xref: /freebsd/sys/netinet/sctp_cc_functions.c (revision 7e00348e7605b9906601438008341ffc37c00e2c)
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 #include <netinet/sctp_auth.h>
48 #include <netinet/sctp_asconf.h>
49 #include <netinet/sctp_dtrace_declare.h>
50 
51 #define SHIFT_MPTCP_MULTI_N 40
52 #define SHIFT_MPTCP_MULTI_Z 16
53 #define SHIFT_MPTCP_MULTI 8
54 
55 static void
56 sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
57 {
58 	struct sctp_association *assoc;
59 	uint32_t cwnd_in_mtu;
60 
61 	assoc = &stcb->asoc;
62 	cwnd_in_mtu = SCTP_BASE_SYSCTL(sctp_initial_cwnd);
63 	if (cwnd_in_mtu == 0) {
64 		/* Using 0 means that the value of RFC 4960 is used. */
65 		net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
66 	} else {
67 		/*
68 		 * We take the minimum of the burst limit and the initial
69 		 * congestion window.
70 		 */
71 		if ((assoc->max_burst > 0) && (cwnd_in_mtu > assoc->max_burst))
72 			cwnd_in_mtu = assoc->max_burst;
73 		net->cwnd = (net->mtu - sizeof(struct sctphdr)) * cwnd_in_mtu;
74 	}
75 	if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) ||
76 	    (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2)) {
77 		/* In case of resource pooling initialize appropriately */
78 		net->cwnd /= assoc->numnets;
79 		if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) {
80 			net->cwnd = net->mtu - sizeof(struct sctphdr);
81 		}
82 	}
83 	net->ssthresh = assoc->peers_rwnd;
84 	SDT_PROBE(sctp, cwnd, net, init,
85 	    stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net,
86 	    0, net->cwnd);
87 	if (SCTP_BASE_SYSCTL(sctp_logging_level) &
88 	    (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
89 		sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
90 	}
91 }
92 
93 static void
94 sctp_cwnd_update_after_fr(struct sctp_tcb *stcb,
95     struct sctp_association *asoc)
96 {
97 	struct sctp_nets *net;
98 	uint32_t t_ssthresh, t_cwnd;
99 	uint64_t t_ucwnd_sbw;
100 
101 	/* MT FIXME: Don't compute this over and over again */
102 	t_ssthresh = 0;
103 	t_cwnd = 0;
104 	t_ucwnd_sbw = 0;
105 	if ((asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) ||
106 	    (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2)) {
107 		TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
108 			t_ssthresh += net->ssthresh;
109 			t_cwnd += net->cwnd;
110 			if (net->lastsa > 0) {
111 				t_ucwnd_sbw += (uint64_t) net->cwnd / (uint64_t) net->lastsa;
112 			}
113 		}
114 		if (t_ucwnd_sbw == 0) {
115 			t_ucwnd_sbw = 1;
116 		}
117 	}
118 	/*-
119 	 * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) &&
120 	 * (net->fast_retran_loss_recovery == 0)))
121 	 */
122 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
123 		if ((asoc->fast_retran_loss_recovery == 0) ||
124 		    (asoc->sctp_cmt_on_off > 0)) {
125 			/* out of a RFC2582 Fast recovery window? */
126 			if (net->net_ack > 0) {
127 				/*
128 				 * per section 7.2.3, are there any
129 				 * destinations that had a fast retransmit
130 				 * to them. If so what we need to do is
131 				 * adjust ssthresh and cwnd.
132 				 */
133 				struct sctp_tmit_chunk *lchk;
134 				int old_cwnd = net->cwnd;
135 
136 				if ((asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) ||
137 				    (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2)) {
138 					if (asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) {
139 						net->ssthresh = (uint32_t) (((uint64_t) 4 *
140 						    (uint64_t) net->mtu *
141 						    (uint64_t) net->ssthresh) /
142 						    (uint64_t) t_ssthresh);
143 
144 					}
145 					if (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2) {
146 						uint32_t srtt;
147 
148 						srtt = net->lastsa;
149 						/*
150 						 * lastsa>>3;  we don't need
151 						 * to devide ...
152 						 */
153 						if (srtt == 0) {
154 							srtt = 1;
155 						}
156 						/*
157 						 * Short Version => Equal to
158 						 * Contel Version MBe
159 						 */
160 						net->ssthresh = (uint32_t) (((uint64_t) 4 *
161 						    (uint64_t) net->mtu *
162 						    (uint64_t) net->cwnd) /
163 						    ((uint64_t) srtt *
164 						    t_ucwnd_sbw));
165 						 /* INCREASE FACTOR */ ;
166 					}
167 					if ((net->cwnd > t_cwnd / 2) &&
168 					    (net->ssthresh < net->cwnd - t_cwnd / 2)) {
169 						net->ssthresh = net->cwnd - t_cwnd / 2;
170 					}
171 					if (net->ssthresh < net->mtu) {
172 						net->ssthresh = net->mtu;
173 					}
174 				} else {
175 					net->ssthresh = net->cwnd / 2;
176 					if (net->ssthresh < (net->mtu * 2)) {
177 						net->ssthresh = 2 * net->mtu;
178 					}
179 				}
180 				net->cwnd = net->ssthresh;
181 				SDT_PROBE(sctp, cwnd, net, fr,
182 				    stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net,
183 				    old_cwnd, net->cwnd);
184 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
185 					sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
186 					    SCTP_CWND_LOG_FROM_FR);
187 				}
188 				lchk = TAILQ_FIRST(&asoc->send_queue);
189 
190 				net->partial_bytes_acked = 0;
191 				/* Turn on fast recovery window */
192 				asoc->fast_retran_loss_recovery = 1;
193 				if (lchk == NULL) {
194 					/* Mark end of the window */
195 					asoc->fast_recovery_tsn = asoc->sending_seq - 1;
196 				} else {
197 					asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
198 				}
199 
200 				/*
201 				 * CMT fast recovery -- per destination
202 				 * recovery variable.
203 				 */
204 				net->fast_retran_loss_recovery = 1;
205 
206 				if (lchk == NULL) {
207 					/* Mark end of the window */
208 					net->fast_recovery_tsn = asoc->sending_seq - 1;
209 				} else {
210 					net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
211 				}
212 
213 				sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
214 				    stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
215 				sctp_timer_start(SCTP_TIMER_TYPE_SEND,
216 				    stcb->sctp_ep, stcb, net);
217 			}
218 		} else if (net->net_ack > 0) {
219 			/*
220 			 * Mark a peg that we WOULD have done a cwnd
221 			 * reduction but RFC2582 prevented this action.
222 			 */
223 			SCTP_STAT_INCR(sctps_fastretransinrtt);
224 		}
225 	}
226 }
227 
228 /* Defines for instantaneous bw decisions */
229 #define SCTP_INST_LOOSING 1	/* Loosing to other flows */
230 #define SCTP_INST_NEUTRAL 2	/* Neutral, no indication */
231 #define SCTP_INST_GAINING 3	/* Gaining, step down possible */
232 
233 
234 static int
235 cc_bw_same(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw,
236     uint64_t rtt_offset, uint64_t vtag, uint8_t inst_ind)
237 {
238 	uint64_t oth, probepoint;
239 
240 	probepoint = (((uint64_t) net->cwnd) << 32);
241 	if (net->rtt > net->cc_mod.rtcc.lbw_rtt + rtt_offset) {
242 		/*
243 		 * rtt increased we don't update bw.. so we don't update the
244 		 * rtt either.
245 		 */
246 		/* Probe point 5 */
247 		probepoint |= ((5 << 16) | 1);
248 		SDT_PROBE(sctp, cwnd, net, rttvar,
249 		    vtag,
250 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
251 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
252 		    net->flight_size,
253 		    probepoint);
254 		if ((net->cc_mod.rtcc.steady_step) && (inst_ind != SCTP_INST_LOOSING)) {
255 			if (net->cc_mod.rtcc.last_step_state == 5)
256 				net->cc_mod.rtcc.step_cnt++;
257 			else
258 				net->cc_mod.rtcc.step_cnt = 1;
259 			net->cc_mod.rtcc.last_step_state = 5;
260 			if ((net->cc_mod.rtcc.step_cnt == net->cc_mod.rtcc.steady_step) ||
261 			    ((net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step) &&
262 			    ((net->cc_mod.rtcc.step_cnt % net->cc_mod.rtcc.steady_step) == 0))) {
263 				/* Try a step down */
264 				oth = net->cc_mod.rtcc.vol_reduce;
265 				oth <<= 16;
266 				oth |= net->cc_mod.rtcc.step_cnt;
267 				oth <<= 16;
268 				oth |= net->cc_mod.rtcc.last_step_state;
269 				SDT_PROBE(sctp, cwnd, net, rttstep,
270 				    vtag,
271 				    ((net->cc_mod.rtcc.lbw << 32) | nbw),
272 				    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
273 				    oth,
274 				    probepoint);
275 				if (net->cwnd > (4 * net->mtu)) {
276 					net->cwnd -= net->mtu;
277 					net->cc_mod.rtcc.vol_reduce++;
278 				} else {
279 					net->cc_mod.rtcc.step_cnt = 0;
280 				}
281 			}
282 		}
283 		return (1);
284 	}
285 	if (net->rtt < net->cc_mod.rtcc.lbw_rtt - rtt_offset) {
286 		/*
287 		 * rtt decreased, there could be more room. we update both
288 		 * the bw and the rtt here to lock this in as a good step
289 		 * down.
