl->sinfo_ssn, stseq->sequence)) {
5578					/* We are past our victim SSN */
5579					break;
5580				}
5581			}
5582			strm = &asoc->strmin[stseq->stream];
5583			if (SCTP_SSN_GT(stseq->sequence, strm->last_sequence_delivered)) {
5584				/* Update the sequence number */
5585				strm->last_sequence_delivered = stseq->sequence;
5586			}
5587			/* now kick the stream the new way */
5588			/* sa_ignore NO_NULL_CHK */
5589			sctp_kick_prsctp_reorder_queue(stcb, strm);
5590		}
5591		SCTP_INP_READ_UNLOCK(stcb->sctp_ep);
5592	}
5593	/*
5594	 * Now slide thing forward.
5595	 */
5596	sctp_slide_mapping_arrays(stcb);
5597
5598	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
5599		/* now lets kick out and check for more fragmented delivery */
5600		/* sa_ignore NO_NULL_CHK */
5601		sctp_deliver_reasm_check(stcb, &stcb->asoc);
5602	}
5603}
5604