290 		 */
291 		/* Probe point 6 */
292 		probepoint |= ((6 << 16) | 0);
293 		SDT_PROBE(sctp, cwnd, net, rttvar,
294 		    vtag,
295 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
296 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
297 		    net->flight_size,
298 		    probepoint);
299 		if (net->cc_mod.rtcc.steady_step) {
300 			oth = net->cc_mod.rtcc.vol_reduce;
301 			oth <<= 16;
302 			oth |= net->cc_mod.rtcc.step_cnt;
303 			oth <<= 16;
304 			oth |= net->cc_mod.rtcc.last_step_state;
305 			SDT_PROBE(sctp, cwnd, net, rttstep,
306 			    vtag,
307 			    ((net->cc_mod.rtcc.lbw << 32) | nbw),
308 			    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
309 			    oth,
310 			    probepoint);
311 			if ((net->cc_mod.rtcc.last_step_state == 5) &&
312 			    (net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step)) {
313 				/* Step down worked */
314 				net->cc_mod.rtcc.step_cnt = 0;
315 				return (1);
316 			} else {
317 				net->cc_mod.rtcc.last_step_state = 6;
318 				net->cc_mod.rtcc.step_cnt = 0;
319 			}
320 		}
321 		net->cc_mod.rtcc.lbw = nbw;
322 		net->cc_mod.rtcc.lbw_rtt = net->rtt;
323 		net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd;
324 		if (inst_ind == SCTP_INST_GAINING)
325 			return (1);
326 		else if (inst_ind == SCTP_INST_NEUTRAL)
327 			return (1);
328 		else
329 			return (0);
330 	}
331 	/*
332 	 * Ok bw and rtt remained the same .. no update to any
333 	 */
334 	/* Probe point 7 */
335 	probepoint |= ((7 << 16) | net->cc_mod.rtcc.ret_from_eq);
336 	SDT_PROBE(sctp, cwnd, net, rttvar,
337 	    vtag,
338 	    ((net->cc_mod.rtcc.lbw << 32) | nbw),
339 	    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
340 	    net->flight_size,
341 	    probepoint);
342 	if ((net->cc_mod.rtcc.steady_step) && (inst_ind != SCTP_INST_LOOSING)) {
343 		if (net->cc_mod.rtcc.last_step_state == 5)
344 			net->cc_mod.rtcc.step_cnt++;
345 		else
346 			net->cc_mod.rtcc.step_cnt = 1;
347 		net->cc_mod.rtcc.last_step_state = 5;
348 		if ((net->cc_mod.rtcc.step_cnt == net->cc_mod.rtcc.steady_step) ||
349 		    ((net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step) &&
350 		    ((net->cc_mod.rtcc.step_cnt % net->cc_mod.rtcc.steady_step) == 0))) {
351 			/* Try a step down */
352 			if (net->cwnd > (4 * net->mtu)) {
353 				net->cwnd -= net->mtu;
354 				net->cc_mod.rtcc.vol_reduce++;
355 				return (1);
356 			} else {
357 				net->cc_mod.rtcc.step_cnt = 0;
358 			}
359 		}
360 	}
361 	if (inst_ind == SCTP_INST_GAINING)
362 		return (1);
363 	else if (inst_ind == SCTP_INST_NEUTRAL)
364 		return (1);
365 	else
366 		return ((int)net->cc_mod.rtcc.ret_from_eq);
367 }
368 
369 static int
370 cc_bw_decrease(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw, uint64_t rtt_offset,
371     uint64_t vtag, uint8_t inst_ind)
372 {
373 	uint64_t oth, probepoint;
374 
375 	/* Bandwidth decreased. */
376 	probepoint = (((uint64_t) net->cwnd) << 32);
377 	if (net->rtt > net->cc_mod.rtcc.lbw_rtt + rtt_offset) {
378 		/* rtt increased */
379 		/* Did we add more */
380 		if ((net->cwnd > net->cc_mod.rtcc.cwnd_at_bw_set) &&
381 		    (inst_ind != SCTP_INST_LOOSING)) {
382 			/* We caused it maybe.. back off? */
383 			/* PROBE POINT 1 */
384 			probepoint |= ((1 << 16) | 1);
385 			SDT_PROBE(sctp, cwnd, net, rttvar,
386 			    vtag,
387 			    ((net->cc_mod.rtcc.lbw << 32) | nbw),
388 			    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
389 			    net->flight_size,
390 			    probepoint);
391 			if (net->cc_mod.rtcc.ret_from_eq) {
392 				/*
393 				 * Switch over to CA if we are less
394 				 * aggressive
395 				 */
396 				net->ssthresh = net->cwnd - 1;
397 				net->partial_bytes_acked = 0;
398 			}
399 			return (1);
400 		}
401 		/* Probe point 2 */
402 		probepoint |= ((2 << 16) | 0);
403 		SDT_PROBE(sctp, cwnd, net, rttvar,
404 		    vtag,
405 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
406 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
407 		    net->flight_size,
408 		    probepoint);
409 		/* Someone else - fight for more? */
410 		if (net->cc_mod.rtcc.steady_step) {
411 			oth = net->cc_mod.rtcc.vol_reduce;
412 			oth <<= 16;
413 			oth |= net->cc_mod.rtcc.step_cnt;
414 			oth <<= 16;
415 			oth |= net->cc_mod.rtcc.last_step_state;
416 			SDT_PROBE(sctp, cwnd, net, rttstep,
417 			    vtag,
418 			    ((net->cc_mod.rtcc.lbw << 32) | nbw),
419 			    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
420 			    oth,
421 			    probepoint);
422 			/*
423 			 * Did we voluntarily give up some? if so take one
424 			 * back please
425 			 */
426 			if ((net->cc_mod.rtcc.vol_reduce) &&
427 			    (inst_ind != SCTP_INST_GAINING)) {
428 				net->cwnd += net->mtu;
429 				net->cc_mod.rtcc.vol_reduce--;
430 			}
431 			net->cc_mod.rtcc.last_step_state = 2;
432 			net->cc_mod.rtcc.step_cnt = 0;
433 		}
434 		goto out_decision;
435 	} else if (net->rtt < net->cc_mod.rtcc.lbw_rtt - rtt_offset) {
436 		/* bw & rtt decreased */
437 		/* Probe point 3 */
438 		probepoint |= ((3 << 16) | 0);
439 		SDT_PROBE(sctp, cwnd, net, rttvar,
440 		    vtag,
441 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
442 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
443 		    net->flight_size,
444 		    probepoint);
445 		if (net->cc_mod.rtcc.steady_step) {
446 			oth = net->cc_mod.rtcc.vol_reduce;
447 			oth <<= 16;
448 			oth |= net->cc_mod.rtcc.step_cnt;
449 			oth <<= 16;
450 			oth |= net->cc_mod.rtcc.last_step_state;
451 			SDT_PROBE(sctp, cwnd, net, rttstep,
452 			    vtag,
453 			    ((net->cc_mod.rtcc.lbw << 32) | nbw),
454 			    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
455 			    oth,
456 			    probepoint);
457 			if ((net->cc_mod.rtcc.vol_reduce) &&
458 			    (inst_ind != SCTP_INST_GAINING)) {
459 				net->cwnd += net->mtu;
460 				net->cc_mod.rtcc.vol_reduce--;
461 			}
462 			net->cc_mod.rtcc.last_step_state = 3;
463 			net->cc_mod.rtcc.step_cnt = 0;
464 		}
465 		goto out_decision;
466 	}
467 	/* The bw decreased but rtt stayed the same */
468 	/* Probe point 4 */
469 	probepoint |= ((4 << 16) | 0);
470 	SDT_PROBE(sctp, cwnd, net, rttvar,
471 	    vtag,
472 	    ((net->cc_mod.rtcc.lbw << 32) | nbw),
473 	    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
474 	    net->flight_size,
475 	    probepoint);
476 	if (net->cc_mod.rtcc.steady_step) {
477 		oth = net->cc_mod.rtcc.vol_reduce;
478 		oth <<= 16;
479 		oth |= net->cc_mod.rtcc.step_cnt;
480 		oth <<= 16;
481 		oth |= net->cc_mod.rtcc.last_step_state;
482 		SDT_PROBE(sctp, cwnd, net, rttstep,
483 		    vtag,
484 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
485 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
486 		    oth,
487 		    probepoint);
488 		if ((net->cc_mod.rtcc.vol_reduce) &&
489 		    (inst_ind != SCTP_INST_GAINING)) {
490 			net->cwnd += net->mtu;
491 			net->cc_mod.rtcc.vol_reduce--;
492 		}
493 		net->cc_mod.rtcc.last_step_state = 4;
494 		net->cc_mod.rtcc.step_cnt = 0;
495 	}
496 out_decision:
497 	net->cc_mod.rtcc.lbw = nbw;
498 	net->cc_mod.rtcc.lbw_rtt = net->rtt;
499 	net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd;
500 	if (inst_ind == SCTP_INST_GAINING) {
501 		return (1);
502 	} else {
503 		return (0);
504 	}
505 }
506 
507 static int
508 cc_bw_increase(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw, uint64_t vtag)
509 {
510 	uint64_t oth, probepoint;
511 
512 	/*
513 	 * BW increased, so update and return 0, since all actions in our
514 	 * table say to do the normal CC update. Note that we pay no
515 	 * attention to the inst_ind since our overall sum is increasing.
516 	 */
517 	/* PROBE POINT 0 */
518 	probepoint = (((uint64_t) net->cwnd) << 32);
519 	SDT_PROBE(sctp, cwnd, net, rttvar,
520 	    vtag,
521 	    ((net->cc_mod.rtcc.lbw << 32) | nbw),
522 	    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
523 	    net->flight_size,
524 	    probepoint);
525 	if (net->cc_mod.rtcc.steady_step) {
526 		oth = net->cc_mod.rtcc.vol_reduce;
527 		oth <<= 16;
528 		oth |= net->cc_mod.rtcc.step_cnt;
529 		oth <<= 16;
530 		oth |= net->cc_mod.rtcc.last_step_state;
531 		SDT_PROBE(sctp, cwnd, net, rttstep,
532 		    vtag,
533 		    ((net->cc_mod.rtcc.lbw << 32) | nbw),
534 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
535 		    oth,
536 		    probepoint);
537 		net->cc_mod.rtcc.last_step_state = 0;
538 		net->cc_mod.rtcc.step_cnt = 0;
539 		net->cc_mod.rtcc.vol_reduce = 0;
540 	}
541 	net->cc_mod.rtcc.lbw = nbw;
542 	net->cc_mod.rtcc.lbw_rtt = net->rtt;
543 	net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd;
544 	return (0);
545 }
546 
547 /* RTCC Algoritm to limit growth of cwnd, return
548  * true if you want to NOT allow cwnd growth
549  */
550 static int
551 cc_bw_limit(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw)
552 {
553 	uint64_t bw_offset, rtt_offset;
554 	uint64_t probepoint, rtt, vtag;
555 	uint64_t bytes_for_this_rtt, inst_bw;
556 	uint64_t div, inst_off;
557 	int bw_shift;
558 	uint8_t inst_ind;
559 	int ret;
560 
561 	/*-
562 	 * Here we need to see if we want
563 	 * to limit cwnd growth due to increase
564 	 * in overall rtt but no increase in bw.
565 	 * We use the following table to figure
566 	 * out what we should do. When we return
567 	 * 0, cc update goes on as planned. If we
568 	 * return 1, then no cc update happens and cwnd
569 	 * stays where it is at.
570 	 * ----------------------------------
571 	 *   BW    |    RTT   | Action
572 	 * *********************************
573 	 *   INC   |    INC   | return 0
574 	 * ----------------------------------
575 	 *   INC   |    SAME  | return 0
576 	 * ----------------------------------
577 	 *   INC   |    DECR  | return 0
578 	 * ----------------------------------
579 	 *   SAME  |    INC   | return 1
580 	 * ----------------------------------
581 	 *   SAME  |    SAME  | return 1
582 	 * ----------------------------------
583 	 *   SAME  |    DECR  | return 0
584 	 * ----------------------------------
585 	 *   DECR  |    INC   | return 0 or 1 based on if we caused.
586 	 * ----------------------------------
587 	 *   DECR  |    SAME  | return 0
588 	 * ----------------------------------
589 	 *   DECR  |    DECR  | return 0
590 	 * ----------------------------------
591 	 *
592 	 * We are a bit fuzz on what an increase or
593 	 * decrease is. For BW it is the same if
594 	 * it did not change within 1/64th. For
595 	 * RTT it stayed the same if it did not
596 	 * change within 1/32nd
597 	 */
598 	bw_shift = SCTP_BASE_SYSCTL(sctp_rttvar_bw);
599 	rtt = stcb->asoc.my_vtag;
600 	vtag = (rtt << 32) | (((uint32_t) (stcb->sctp_ep->sctp_lport)) << 16) | (stcb->rport);
601 	probepoint = (((uint64_t) net->cwnd) << 32);
602 	rtt = net->rtt;
603 	if (net->cc_mod.rtcc.rtt_set_this_sack) {
604 		net->cc_mod.rtcc.rtt_set_this_sack = 0;
605 		bytes_for_this_rtt = net->cc_mod.rtcc.bw_bytes - net->cc_mod.rtcc.bw_bytes_at_last_rttc;
606 		net->cc_mod.rtcc.bw_bytes_at_last_rttc = net->cc_mod.rtcc.bw_bytes;
607 		if (net->rtt) {
608 			div = net->rtt / 1000;
609 			if (div) {
610 				inst_bw = bytes_for_this_rtt / div;
611 				inst_off = inst_bw >> bw_shift;
612 				if (inst_bw > nbw)
613 					inst_ind = SCTP_INST_GAINING;
614 				else if ((inst_bw + inst_off) < nbw)
615 					inst_ind = SCTP_INST_LOOSING;
616 				else
617 					inst_ind = SCTP_INST_NEUTRAL;
618 				probepoint |= ((0xb << 16) | inst_ind);
619 			} else {
620 				inst_ind = net->cc_mod.rtcc.last_inst_ind;
621 				inst_bw = bytes_for_this_rtt / (uint64_t) (net->rtt);
622 				/* Can't determine do not change */
623 				probepoint |= ((0xc << 16) | inst_ind);
624 			}
625 		} else {
626 			inst_ind = net->cc_mod.rtcc.last_inst_ind;
627 			inst_bw = bytes_for_this_rtt;
628 			/* Can't determine do not change */
629 			probepoint |= ((0xd << 16) | inst_ind);
630 		}
631 		SDT_PROBE(sctp, cwnd, net, rttvar,
632 		    vtag,
633 		    ((nbw << 32) | inst_bw),
634 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | rtt),
635 		    net->flight_size,
636 		    probepoint);
637 	} else {
638 		/* No rtt measurement, use last one */
639 		inst_ind = net->cc_mod.rtcc.last_inst_ind;
640 	}
641 	bw_offset = net->cc_mod.rtcc.lbw >> bw_shift;
642 	if (nbw > net->cc_mod.rtcc.lbw + bw_offset) {
643 		ret = cc_bw_increase(stcb, net, nbw, vtag);
644 		goto out;
645 	}
646 	rtt_offset = net->cc_mod.rtcc.lbw_rtt >> SCTP_BASE_SYSCTL(sctp_rttvar_rtt);
647 	if (nbw < net->cc_mod.rtcc.lbw - bw_offset) {
648 		ret = cc_bw_decrease(stcb, net, nbw, rtt_offset, vtag, inst_ind);
649 		goto out;
650 	}
651 	/*
652 	 * If we reach here then we are in a situation where the bw stayed
653 	 * the same.
654 	 */
655 	ret = cc_bw_same(stcb, net, nbw, rtt_offset, vtag, inst_ind);
656 out:
657 	net->cc_mod.rtcc.last_inst_ind = inst_ind;
658 	return (ret);
659 }
660 
661 static void
662 sctp_cwnd_update_after_sack_common(struct sctp_tcb *stcb,
663     struct sctp_association *asoc,
664     int accum_moved, int reneged_all SCTP_UNUSED, int will_exit, int use_rtcc)
665 {
666 	struct sctp_nets *net;
667 	int old_cwnd;
668 	uint32_t t_ssthresh, t_cwnd, incr;
669 	uint64_t t_ucwnd_sbw;
670 	uint64_t t_path_mptcp;
671 	uint64_t mptcp_like_alpha;
672 	uint32_t srtt;
673 	uint64_t max_path;
674 
675 	/* MT FIXME: Don't compute this over and over again */
676 	t_ssthresh = 0;
677 	t_cwnd = 0;
678 	t_ucwnd_sbw = 0;
679 	t_path_mptcp = 0;
680 	mptcp_like_alpha = 1;
681 	if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) ||
682 	    (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2) ||
683 	    (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_MPTCP)) {
684 		max_path = 0;
685 		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
686 			t_ssthresh += net->ssthresh;
687 			t_cwnd += net->cwnd;
688 			/* lastsa>>3;  we don't need to devide ... */
689 			srtt = net->lastsa;
690 			if (srtt > 0) {
691 				uint64_t tmp;
692 
693 				t_ucwnd_sbw += (uint64_t) net->cwnd / (uint64_t) srtt;
694 				t_path_mptcp += (((uint64_t) net->cwnd) << SHIFT_MPTCP_MULTI_Z) /
695 				    (((uint64_t) net->mtu) * (uint64_t) srtt);
696 				tmp = (((uint64_t) net->cwnd) << SHIFT_MPTCP_MULTI_N) /
697 				    ((uint64_t) net->mtu * (uint64_t) (srtt * srtt));
698 				if (tmp > max_path) {
699 					max_path = tmp;
700 				}
701 			}
702 		}
703 		if (t_path_mptcp > 0) {
704 			mptcp_like_alpha = max_path / (t_path_mptcp * t_path_mptcp);
705 		} else {
706 			mptcp_like_alpha = 1;
707 		}
708 	}
709 	if (t_ssthresh == 0) {
710 		t_ssthresh = 1;
711 	}
712 	if (t_ucwnd_sbw == 0) {
713 		t_ucwnd_sbw = 1;
714 	}
715 	/******************************/
716 	/* update cwnd and Early FR   */
717 	/******************************/
718 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
719 
720 #ifdef JANA_CMT_FAST_RECOVERY
721 		/*
722 		 * CMT fast recovery code. Need to debug.
723 		 */
724 		if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
725 			if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) ||
726 			    SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) {
727 				net->will_exit_fast_recovery = 1;
728 			}
729 		}
730 #endif
731 		/* if nothing was acked on this destination skip it */
732 		if (net->net_ack == 0) {
733 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
734 				sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
735 			}
736 			continue;
737 		}
738 #ifdef JANA_CMT_FAST_RECOVERY
739 		/*
740 		 * CMT fast recovery code
741 		 */
742 		/*
743 		 * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery
744 		 * && net->will_exit_fast_recovery == 0) { @@@ Do something
745 		 * } else if (sctp_cmt_on_off == 0 &&
746 		 * asoc->fast_retran_loss_recovery && will_exit == 0) {
747 		 */
748 #endif
749 
750 		if (asoc->fast_retran_loss_recovery &&
751 		    (will_exit == 0) &&
752 		    (asoc->sctp_cmt_on_off == 0)) {
753 			/*
754 			 * If we are in loss recovery we skip any cwnd
755 			 * update
756 			 */
757 			return;
758 		}
759 		/*
760 		 * Did any measurements go on for this network?
761 		 */
762 		if (use_rtcc && (net->cc_mod.rtcc.tls_needs_set > 0)) {
763 			uint64_t nbw;
764 
765 			/*
766 			 * At this point our bw_bytes has been updated by
767 			 * incoming sack information.
768 			 *
769 			 * But our bw may not yet be set.
770 			 *
771 			 */
772 			if ((net->cc_mod.rtcc.new_tot_time / 1000) > 0) {
773 				nbw = net->cc_mod.rtcc.bw_bytes / (net->cc_mod.rtcc.new_tot_time / 1000);
774 			} else {
775 				nbw = net->cc_mod.rtcc.bw_bytes;
776 			}
777 			if (net->cc_mod.rtcc.lbw) {
778 				if (cc_bw_limit(stcb, net, nbw)) {
779 					/* Hold here, no update */
780 					continue;
781 				}
782 			} else {
783 				uint64_t vtag, probepoint;
784 
785 				probepoint = (((uint64_t) net->cwnd) << 32);
786 				probepoint |= ((0xa << 16) | 0);
787 				vtag = (net->rtt << 32) |
788 				    (((uint32_t) (stcb->sctp_ep->sctp_lport)) << 16) |
789 				    (stcb->rport);
790 
791 				SDT_PROBE(sctp, cwnd, net, rttvar,
792 				    vtag,
793 				    nbw,
794 				    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
795 				    net->flight_size,
796 				    probepoint);
797 				net->cc_mod.rtcc.lbw = nbw;
798 				net->cc_mod.rtcc.lbw_rtt = net->rtt;
799 				if (net->cc_mod.rtcc.rtt_set_this_sack) {
800 					net->cc_mod.rtcc.rtt_set_this_sack = 0;
801 					net->cc_mod.rtcc.bw_bytes_at_last_rttc = net->cc_mod.rtcc.bw_bytes;
802 				}
803 			}
804 		}
805 		/*
806 		 * CMT: CUC algorithm. Update cwnd if pseudo-cumack has
807 		 * moved.
808 		 */
809 		if (accum_moved ||
810 		    ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) {
811 			/* If the cumulative ack moved we can proceed */
812 			if (net->cwnd <= net->ssthresh) {
813 				/* We are in slow start */
814 				if (net->flight_size + net->net_ack >= net->cwnd) {
815 					uint32_t limit;
816 
817 					old_cwnd = net->cwnd;
818 					switch (asoc->sctp_cmt_on_off) {
819 					case SCTP_CMT_RPV1:
820 						limit = (uint32_t) (((uint64_t) net->mtu *
821 						    (uint64_t) SCTP_BASE_SYSCTL(sctp_L2_abc_variable) *
822 						    (uint64_t) net->ssthresh) /
823 						    (uint64_t) t_ssthresh);
824 						incr = (uint32_t) (((uint64_t) net->net_ack *
825 						    (uint64_t) net->ssthresh) /
826 						    (uint64_t) t_ssthresh);
827 						if (incr > limit) {
828 							incr = limit;
829 						}
830 						if (incr == 0) {
831 							incr = 1;
832 						}
833 						break;
834 					case SCTP_CMT_RPV2:
835 						/*
836 						 * lastsa>>3;  we don't need
837 						 * to divide ...
838 						 */
839 						srtt = net->lastsa;
840 						if (srtt == 0) {
841 							srtt = 1;
842 						}
843 						limit = (uint32_t) (((uint64_t) net->mtu *
844 						    (uint64_t) SCTP_BASE_SYSCTL(sctp_L2_abc_variable) *
845 						    (uint64_t) net->cwnd) /
846 						    ((uint64_t) srtt * t_ucwnd_sbw));
847 						/* INCREASE FACTOR */
848 						incr = (uint32_t) (((uint64_t) net->net_ack *
849 						    (uint64_t) net->cwnd) /
850 						    ((uint64_t) srtt * t_ucwnd_sbw));
851 						/* INCREASE FACTOR */
852 						if (incr > limit) {
853 							incr = limit;
854 						}
855 						if (incr == 0) {
856 							incr = 1;
857 						}
858 						break;
859 					case SCTP_CMT_MPTCP:
860 						limit = (uint32_t) (((uint64_t) net->mtu *
861 						    mptcp_like_alpha *
862 						    (uint64_t) SCTP_BASE_SYSCTL(sctp_L2_abc_variable)) >>
863 						    SHIFT_MPTCP_MULTI);
864 						incr = (uint32_t) (((uint64_t) net->net_ack *
865 						    mptcp_like_alpha) >>
866 						    SHIFT_MPTCP_MULTI);
867 						if (incr > limit) {
868 							incr = limit;
869 						}
870 						if (incr > net->net_ack) {
871 							incr = net->net_ack;
872 						}
873 						if (incr > net->mtu) {
874 							incr = net->mtu;
875 						}
876 						break;
877 					default:
878 						incr = net->net_ack;
879 						if (incr > net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable)) {
880 							incr = net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable);
881 						}
882 						break;
883 					}
884 					net->cwnd += incr;
885 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
886 						sctp_log_cwnd(stcb, net, incr,
887 						    SCTP_CWND_LOG_FROM_SS);
888 					}
889 					SDT_PROBE(sctp, cwnd, net, ack,
890 					    stcb->asoc.my_vtag,
891 					    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
892 					    net,
893 					    old_cwnd, net->cwnd);
894 				} else {
895 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
896 						sctp_log_cwnd(stcb, net, net->net_ack,
897 						    SCTP_CWND_LOG_NOADV_SS);
898 					}
899 				}
900 			} else {
901 				/* We are in congestion avoidance */
902 				/*
903 				 * Add to pba
904 				 */
905 				net->partial_bytes_acked += net->net_ack;
906 
907 				if ((net->flight_size + net->net_ack >= net->cwnd) &&
908 				    (net->partial_bytes_acked >= net->cwnd)) {
909 					net->partial_bytes_acked -= net->cwnd;
910 					old_cwnd = net->cwnd;
911 					switch (asoc->sctp_cmt_on_off) {
912 					case SCTP_CMT_RPV1:
913 						incr = (uint32_t) (((uint64_t) net->mtu *
914 						    (uint64_t) net->ssthresh) /
915 						    (uint64_t) t_ssthresh);
916 						if (incr == 0) {
917 							incr = 1;
918 						}
919 						break;
920 					case SCTP_CMT_RPV2:
921 						/*
922 						 * lastsa>>3;  we don't need
923 						 * to divide ...
924 						 */
925 						srtt = net->lastsa;
926 						if (srtt == 0) {
927 							srtt = 1;
928 						}
929 						incr = (uint32_t) ((uint64_t) net->mtu *
930 						    (uint64_t) net->cwnd /
931 						    ((uint64_t) srtt *
932 						    t_ucwnd_sbw));
933 						/* INCREASE FACTOR */
934 						if (incr == 0) {
935 							incr = 1;
936 						}
937 						break;
938 					case SCTP_CMT_MPTCP:
939 						incr = (uint32_t) ((mptcp_like_alpha *
940 						    (uint64_t) net->cwnd) >>
941 						    SHIFT_MPTCP_MULTI);
942 						if (incr > net->mtu) {
943 							incr = net->mtu;
944 						}
945 						break;
946 					default:
947 						incr = net->mtu;
948 						break;
949 					}
950 					net->cwnd += incr;
951 					SDT_PROBE(sctp, cwnd, net, ack,
952 					    stcb->asoc.my_vtag,
953 					    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
954 					    net,
955 					    old_cwnd, net->cwnd);
956 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
957 						sctp_log_cwnd(stcb, net, net->mtu,
958 						    SCTP_CWND_LOG_FROM_CA);
959 					}
960 				} else {
961 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
962 						sctp_log_cwnd(stcb, net, net->net_ack,
963 						    SCTP_CWND_LOG_NOADV_CA);
964 					}
965 				}
966 			}
967 		} else {
968 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
969 				sctp_log_cwnd(stcb, net, net->mtu,
970 				    SCTP_CWND_LOG_NO_CUMACK);
971 			}
972 		}
973 	}
974 }
975 
976 static void
977 sctp_cwnd_update_exit_pf_common(struct sctp_tcb *stcb, struct sctp_nets *net)
978 {
979 	int old_cwnd;
980 
981 	old_cwnd = net->cwnd;
982 	net->cwnd = net->mtu;
983 	SDT_PROBE(sctp, cwnd, net, ack,
984 	    stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net,
985 	    old_cwnd, net->cwnd);
986 	SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n",
987 	    (void *)net, net->cwnd);
988 }
989 
990 
991 static void
992 sctp_cwnd_update_after_timeout(struct sctp_tcb *stcb, struct sctp_nets *net)
993 {
994 	int old_cwnd = net->cwnd;
995 	uint32_t t_ssthresh, t_cwnd;
996 	uint64_t t_ucwnd_sbw;
997 
998 	/* MT FIXME: Don't compute this over and over again */
999 	t_ssthresh = 0;
1000 	t_cwnd = 0;
1001 	if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) ||
1002 	    (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2)) {
1003 		struct sctp_nets *lnet;
1004 		uint32_t srtt;
1005 
1006 		t_ucwnd_sbw = 0;
1007 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
1008 			t_ssthresh += lnet->ssthresh;
1009 			t_cwnd += lnet->cwnd;
1010 			srtt = lnet->lastsa;
1011 			/* lastsa>>3;  we don't need to divide ... */
1012 			if (srtt > 0) {
1013 				t_ucwnd_sbw += (uint64_t) lnet->cwnd / (uint64_t) srtt;
1014 			}
1015 		}
1016 		if (t_ssthresh < 1) {
1017 			t_ssthresh = 1;
1018 		}
1019 		if (t_ucwnd_sbw < 1) {
1020 			t_ucwnd_sbw = 1;
1021 		}
1022 		if (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) {
1023 			net->ssthresh = (uint32_t) (((uint64_t) 4 *
1024 			    (uint64_t) net->mtu *
1025 			    (uint64_t) net->ssthresh) /
1026 			    (uint64_t) t_ssthresh);
1027 		} else {
1028 			uint64_t cc_delta;
1029 
1030 			srtt = net->lastsa;
1031 			/* lastsa>>3;  we don't need to divide ... */
1032 			if (srtt == 0) {
1033 				srtt = 1;
1034 			}
1035 			cc_delta = t_ucwnd_sbw * (uint64_t) srtt / 2;
1036 			if (cc_delta < t_cwnd) {
1037 				net->ssthresh = (uint32_t) ((uint64_t) t_cwnd - cc_delta);
1038 			} else {
1039 				net->ssthresh = net->mtu;
1040 			}
1041 		}
1042 		if ((net->cwnd > t_cwnd / 2) &&
1043 		    (net->ssthresh < net->cwnd - t_cwnd / 2)) {
1044 			net->ssthresh = net->cwnd - t_cwnd / 2;
1045 		}
1046 		if (net->ssthresh < net->mtu) {
1047 			net->ssthresh = net->mtu;
1048 		}
1049 	} else {
1050 		net->ssthresh = max(net->cwnd / 2, 4 * net->mtu);
1051 	}
1052 	net->cwnd = net->mtu;
1053 	net->partial_bytes_acked = 0;
1054 	SDT_PROBE(sctp, cwnd, net, to,
1055 	    stcb->asoc.my_vtag,
1056 	    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
1057 	    net,
1058 	    old_cwnd, net->cwnd);
1059 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1060 		sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX);
1061 	}
1062 }
1063 
1064 static void
1065 sctp_cwnd_update_after_ecn_echo_common(struct sctp_tcb *stcb, struct sctp_nets *net,
1066     int in_window, int num_pkt_lost, int use_rtcc)
1067 {
1068 	int old_cwnd = net->cwnd;
1069 
1070 	if ((use_rtcc) && (net->lan_type == SCTP_LAN_LOCAL) && (net->cc_mod.rtcc.use_dccc_ecn)) {
1071 		/* Data center Congestion Control */
1072 		if (in_window == 0) {
1073 			/*
1074 			 * Go to CA with the cwnd at the point we sent the
1075 			 * TSN that was marked with a CE.
1076 			 */
1077 			if (net->ecn_prev_cwnd < net->cwnd) {
1078 				/* Restore to prev cwnd */
1079 				net->cwnd = net->ecn_prev_cwnd - (net->mtu * num_pkt_lost);
1080 			} else {
1081 				/* Just cut in 1/2 */
1082 				net->cwnd /= 2;
1083 			}
1084 			/* Drop to CA */
1085 			net->ssthresh = net->cwnd - (num_pkt_lost * net->mtu);
1086 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1087 				sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
1088 			}
1089 		} else {
1090 			/*
1091 			 * Further tuning down required over the drastic
1092 			 * orginal cut
1093 			 */
1094 			net->ssthresh -= (net->mtu * num_pkt_lost);
1095 			net->cwnd -= (net->mtu * num_pkt_lost);
1096 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1097 				sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
1098 			}
1099 		}
1100 		SCTP_STAT_INCR(sctps_ecnereducedcwnd);
1101 	} else {
1102 		if (in_window == 0) {
1103 			SCTP_STAT_INCR(sctps_ecnereducedcwnd);
1104 			net->ssthresh = net->cwnd / 2;
1105 			if (net->ssthresh < net->mtu) {
1106 				net->ssthresh = net->mtu;
1107 				/*
1108 				 * here back off the timer as well, to slow
1109 				 * us down
1110 				 */
1111 				net->RTO <<= 1;
1112 			}
1113 			net->cwnd = net->ssthresh;
1114 			SDT_PROBE(sctp, cwnd, net, ecn,
1115 			    stcb->asoc.my_vtag,
1116 			    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
1117 			    net,
1118 			    old_cwnd, net->cwnd);
1119 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1120 				sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
1121 			}
1122 		}
1123 	}
1124 
1125 }
1126 
1127 static void
1128 sctp_cwnd_update_after_packet_dropped(struct sctp_tcb *stcb,
1129     struct sctp_nets *net, struct sctp_pktdrop_chunk *cp,
1130     uint32_t * bottle_bw, uint32_t * on_queue)
1131 {
1132 	uint32_t bw_avail;
1133 	unsigned int incr;
1134 	int old_cwnd = net->cwnd;
1135 
1136 	/* get bottle neck bw */
1137 	*bottle_bw = ntohl(cp->bottle_bw);
1138 	/* and whats on queue */
1139 	*on_queue = ntohl(cp->current_onq);
1140 	/*
1141 	 * adjust the on-queue if our flight is more it could be that the
1142 	 * router has not yet gotten data "in-flight" to it
1143 	 */
1144 	if (*on_queue < net->flight_size) {
1145 		*on_queue = net->flight_size;
1146 	}
1147 	/* rtt is measured in micro seconds, bottle_bw in bytes per second */
1148 	bw_avail = (uint32_t) (((uint64_t) (*bottle_bw) * net->rtt) / (uint64_t) 1000000);
1149 	if (bw_avail > *bottle_bw) {
1150 		/*
1151 		 * Cap the growth to no more than the bottle neck. This can
1152 		 * happen as RTT slides up due to queues. It also means if
1153 		 * you have more than a 1 second RTT with a empty queue you
1154 		 * will be limited to the bottle_bw per second no matter if
1155 		 * other points have 1/2 the RTT and you could get more
1156 		 * out...
1157 		 */
1158 		bw_avail = *bottle_bw;
1159 	}
1160 	if (*on_queue > bw_avail) {
1161 		/*
1162 		 * No room for anything else don't allow anything else to be
1163 		 * "added to the fire".
1164 		 */
1165 		int seg_inflight, seg_onqueue, my_portion;
1166 
1167 		net->partial_bytes_acked = 0;
1168 		/* how much are we over queue size? */
1169 		incr = *on_queue - bw_avail;
1170 		if (stcb->asoc.seen_a_sack_this_pkt) {
1171 			/*
1172 			 * undo any cwnd adjustment that the sack might have
1173 			 * made
1174 			 */
1175 			net->cwnd = net->prev_cwnd;
1176 		}
1177 		/* Now how much of that is mine? */
1178 		seg_inflight = net->flight_size / net->mtu;
1179 		seg_onqueue = *on_queue / net->mtu;
1180 		my_portion = (incr * seg_inflight) / seg_onqueue;
1181 
1182 		/* Have I made an adjustment already */
1183 		if (net->cwnd > net->flight_size) {
1184 			/*
1185 			 * for this flight I made an adjustment we need to
1186 			 * decrease the portion by a share our previous
1187 			 * adjustment.
1188 			 */
1189 			int diff_adj;
1190 
1191 			diff_adj = net->cwnd - net->flight_size;
1192 			if (diff_adj > my_portion)
1193 				my_portion = 0;
1194 			else
1195 				my_portion -= diff_adj;
1196 		}
1197 		/*
1198 		 * back down to the previous cwnd (assume we have had a sack
1199 		 * before this packet). minus what ever portion of the
1200 		 * overage is my fault.
1201 		 */
1202 		net->cwnd -= my_portion;
1203 
1204 		/* we will NOT back down more than 1 MTU */
1205 		if (net->cwnd <= net->mtu) {
1206 			net->cwnd = net->mtu;
1207 		}
1208 		/* force into CA */
1209 		net->ssthresh = net->cwnd - 1;
1210 	} else {
1211 		/*
1212 		 * Take 1/4 of the space left or max burst up .. whichever
1213 		 * is less.
1214 		 */
1215 		incr = (bw_avail - *on_queue) >> 2;
1216 		if ((stcb->asoc.max_burst > 0) &&
1217 		    (stcb->asoc.max_burst * net->mtu < incr)) {
1218 			incr = stcb->asoc.max_burst * net->mtu;
1219 		}
1220 		net->cwnd += incr;
1221 	}
1222 	if (net->cwnd > bw_avail) {
1223 		/* We can't exceed the pipe size */
1224 		net->cwnd = bw_avail;
1225 	}
1226 	if (net->cwnd < net->mtu) {
1227 		/* We always have 1 MTU */
1228 		net->cwnd = net->mtu;
1229 	}
1230 	if (net->cwnd - old_cwnd != 0) {
1231 		/* log only changes */
1232 		SDT_PROBE(sctp, cwnd, net, pd,
1233 		    stcb->asoc.my_vtag,
1234 		    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
1235 		    net,
1236 		    old_cwnd, net->cwnd);
1237 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1238 			sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
1239 			    SCTP_CWND_LOG_FROM_SAT);
1240 		}
1241 	}
1242 }
1243 
1244 static void
1245 sctp_cwnd_update_after_output(struct sctp_tcb *stcb,
1246     struct sctp_nets *net, int burst_limit)
1247 {
1248 	int old_cwnd = net->cwnd;
1249 
1250 	if (net->ssthresh < net->cwnd)
1251 		net->ssthresh = net->cwnd;
1252 	if (burst_limit) {
1253 		net->cwnd = (net->flight_size + (burst_limit * net->mtu));
1254 		SDT_PROBE(sctp, cwnd, net, bl,
1255 		    stcb->asoc.my_vtag,
1256 		    ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)),
1257 		    net,
1258 		    old_cwnd, net->cwnd);
1259 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1260 			sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_BRST);
1261 		}
1262 	}
1263 }
1264 
1265 static void
1266 sctp_cwnd_update_after_sack(struct sctp_tcb *stcb,
1267     struct sctp_association *asoc,
1268     int accum_moved, int reneged_all, int will_exit)
1269 {
1270 	/* Passing a zero argument in last disables the rtcc algoritm */
1271 	sctp_cwnd_update_after_sack_common(stcb, asoc, accum_moved, reneged_all, will_exit, 0);
1272 }
1273 
1274 static void
1275 sctp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net,
1276     int in_window, int num_pkt_lost)
1277 {
1278 	/* Passing a zero argument in last disables the rtcc algoritm */
1279 	sctp_cwnd_update_after_ecn_echo_common(stcb, net, in_window, num_pkt_lost, 0);
1280 }
1281 
1282 /* Here starts the RTCCVAR type CC invented by RRS which
1283  * is a slight mod to RFC2581. We reuse a common routine or
1284  * two since these algoritms are so close and need to
1285  * remain the same.
1286  */
1287 static void
1288 sctp_cwnd_update_rtcc_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net,
1289     int in_window, int num_pkt_lost)
1290 {
1291 	sctp_cwnd_update_after_ecn_echo_common(stcb, net, in_window, num_pkt_lost, 1);
1292 }
1293 
1294 
1295 static
1296 void
1297 sctp_cwnd_update_rtcc_tsn_acknowledged(struct sctp_nets *net,
1298     struct sctp_tmit_chunk *tp1)
1299 {
1300 	net->cc_mod.rtcc.bw_bytes += tp1->send_size;
1301 }
1302 
1303 static void
1304 sctp_cwnd_prepare_rtcc_net_for_sack(struct sctp_tcb *stcb SCTP_UNUSED,
1305     struct sctp_nets *net)
1306 {
1307 	if (net->cc_mod.rtcc.tls_needs_set > 0) {
1308 		/* We had a bw measurment going on */
1309 		struct timeval ltls;
1310 
1311 		SCTP_GETPTIME_TIMEVAL(&ltls);
1312 		timevalsub(&ltls, &net->cc_mod.rtcc.tls);
1313 		net->cc_mod.rtcc.new_tot_time = (ltls.tv_sec * 1000000) + ltls.tv_usec;
1314 	}
1315 }
1316 
1317 static void
1318 sctp_cwnd_new_rtcc_transmission_begins(struct sctp_tcb *stcb,
1319     struct sctp_nets *net)
1320 {
1321 	uint64_t vtag, probepoint;
1322 
1323 	if (net->cc_mod.rtcc.lbw) {
1324 		/* Clear the old bw.. we went to 0 in-flight */
1325 		vtag = (net->rtt << 32) | (((uint32_t) (stcb->sctp_ep->sctp_lport)) << 16) |
1326 		    (stcb->rport);
1327 		probepoint = (((uint64_t) net->cwnd) << 32);
1328 		/* Probe point 8 */
1329 		probepoint |= ((8 << 16) | 0);
1330 		SDT_PROBE(sctp, cwnd, net, rttvar,
1331 		    vtag,
1332 		    ((net->cc_mod.rtcc.lbw << 32) | 0),
1333 		    ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt),
1334 		    net->flight_size,
1335 		    probepoint);
1336 		net->cc_mod.rtcc.lbw_rtt = 0;
1337 		net->cc_mod.rtcc.cwnd_at_bw_set = 0;
1338 		net->cc_mod.rtcc.lbw = 0;
1339 		net->cc_mod.rtcc.bw_bytes_at_last_rttc = 0;
1340 		net->cc_mod.rtcc.vol_reduce = 0;
1341 		net->cc_mod.rtcc.bw_tot_time = 0;
1342 		net->cc_mod.rtcc.bw_bytes = 0;
1343 		net->cc_mod.rtcc.tls_needs_set = 0;
1344 		if (net->cc_mod.rtcc.steady_step) {
1345 			net->cc_mod.rtcc.vol_reduce = 0;
1346 			net->cc_mod.rtcc.step_cnt = 0;
1347 			net->cc_mod.rtcc.last_step_state = 0;
1348 		}
1349 		if (net->cc_mod.rtcc.ret_from_eq) {
1350 			/* less aggressive one - reset cwnd too */
1351 			uint32_t cwnd_in_mtu, cwnd;
1352 
1353 			cwnd_in_mtu = SCTP_BASE_SYSCTL(sctp_initial_cwnd);
1354 			if (cwnd_in_mtu == 0) {
1355 				/*
1356 				 * Using 0 means that the value of RFC 4960
1357 				 * is used.
1358 				 */
1359 				cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
1360 			} else {
1361 				/*
1362 				 * We take the minimum of the burst limit
1363 				 * and the initial congestion window.
1364 				 */
1365 				if ((stcb->asoc.max_burst > 0) && (cwnd_in_mtu > stcb->asoc.max_burst))
1366 					cwnd_in_mtu = stcb->asoc.max_burst;
1367 				cwnd = (net->mtu - sizeof(struct sctphdr)) * cwnd_in_mtu;
1368 			}
1369 			if (net->cwnd > cwnd) {
1370 				/*
1371 				 * Only set if we are not a timeout (i.e.
1372 				 * down to 1 mtu)
1373 				 */
1374 				net->cwnd = cwnd;
1375 			}
1376 		}
1377 	}
1378 }
1379 
1380 static void
1381 sctp_set_rtcc_initial_cc_param(struct sctp_tcb *stcb,
1382     struct sctp_nets *net)
1383 {
1384 	uint64_t vtag, probepoint;
1385 
1386 	sctp_set_initial_cc_param(stcb, net);
1387 	stcb->asoc.use_precise_time = 1;
1388 	probepoint = (((uint64_t) net->cwnd) << 32);
1389 	probepoint |= ((9 << 16) | 0);
1390 	vtag = (net->rtt << 32) |
1391 	    (((uint32_t) (stcb->sctp_ep->sctp_lport)) << 16) |
1392 	    (stcb->rport);
1393 	SDT_PROBE(sctp, cwnd, net, rttvar,
1394 	    vtag,
1395 	    0,
1396 	    0,
1397 	    0,
1398 	    probepoint);
1399 	net->cc_mod.rtcc.lbw_rtt = 0;
1400 	net->cc_mod.rtcc.cwnd_at_bw_set = 0;
1401 	net->cc_mod.rtcc.vol_reduce = 0;
1402 	net->cc_mod.rtcc.lbw = 0;
1403 	net->cc_mod.rtcc.vol_reduce = 0;
1404 	net->cc_mod.rtcc.bw_bytes_at_last_rttc = 0;
1405 	net->cc_mod.rtcc.bw_tot_time = 0;
1406 	net->cc_mod.rtcc.bw_bytes = 0;
1407 	net->cc_mod.rtcc.tls_needs_set = 0;
1408 	net->cc_mod.rtcc.ret_from_eq = SCTP_BASE_SYSCTL(sctp_rttvar_eqret);
1409 	net->cc_mod.rtcc.steady_step = SCTP_BASE_SYSCTL(sctp_steady_step);
1410 	net->cc_mod.rtcc.use_dccc_ecn = SCTP_BASE_SYSCTL(sctp_use_dccc_ecn);
1411 	net->cc_mod.rtcc.step_cnt = 0;
1412 	net->cc_mod.rtcc.last_step_state = 0;
1413 
1414 
1415 }
1416 
1417 static int
1418 sctp_cwnd_rtcc_socket_option(struct sctp_tcb *stcb, int setorget,
1419     struct sctp_cc_option *cc_opt)
1420 {
1421 	struct sctp_nets *net;
1422 
1423 	if (setorget == 1) {
1424 		/* a set */
1425 		if (cc_opt->option == SCTP_CC_OPT_RTCC_SETMODE) {
1426 			if ((cc_opt->aid_value.assoc_value != 0) &&
1427 			    (cc_opt->aid_value.assoc_value != 1)) {
1428 				return (EINVAL);
1429 			}
1430 			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1431 				net->cc_mod.rtcc.ret_from_eq = cc_opt->aid_value.assoc_value;
1432 			}
1433 		} else if (cc_opt->option == SCTP_CC_OPT_USE_DCCC_ECN) {
1434 			if ((cc_opt->aid_value.assoc_value != 0) &&
1435 			    (cc_opt->aid_value.assoc_value != 1)) {
1436 				return (EINVAL);
1437 			}
1438 			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1439 				net->cc_mod.rtcc.use_dccc_ecn = cc_opt->aid_value.assoc_value;
1440 			}
1441 		} else if (cc_opt->option == SCTP_CC_OPT_STEADY_STEP) {
1442 			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1443 				net->cc_mod.rtcc.steady_step = cc_opt->aid_value.assoc_value;
1444 			}
1445 		} else {
1446 			return (EINVAL);
1447 		}
1448 	} else {
1449 		/* a get */
1450 		if (cc_opt->option == SCTP_CC_OPT_RTCC_SETMODE) {
1451 			net = TAILQ_FIRST(&stcb->asoc.nets);
1452 			if (net == NULL) {
1453 				return (EFAULT);
1454 			}
1455 			cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.ret_from_eq;
1456 		} else if (cc_opt->option == SCTP_CC_OPT_USE_DCCC_ECN) {
1457 			net = TAILQ_FIRST(&stcb->asoc.nets);
1458 			if (net == NULL) {
1459 				return (EFAULT);
1460 			}
1461 			cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.use_dccc_ecn;
1462 		} else if (cc_opt->option == SCTP_CC_OPT_STEADY_STEP) {
1463 			net = TAILQ_FIRST(&stcb->asoc.nets);
1464 			if (net == NULL) {
1465 				return (EFAULT);
1466 			}
1467 			cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.steady_step;
1468 		} else {
1469 			return (EINVAL);
1470 		}
1471 	}
1472 	return (0);
1473 }
1474 
1475 static void
1476 sctp_cwnd_update_rtcc_packet_transmitted(struct sctp_tcb *stcb SCTP_UNUSED,
1477     struct sctp_nets *net)
1478 {
1479 	if (net->cc_mod.rtcc.tls_needs_set == 0) {
1480 		SCTP_GETPTIME_TIMEVAL(&net->cc_mod.rtcc.tls);
1481 		net->cc_mod.rtcc.tls_needs_set = 2;
1482 	}
1483 }
1484 
1485 static void
1486 sctp_cwnd_update_rtcc_after_sack(struct sctp_tcb *stcb,
1487     struct sctp_association *asoc,
1488     int accum_moved, int reneged_all, int will_exit)
1489 {
1490 	/* Passing a one argument at the last enables the rtcc algoritm */
1491 	sctp_cwnd_update_after_sack_common(stcb, asoc, accum_moved, reneged_all, will_exit, 1);
1492 }
1493 
1494 static void
1495 sctp_rtt_rtcc_calculated(struct sctp_tcb *stcb SCTP_UNUSED,
1496     struct sctp_nets *net,
1497     struct timeval *now SCTP_UNUSED)
1498 {
1499 	net->cc_mod.rtcc.rtt_set_this_sack = 1;
1500 }
1501 
1502 /* Here starts Sally Floyds HS-TCP */
1503 
1504 struct sctp_hs_raise_drop {
1505 	int32_t cwnd;
1506 	int32_t increase;
1507 	int32_t drop_percent;
1508 };
1509 
1510 #define SCTP_HS_TABLE_SIZE 73
1511 
1512 struct sctp_hs_raise_drop sctp_cwnd_adjust[SCTP_HS_TABLE_SIZE] = {
1513 	{38, 1, 50},		/* 0   */
1514 	{118, 2, 44},		/* 1   */
1515 	{221, 3, 41},		/* 2   */
1516 	{347, 4, 38},		/* 3   */
1517 	{495, 5, 37},		/* 4   */
1518 	{663, 6, 35},		/* 5   */
1519 	{851, 7, 34},		/* 6   */
1520 	{1058, 8, 33},		/* 7   */
1521 	{1284, 9, 32},		/* 8   */
1522 	{1529, 10, 31},		/* 9   */
1523 	{1793, 11, 30},		/* 10  */
1524 	{2076, 12, 29},		/* 11  */
1525 	{2378, 13, 28},		/* 12  */
1526 	{2699, 14, 28},		/* 13  */
1527 	{3039, 15, 27},		/* 14  */
1528 	{3399, 16, 27},		/* 15  */
1529 	{3778, 17, 26},		/* 16  */
1530 	{4177, 18, 26},		/* 17  */
1531 	{4596, 19, 25},		/* 18  */
1532 	{5036, 20, 25},		/* 19  */
1533 	{5497, 21, 24},		/* 20  */
1534 	{5979, 22, 24},		/* 21  */
1535 	{6483, 23, 23},		/* 22  */
1536 	{7009, 24, 23},		/* 23  */
1537 	{7558, 25, 22},		/* 24  */
1538 	{8130, 26, 22},		/* 25  */
1539 	{8726, 27, 22},		/* 26  */
1540 	{9346, 28, 21},		/* 27  */
1541 	{9991, 29, 21},		/* 28  */
1542 	{10661, 30, 21},	/* 29  */
1543 	{11358, 31, 20},	/* 30  */
1544 	{12082, 32, 20},	/* 31  */
1545 	{12834, 33, 20},	/* 32  */
1546 	{13614, 34, 19},	/* 33  */
1547 	{14424, 35, 19},	/* 34  */
1548 	{15265, 36, 19},	/* 35  */
1549 	{16137, 37, 19},	/* 36  */
1550 	{17042, 38, 18},	/* 37  */
1551 	{17981, 39, 18},	/* 38  */
1552 	{18955, 40, 18},	/* 39  */
1553 	{19965, 41, 17},	/* 40  */
1554 	{21013, 42, 17},	/* 41  */
1555 	{22101, 43, 17},	/* 42  */
1556 	{23230, 44, 17},	/* 43  */
1557 	{24402, 45, 16},	/* 44  */
1558 	{25618, 46, 16},	/* 45  */
1559 	{26881, 47, 16},	/* 46  */
1560 	{28193, 48, 16},	/* 47  */
1561 	{29557, 49, 15},	/* 48  */
1562 	{30975, 50, 15},	/* 49  */
1563 	{32450, 51, 15},	/* 50  */
1564 	{33986, 52, 15},	/* 51  */
1565 	{35586, 53, 14},	/* 52  */
1566 	{37253, 54, 14},	/* 53  */
1567 	{38992, 55, 14},	/* 54  */
1568 	{40808, 56, 14},	/* 55  */
1569 	{42707, 57, 13},	/* 56  */
1570 	{44694, 58, 13},	/* 57  */
1571 	{46776, 59, 13},	/* 58  */
1572 	{48961, 60, 13},	/* 59  */
1573 	{51258, 61, 13},	/* 60  */
1574 	{53677, 62, 12},	/* 61  */
1575 	{56230, 63, 12},	/* 62  */
1576 	{58932, 64, 12},	/* 63  */
1577 	{61799, 65, 12},	/* 64  */
1578 	{64851, 66, 11},	/* 65  */
1579 	{68113, 67, 11},	/* 66  */
1580 	{71617, 68, 11},	/* 67  */
1581 	{75401, 69, 10},	/* 68  */
1582 	{79517, 70, 10},	/* 69  */
1583 	{84035, 71, 10},	/* 70  */
1584 	{89053, 72, 10},	/* 71  */
1585 	{94717, 73, 9}		/* 72  */
1586 };
1587 
1588 static void
1589 sctp_hs_cwnd_increase(struct sctp_tcb *stcb, struct sctp_nets *net)
1590 {
1591 	int cur_val, i, indx, incr;
1592 
1593 	cur_val = net->cwnd >> 10;
1594 	indx = SCTP_HS_TABLE_SIZE - 1;
1595 
1596 	if (cur_val < sctp_cwnd_adjust[0].cwnd) {
1597 		/* normal mode */
1598 		if (net->net_ack > net->mtu) {
1599 			net->cwnd += net->mtu;
1600 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1601 				sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_SS);
1602 			}
1603 		} else {
1604 			net->cwnd += net->net_ack;
1605 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1606 				sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_FROM_SS);
1607 			}
1608 		}
1609 	} else {
1610 		for (i = net->last_hs_used; i < SCTP_HS_TABLE_SIZE; i++) {
1611 			if (cur_val < sctp_cwnd_adjust[i].cwnd) {
1612 				indx = i;
1613 				break;
1614 			}
1615 		}
1616 		net->last_hs_used = indx;
1617 		incr = ((sctp_cwnd_adjust[indx].increase) << 10);
1618 		net->cwnd += incr;
1619 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1620 			sctp_log_cwnd(stcb, net, incr, SCTP_CWND_LOG_FROM_SS);
1621 		}
1622 	}
1623 }
1624 
1625 static void
1626 sctp_hs_cwnd_decrease(struct sctp_tcb *stcb, struct sctp_nets *net)
1627 {
1628 	int cur_val, i, indx;
1629 	int old_cwnd = net->cwnd;
1630 
1631 	cur_val = net->cwnd >> 10;
1632 	if (cur_val < sctp_cwnd_adjust[0].cwnd) {
1633 		/* normal mode */
1634 		net->ssthresh = net->cwnd / 2;
1635 		if (net->ssthresh < (net->mtu * 2)) {
1636 			net->ssthresh = 2 * net->mtu;
1637 		}
1638 		net->cwnd = net->ssthresh;
1639 	} else {
1640 		/* drop by the proper amount */
1641 		net->ssthresh = net->cwnd - (int)((net->cwnd / 100) *
1642 		    sctp_cwnd_adjust[net->last_hs_used].drop_percent);
1643 		net->cwnd = net->ssthresh;
1644 		/* now where are we */
1645 		indx = net->last_hs_used;
1646 		cur_val = net->cwnd >> 10;
1647 		/* reset where we are in the table */
1648 		if (cur_val < sctp_cwnd_adjust[0].cwnd) {
1649 			/* feel out of hs */
1650 			net->last_hs_used = 0;
1651 		} else {
1652 			for (i = indx; i >= 1; i--) {
1653 				if (cur_val > sctp_cwnd_adjust[i - 1].cwnd) {
1654 					break;
1655 				}
1656 			}
1657 			net->last_hs_used = indx;
1658 		}
1659 	}
1660 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1661 		sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR);
1662 	}
1663 }
1664 
1665 static void
1666 sctp_hs_cwnd_update_after_fr(struct sctp_tcb *stcb,
1667     struct sctp_association *asoc)
1668 {
1669 	struct sctp_nets *net;
1670 
1671 	/*
1672 	 * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) &&
1673 	 * (net->fast_retran_loss_recovery == 0)))
1674 	 */
1675 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
1676 		if ((asoc->fast_retran_loss_recovery == 0) ||
1677 		    (asoc->sctp_cmt_on_off > 0)) {
1678 			/* out of a RFC2582 Fast recovery window? */
1679 			if (net->net_ack > 0) {
1680 				/*
1681 				 * per section 7.2.3, are there any
1682 				 * destinations that had a fast retransmit
1683 				 * to them. If so what we need to do is
1684 				 * adjust ssthresh and cwnd.
1685 				 */
1686 				struct sctp_tmit_chunk *lchk;
1687 
1688 				sctp_hs_cwnd_decrease(stcb, net);
1689 
1690 				lchk = TAILQ_FIRST(&asoc->send_queue);
1691 
1692 				net->partial_bytes_acked = 0;
1693 				/* Turn on fast recovery window */
1694 				asoc->fast_retran_loss_recovery = 1;
1695 				if (lchk == NULL) {
1696 					/* Mark end of the window */
1697 					asoc->fast_recovery_tsn = asoc->sending_seq - 1;
1698 				} else {
1699 					asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
1700 				}
1701 
1702 				/*
1703 				 * CMT fast recovery -- per destination
1704 				 * recovery variable.
1705 				 */
1706 				net->fast_retran_loss_recovery = 1;
1707 
1708 				if (lchk == NULL) {
1709 					/* Mark end of the window */
1710 					net->fast_recovery_tsn = asoc->sending_seq - 1;
1711 				} else {
1712 					net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
1713 				}
1714 
1715 				sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
1716 				    stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
1717 				sctp_timer_start(SCTP_TIMER_TYPE_SEND,
1718 				    stcb->sctp_ep, stcb, net);
1719 			}
1720 		} else if (net->net_ack > 0) {
1721 			/*
1722 			 * Mark a peg that we WOULD have done a cwnd
1723 			 * reduction but RFC2582 prevented this action.
1724 			 */
1725 			SCTP_STAT_INCR(sctps_fastretransinrtt);
1726 		}
1727 	}
1728 }
1729 
1730 static void
1731 sctp_hs_cwnd_update_after_sack(struct sctp_tcb *stcb,
1732     struct sctp_association *asoc,
1733     int accum_moved, int reneged_all SCTP_UNUSED, int will_exit)
1734 {
1735 	struct sctp_nets *net;
1736 
1737 	/******************************/
1738 	/* update cwnd and Early FR   */
1739 	/******************************/
1740 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
1741 
1742 #ifdef JANA_CMT_FAST_RECOVERY
1743 		/*
1744 		 * CMT fast recovery code. Need to debug.
1745 		 */
1746 		if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
1747 			if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) ||
1748 			    SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) {
1749 				net->will_exit_fast_recovery = 1;
1750 			}
1751 		}
1752 #endif
1753 		/* if nothing was acked on this destination skip it */
1754 		if (net->net_ack == 0) {
1755 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
1756 				sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
1757 			}
1758 			continue;
1759 		}
1760 #ifdef JANA_CMT_FAST_RECOVERY
1761 		/*
1762 		 * CMT fast recovery code
1763 		 */
1764 		/*
1765 		 * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery
1766 		 * && net->will_exit_fast_recovery == 0) { @@@ Do something
1767 		 * } else if (sctp_cmt_on_off == 0 &&
1768 		 * asoc->fast_retran_loss_recovery && will_exit == 0) {
1769 		 */
1770 #endif
1771 
1772 		if (asoc->fast_retran_loss_recovery &&
1773 		    (will_exit == 0) &&
1774 		    (asoc->sctp_cmt_on_off == 0)) {
1775 			/*
1776 			 * If we are in loss recovery we skip any cwnd
1777 			 * update
1778 			 */
1779 			return;
1780 		}
1781 		/*
1782 		 * CMT: CUC algorithm. Update cwnd if pseudo-cumack has
1783 		 * moved.
1784 		 */
1785 		if (accum_moved ||
1786 		    ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) {
1787 			/* If the cumulative ack moved we can proceed */
1788 			if (net->cwnd <= net->ssthresh) {
1789 				/* We are in slow start */
1790 				if (net->flight_size + net->net_ack >= net->cwnd) {
1791 
1792 					sctp_hs_cwnd_increase(stcb, net);
1793 
1794 				} else {
1795 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
1796 						sctp_log_cwnd(stcb, net, net->net_ack,
1797 						    SCTP_CWND_LOG_NOADV_SS);
1798 					}
1799 				}
1800 			} else {
1801 				/* We are in congestion avoidance */
1802 				net->partial_bytes_acked += net->net_ack;
1803 				if ((net->flight_size + net->net_ack >= net->cwnd) &&
1804 				    (net->partial_bytes_acked >= net->cwnd)) {
1805 					net->partial_bytes_acked -= net->cwnd;
1806 					net->cwnd += net->mtu;
1807 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1808 						sctp_log_cwnd(stcb, net, net->mtu,
1809 						    SCTP_CWND_LOG_FROM_CA);
1810 					}
1811 				} else {
1812 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
1813 						sctp_log_cwnd(stcb, net, net->net_ack,
1814 						    SCTP_CWND_LOG_NOADV_CA);
1815 					}
1816 				}
1817 			}
1818 		} else {
1819 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
1820 				sctp_log_cwnd(stcb, net, net->mtu,
1821 				    SCTP_CWND_LOG_NO_CUMACK);
1822 			}
1823 		}
1824 	}
1825 }
1826 
1827 
1828 /*
1829  * H-TCP congestion control. The algorithm is detailed in:
1830  * R.N.Shorten, D.J.Leith:
1831  *   "H-TCP: TCP for high-speed and long-distance networks"
1832  *   Proc. PFLDnet, Argonne, 2004.
1833  * http://www.hamilton.ie/net/htcp3.pdf
1834  */
1835 
1836 
1837 static int use_rtt_scaling = 1;
1838 static int use_bandwidth_switch = 1;
1839 
1840 static inline int
1841 between(uint32_t seq1, uint32_t seq2, uint32_t seq3)
1842 {
1843 	return (seq3 - seq2 >= seq1 - seq2);
1844 }
1845 
1846 static inline uint32_t
1847 htcp_cong_time(struct htcp *ca)
1848 {
1849 	return (sctp_get_tick_count() - ca->last_cong);
1850 }
1851 
1852 static inline uint32_t
1853 htcp_ccount(struct htcp *ca)
1854 {
1855 	return (htcp_cong_time(ca) / ca->minRTT);
1856 }
1857 
1858 static inline void
1859 htcp_reset(struct htcp *ca)
1860 {
1861 	ca->undo_last_cong = ca->last_cong;
1862 	ca->undo_maxRTT = ca->maxRTT;
1863 	ca->undo_old_maxB = ca->old_maxB;
1864 	ca->last_cong = sctp_get_tick_count();
1865 }
1866 
1867 #ifdef SCTP_NOT_USED
1868 
1869 static uint32_t
1870 htcp_cwnd_undo(struct sctp_tcb *stcb, struct sctp_nets *net)
1871 {
1872 	net->cc_mod.htcp_ca.last_cong = net->cc_mod.htcp_ca.undo_last_cong;
1873 	net->cc_mod.htcp_ca.maxRTT = net->cc_mod.htcp_ca.undo_maxRTT;
1874 	net->cc_mod.htcp_ca.old_maxB = net->cc_mod.htcp_ca.undo_old_maxB;
1875 	return (max(net->cwnd, ((net->ssthresh / net->mtu << 7) / net->cc_mod.htcp_ca.beta) * net->mtu));
1876 }
1877 
1878 #endif
1879 
1880 static inline void
1881 measure_rtt(struct sctp_nets *net)
1882 {
1883 	uint32_t srtt = net->lastsa >> SCTP_RTT_SHIFT;
1884 
1885 	/* keep track of minimum RTT seen so far, minRTT is zero at first */
1886 	if (net->cc_mod.htcp_ca.minRTT > srtt || !net->cc_mod.htcp_ca.minRTT)
1887 		net->cc_mod.htcp_ca.minRTT = srtt;
1888 
1889 	/* max RTT */
1890 	if (net->fast_retran_ip == 0 && net->ssthresh < 0xFFFF && htcp_ccount(&net->cc_mod.htcp_ca) > 3) {
1891 		if (net->cc_mod.htcp_ca.maxRTT < net->cc_mod.htcp_ca.minRTT)
1892 			net->cc_mod.htcp_ca.maxRTT = net->cc_mod.htcp_ca.minRTT;
1893 		if (net->cc_mod.htcp_ca.maxRTT < srtt && srtt <= net->cc_mod.htcp_ca.maxRTT + MSEC_TO_TICKS(20))
1894 			net->cc_mod.htcp_ca.maxRTT = srtt;
1895 	}
1896 }
1897 
1898 static void
1899 measure_achieved_throughput(struct sctp_nets *net)
1900 {
1901 	uint32_t now = sctp_get_tick_count();
1902 
1903 	if (net->fast_retran_ip == 0)
1904 		net->cc_mod.htcp_ca.bytes_acked = net->net_ack;
1905 
1906 	if (!use_bandwidth_switch)
1907 		return;
1908 
1909 	/* achieved throughput calculations */
1910 	/* JRS - not 100% sure of this statement */
1911 	if (net->fast_retran_ip == 1) {
1912 		net->cc_mod.htcp_ca.bytecount = 0;
1913 		net->cc_mod.htcp_ca.lasttime = now;
1914 		return;
1915 	}
1916 	net->cc_mod.htcp_ca.bytecount += net->net_ack;
1917 	if ((net->cc_mod.htcp_ca.bytecount >= net->cwnd - (((net->cc_mod.htcp_ca.alpha >> 7) ? (net->cc_mod.htcp_ca.alpha >> 7) : 1) * net->mtu)) &&
1918 	    (now - net->cc_mod.htcp_ca.lasttime >= net->cc_mod.htcp_ca.minRTT) &&
1919 	    (net->cc_mod.htcp_ca.minRTT > 0)) {
1920 		uint32_t cur_Bi = net->cc_mod.htcp_ca.bytecount / net->mtu * hz / (now - net->cc_mod.htcp_ca.lasttime);
1921 
1922 		if (htcp_ccount(&net->cc_mod.htcp_ca) <= 3) {
1923 			/* just after backoff */
1924 			net->cc_mod.htcp_ca.minB = net->cc_mod.htcp_ca.maxB = net->cc_mod.htcp_ca.Bi = cur_Bi;
1925 		} else {
1926 			net->cc_mod.htcp_ca.Bi = (3 * net->cc_mod.htcp_ca.Bi + cur_Bi) / 4;
1927 			if (net->cc_mod.htcp_ca.Bi > net->cc_mod.htcp_ca.maxB)
1928 				net->cc_mod.htcp_ca.maxB = net->cc_mod.htcp_ca.Bi;
1929 			if (net->cc_mod.htcp_ca.minB > net->cc_mod.htcp_ca.maxB)
1930 				net->cc_mod.htcp_ca.minB = net->cc_mod.htcp_ca.maxB;
1931 		}
1932 		net->cc_mod.htcp_ca.bytecount = 0;
1933 		net->cc_mod.htcp_ca.lasttime = now;
1934 	}
1935 }
1936 
1937 static inline void
1938 htcp_beta_update(struct htcp *ca, uint32_t minRTT, uint32_t maxRTT)
1939 {
1940 	if (use_bandwidth_switch) {
1941 		uint32_t maxB = ca->maxB;
1942 		uint32_t old_maxB = ca->old_maxB;
1943 
1944 		ca->old_maxB = ca->maxB;
1945 
1946 		if (!between(5 * maxB, 4 * old_maxB, 6 * old_maxB)) {
1947 			ca->beta = BETA_MIN;
1948 			ca->modeswitch = 0;
1949 			return;
1950 		}
1951 	}
1952 	if (ca->modeswitch && minRTT > (uint32_t) MSEC_TO_TICKS(10) && maxRTT) {
1953 		ca->beta = (minRTT << 7) / maxRTT;
1954 		if (ca->beta < BETA_MIN)
1955 			ca->beta = BETA_MIN;
1956 		else if (ca->beta > BETA_MAX)
1957 			ca->beta = BETA_MAX;
1958 	} else {
1959 		ca->beta = BETA_MIN;
1960 		ca->modeswitch = 1;
1961 	}
1962 }
1963 
1964 static inline void
1965 htcp_alpha_update(struct htcp *ca)
1966 {
1967 	uint32_t minRTT = ca->minRTT;
1968 	uint32_t factor = 1;
1969 	uint32_t diff = htcp_cong_time(ca);
1970 
1971 	if (diff > (uint32_t) hz) {
1972 		diff -= hz;
1973 		factor = 1 + (10 * diff + ((diff / 2) * (diff / 2) / hz)) / hz;
1974 	}
1975 	if (use_rtt_scaling && minRTT) {
1976 		uint32_t scale = (hz << 3) / (10 * minRTT);
1977 
1978 		scale = min(max(scale, 1U << 2), 10U << 3);	/* clamping ratio to
1979 								 * interval [0.5,10]<<3 */
1980 		factor = (factor << 3) / scale;
1981 		if (!factor)
1982 			factor = 1;
1983 	}
1984 	ca->alpha = 2 * factor * ((1 << 7) - ca->beta);
1985 	if (!ca->alpha)
1986 		ca->alpha = ALPHA_BASE;
1987 }
1988 
1989 /* After we have the rtt data to calculate beta, we'd still prefer to wait one
1990  * rtt before we adjust our beta to ensure we are working from a consistent
1991  * data.
1992  *
1993  * This function should be called when we hit a congestion event since only at
1994  * that point do we really have a real sense of maxRTT (the queues en route
1995  * were getting just too full now).
1996  */
1997 static void
1998 htcp_param_update(struct sctp_nets *net)
1999 {
2000 	uint32_t minRTT = net->cc_mod.htcp_ca.minRTT;
2001 	uint32_t maxRTT = net->cc_mod.htcp_ca.maxRTT;
2002 
2003 	htcp_beta_update(&net->cc_mod.htcp_ca, minRTT, maxRTT);
2004 	htcp_alpha_update(&net->cc_mod.htcp_ca);
2005 
2006 	/*
2007 	 * add slowly fading memory for maxRTT to accommodate routing
2008 	 * changes etc
2009 	 */
2010 	if (minRTT > 0 && maxRTT > minRTT)
2011 		net->cc_mod.htcp_ca.maxRTT = minRTT + ((maxRTT - minRTT) * 95) / 100;
2012 }
2013 
2014 static uint32_t
2015 htcp_recalc_ssthresh(struct sctp_nets *net)
2016 {
2017 	htcp_param_update(net);
2018 	return (max(((net->cwnd / net->mtu * net->cc_mod.htcp_ca.beta) >> 7) * net->mtu, 2U * net->mtu));
2019 }
2020 
2021 static void
2022 htcp_cong_avoid(struct sctp_tcb *stcb, struct sctp_nets *net)
2023 {
2024 	/*-
2025 	 * How to handle these functions?
2026          *	if (!tcp_is_cwnd_limited(sk, in_flight)) RRS - good question.
2027 	 *		return;
2028 	 */
2029 	if (net->cwnd <= net->ssthresh) {
2030 		/* We are in slow start */
2031 		if (net->flight_size + net->net_ack >= net->cwnd) {
2032 			if (net->net_ack > (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable))) {
2033 				net->cwnd += (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable));
2034 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2035 					sctp_log_cwnd(stcb, net, net->mtu,
2036 					    SCTP_CWND_LOG_FROM_SS);
2037 				}
2038 			} else {
2039 				net->cwnd += net->net_ack;
2040 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2041 					sctp_log_cwnd(stcb, net, net->net_ack,
2042 					    SCTP_CWND_LOG_FROM_SS);
2043 				}
2044 			}
2045 		} else {
2046 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
2047 				sctp_log_cwnd(stcb, net, net->net_ack,
2048 				    SCTP_CWND_LOG_NOADV_SS);
2049 			}
2050 		}
2051 	} else {
2052 		measure_rtt(net);
2053 
2054 		/*
2055 		 * In dangerous area, increase slowly. In theory this is
2056 		 * net->cwnd += alpha / net->cwnd
2057 		 */
2058 		/* What is snd_cwnd_cnt?? */
2059 		if (((net->partial_bytes_acked / net->mtu * net->cc_mod.htcp_ca.alpha) >> 7) * net->mtu >= net->cwnd) {
2060 			/*-
2061 			 * Does SCTP have a cwnd clamp?
2062 			 * if (net->snd_cwnd < net->snd_cwnd_clamp) - Nope (RRS).
2063 			 */
2064 			net->cwnd += net->mtu;
2065 			net->partial_bytes_acked = 0;
2066 			htcp_alpha_update(&net->cc_mod.htcp_ca);
2067 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2068 				sctp_log_cwnd(stcb, net, net->mtu,
2069 				    SCTP_CWND_LOG_FROM_CA);
2070 			}
2071 		} else {
2072 			net->partial_bytes_acked += net->net_ack;
2073 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
2074 				sctp_log_cwnd(stcb, net, net->net_ack,
2075 				    SCTP_CWND_LOG_NOADV_CA);
2076 			}
2077 		}
2078 
2079 		net->cc_mod.htcp_ca.bytes_acked = net->mtu;
2080 	}
2081 }
2082 
2083 #ifdef SCTP_NOT_USED
2084 /* Lower bound on congestion window. */
2085 static uint32_t
2086 htcp_min_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net)
2087 {
2088 	return (net->ssthresh);
2089 }
2090 
2091 #endif
2092 
2093 static void
2094 htcp_init(struct sctp_nets *net)
2095 {
2096 	memset(&net->cc_mod.htcp_ca, 0, sizeof(struct htcp));
2097 	net->cc_mod.htcp_ca.alpha = ALPHA_BASE;
2098 	net->cc_mod.htcp_ca.beta = BETA_MIN;
2099 	net->cc_mod.htcp_ca.bytes_acked = net->mtu;
2100 	net->cc_mod.htcp_ca.last_cong = sctp_get_tick_count();
2101 }
2102 
2103 static void
2104 sctp_htcp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
2105 {
2106 	/*
2107 	 * We take the max of the burst limit times a MTU or the
2108 	 * INITIAL_CWND. We then limit this to 4 MTU's of sending.
2109 	 */
2110 	net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
2111 	net->ssthresh = stcb->asoc.peers_rwnd;
2112 	htcp_init(net);
2113 
2114 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
2115 		sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
2116 	}
2117 }
2118 
2119 static void
2120 sctp_htcp_cwnd_update_after_sack(struct sctp_tcb *stcb,
2121     struct sctp_association *asoc,
2122     int accum_moved, int reneged_all SCTP_UNUSED, int will_exit)
2123 {
2124 	struct sctp_nets *net;
2125 
2126 	/******************************/
2127 	/* update cwnd and Early FR   */
2128 	/******************************/
2129 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
2130 
2131 #ifdef JANA_CMT_FAST_RECOVERY
2132 		/*
2133 		 * CMT fast recovery code. Need to debug.
2134 		 */
2135 		if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
2136 			if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) ||
2137 			    SCTP_TSN_GE(net->pseudo_cumack, net->fast_recovery_tsn)) {
2138 				net->will_exit_fast_recovery = 1;
2139 			}
2140 		}
2141 #endif
2142 		/* if nothing was acked on this destination skip it */
2143 		if (net->net_ack == 0) {
2144 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
2145 				sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
2146 			}
2147 			continue;
2148 		}
2149 #ifdef JANA_CMT_FAST_RECOVERY
2150 		/*
2151 		 * CMT fast recovery code
2152 		 */
2153 		/*
2154 		 * if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery
2155 		 * && net->will_exit_fast_recovery == 0) { @@@ Do something
2156 		 * } else if (sctp_cmt_on_off == 0 &&
2157 		 * asoc->fast_retran_loss_recovery && will_exit == 0) {
2158 		 */
2159 #endif
2160 
2161 		if (asoc->fast_retran_loss_recovery &&
2162 		    will_exit == 0 &&
2163 		    (asoc->sctp_cmt_on_off == 0)) {
2164 			/*
2165 			 * If we are in loss recovery we skip any cwnd
2166 			 * update
2167 			 */
2168 			return;
2169 		}
2170 		/*
2171 		 * CMT: CUC algorithm. Update cwnd if pseudo-cumack has
2172 		 * moved.
2173 		 */
2174 		if (accum_moved ||
2175 		    ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) {
2176 			htcp_cong_avoid(stcb, net);
2177 			measure_achieved_throughput(net);
2178 		} else {
2179 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
2180 				sctp_log_cwnd(stcb, net, net->mtu,
2181 				    SCTP_CWND_LOG_NO_CUMACK);
2182 			}
2183 		}
2184 	}
2185 }
2186 
2187 static void
2188 sctp_htcp_cwnd_update_after_fr(struct sctp_tcb *stcb,
2189     struct sctp_association *asoc)
2190 {
2191 	struct sctp_nets *net;
2192 
2193 	/*
2194 	 * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) &&
2195 	 * (net->fast_retran_loss_recovery == 0)))
2196 	 */
2197 	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
2198 		if ((asoc->fast_retran_loss_recovery == 0) ||
2199 		    (asoc->sctp_cmt_on_off > 0)) {
2200 			/* out of a RFC2582 Fast recovery window? */
2201 			if (net->net_ack > 0) {
2202 				/*
2203 				 * per section 7.2.3, are there any
2204 				 * destinations that had a fast retransmit
2205 				 * to them. If so what we need to do is
2206 				 * adjust ssthresh and cwnd.
2207 				 */
2208 				struct sctp_tmit_chunk *lchk;
2209 				int old_cwnd = net->cwnd;
2210 
2211 				/* JRS - reset as if state were changed */
2212 				htcp_reset(&net->cc_mod.htcp_ca);
2213 				net->ssthresh = htcp_recalc_ssthresh(net);
2214 				net->cwnd = net->ssthresh;
2215 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2216 					sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
2217 					    SCTP_CWND_LOG_FROM_FR);
2218 				}
2219 				lchk = TAILQ_FIRST(&asoc->send_queue);
2220 
2221 				net->partial_bytes_acked = 0;
2222 				/* Turn on fast recovery window */
2223 				asoc->fast_retran_loss_recovery = 1;
2224 				if (lchk == NULL) {
2225 					/* Mark end of the window */
2226 					asoc->fast_recovery_tsn = asoc->sending_seq - 1;
2227 				} else {
2228 					asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
2229 				}
2230 
2231 				/*
2232 				 * CMT fast recovery -- per destination
2233 				 * recovery variable.
2234 				 */
2235 				net->fast_retran_loss_recovery = 1;
2236 
2237 				if (lchk == NULL) {
2238 					/* Mark end of the window */
2239 					net->fast_recovery_tsn = asoc->sending_seq - 1;
2240 				} else {
2241 					net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
2242 				}
2243 
2244 				sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
2245 				    stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
2246 				sctp_timer_start(SCTP_TIMER_TYPE_SEND,
2247 				    stcb->sctp_ep, stcb, net);
2248 			}
2249 		} else if (net->net_ack > 0) {
2250 			/*
2251 			 * Mark a peg that we WOULD have done a cwnd
2252 			 * reduction but RFC2582 prevented this action.
2253 			 */
2254 			SCTP_STAT_INCR(sctps_fastretransinrtt);
2255 		}
2256 	}
2257 }
2258 
2259 static void
2260 sctp_htcp_cwnd_update_after_timeout(struct sctp_tcb *stcb,
2261     struct sctp_nets *net)
2262 {
2263 	int old_cwnd = net->cwnd;
2264 
2265 	/* JRS - reset as if the state were being changed to timeout */
2266 	htcp_reset(&net->cc_mod.htcp_ca);
2267 	net->ssthresh = htcp_recalc_ssthresh(net);
2268 	net->cwnd = net->mtu;
2269 	net->partial_bytes_acked = 0;
2270 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2271 		sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX);
2272 	}
2273 }
2274 
2275 static void
2276 sctp_htcp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb,
2277     struct sctp_nets *net, int in_window, int num_pkt_lost SCTP_UNUSED)
2278 {
2279 	int old_cwnd;
2280 
2281 	old_cwnd = net->cwnd;
2282 
2283 	/* JRS - reset hctp as if state changed */
2284 	if (in_window == 0) {
2285 		htcp_reset(&net->cc_mod.htcp_ca);
2286 		SCTP_STAT_INCR(sctps_ecnereducedcwnd);
2287 		net->ssthresh = htcp_recalc_ssthresh(net);
2288 		if (net->ssthresh < net->mtu) {
2289 			net->ssthresh = net->mtu;
2290 			/* here back off the timer as well, to slow us down */
2291 			net->RTO <<= 1;
2292 		}
2293 		net->cwnd = net->ssthresh;
2294 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
2295 			sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
2296 		}
2297 	}
2298 }
2299 
2300 struct sctp_cc_functions sctp_cc_functions[] = {
2301 	{
2302 		.sctp_set_initial_cc_param = sctp_set_initial_cc_param,
2303 		.sctp_cwnd_update_after_sack = sctp_cwnd_update_after_sack,
2304 		.sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common,
2305 		.sctp_cwnd_update_after_fr = sctp_cwnd_update_after_fr,
2306 		.sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout,
2307 		.sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo,
2308 		.sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped,
2309 		.sctp_cwnd_update_after_output = sctp_cwnd_update_after_output,
2310 	},
2311 	{
2312 		.sctp_set_initial_cc_param = sctp_set_initial_cc_param,
2313 		.sctp_cwnd_update_after_sack = sctp_hs_cwnd_update_after_sack,
2314 		.sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common,
2315 		.sctp_cwnd_update_after_fr = sctp_hs_cwnd_update_after_fr,
2316 		.sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout,
2317 		.sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo,
2318 		.sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped,
2319 		.sctp_cwnd_update_after_output = sctp_cwnd_update_after_output,
2320 	},
2321 	{
2322 		.sctp_set_initial_cc_param = sctp_htcp_set_initial_cc_param,
2323 		.sctp_cwnd_update_after_sack = sctp_htcp_cwnd_update_after_sack,
2324 		.sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common,
2325 		.sctp_cwnd_update_after_fr = sctp_htcp_cwnd_update_after_fr,
2326 		.sctp_cwnd_update_after_timeout = sctp_htcp_cwnd_update_after_timeout,
2327 		.sctp_cwnd_update_after_ecn_echo = sctp_htcp_cwnd_update_after_ecn_echo,
2328 		.sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped,
2329 		.sctp_cwnd_update_after_output = sctp_cwnd_update_after_output,
2330 	},
2331 	{
2332 		.sctp_set_initial_cc_param = sctp_set_rtcc_initial_cc_param,
2333 		.sctp_cwnd_update_after_sack = sctp_cwnd_update_rtcc_after_sack,
2334 		.sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common,
2335 		.sctp_cwnd_update_after_fr = sctp_cwnd_update_after_fr,
2336 		.sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout,
2337 		.sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_rtcc_after_ecn_echo,
2338 		.sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped,
2339 		.sctp_cwnd_update_after_output = sctp_cwnd_update_after_output,
2340 		.sctp_cwnd_update_packet_transmitted = sctp_cwnd_update_rtcc_packet_transmitted,
2341 		.sctp_cwnd_update_tsn_acknowledged = sctp_cwnd_update_rtcc_tsn_acknowledged,
2342 		.sctp_cwnd_new_transmission_begins = sctp_cwnd_new_rtcc_transmission_begins,
2343 		.sctp_cwnd_prepare_net_for_sack = sctp_cwnd_prepare_rtcc_net_for_sack,
2344 		.sctp_cwnd_socket_option = sctp_cwnd_rtcc_socket_option,
2345 		.sctp_rtt_calculated = sctp_rtt_rtcc_calculated
2346 	}
2347 };
2348