1 /* Management of Tx window, Tx resend, ACKs and out-of-sequence reception
2  *
3  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4  * Written by David Howells (dhowells@redhat.com)
5  *
6  * This program is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU General Public License
8  * as published by the Free Software Foundation; either version
9  * 2 of the License, or (at your option) any later version.
10  */
11 
12 #include <linux/module.h>
13 #include <linux/circ_buf.h>
14 #include <linux/net.h>
15 #include <linux/skbuff.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <net/sock.h>
19 #include <net/af_rxrpc.h>
20 #include "ar-internal.h"
21 
22 /*
23  * How long to wait before scheduling ACK generation after seeing a
24  * packet with RXRPC_REQUEST_ACK set (in jiffies).
25  */
26 unsigned rxrpc_requested_ack_delay = 1;
27 
28 /*
29  * How long to wait before scheduling an ACK with subtype DELAY (in jiffies).
30  *
31  * We use this when we've received new data packets.  If those packets aren't
32  * all consumed within this time we will send a DELAY ACK if an ACK was not
33  * requested to let the sender know it doesn't need to resend.
34  */
35 unsigned rxrpc_soft_ack_delay = 1 * HZ;
36 
37 /*
38  * How long to wait before scheduling an ACK with subtype IDLE (in jiffies).
39  *
40  * We use this when we've consumed some previously soft-ACK'd packets when
41  * further packets aren't immediately received to decide when to send an IDLE
42  * ACK let the other end know that it can free up its Tx buffer space.
43  */
44 unsigned rxrpc_idle_ack_delay = 0.5 * HZ;
45 
46 /*
47  * Receive window size in packets.  This indicates the maximum number of
48  * unconsumed received packets we're willing to retain in memory.  Once this
49  * limit is hit, we should generate an EXCEEDS_WINDOW ACK and discard further
50  * packets.
51  */
52 unsigned rxrpc_rx_window_size = 32;
53 
54 /*
55  * Maximum Rx MTU size.  This indicates to the sender the size of jumbo packet
56  * made by gluing normal packets together that we're willing to handle.
57  */
58 unsigned rxrpc_rx_mtu = 5692;
59 
60 /*
61  * The maximum number of fragments in a received jumbo packet that we tell the
62  * sender that we're willing to handle.
63  */
64 unsigned rxrpc_rx_jumbo_max = 4;
65 
rxrpc_acks(u8 reason)66 static const char *rxrpc_acks(u8 reason)
67 {
68 	static const char *const str[] = {
69 		"---", "REQ", "DUP", "OOS", "WIN", "MEM", "PNG", "PNR", "DLY",
70 		"IDL", "-?-"
71 	};
72 
73 	if (reason >= ARRAY_SIZE(str))
74 		reason = ARRAY_SIZE(str) - 1;
75 	return str[reason];
76 }
77 
78 static const s8 rxrpc_ack_priority[] = {
79 	[0]				= 0,
80 	[RXRPC_ACK_DELAY]		= 1,
81 	[RXRPC_ACK_REQUESTED]		= 2,
82 	[RXRPC_ACK_IDLE]		= 3,
83 	[RXRPC_ACK_PING_RESPONSE]	= 4,
84 	[RXRPC_ACK_DUPLICATE]		= 5,
85 	[RXRPC_ACK_OUT_OF_SEQUENCE]	= 6,
86 	[RXRPC_ACK_EXCEEDS_WINDOW]	= 7,
87 	[RXRPC_ACK_NOSPACE]		= 8,
88 };
89 
90 /*
91  * propose an ACK be sent
92  */
__rxrpc_propose_ACK(struct rxrpc_call * call,u8 ack_reason,__be32 serial,bool immediate)93 void __rxrpc_propose_ACK(struct rxrpc_call *call, u8 ack_reason,
94 			 __be32 serial, bool immediate)
95 {
96 	unsigned long expiry;
97 	s8 prior = rxrpc_ack_priority[ack_reason];
98 
99 	ASSERTCMP(prior, >, 0);
100 
101 	_enter("{%d},%s,%%%x,%u",
102 	       call->debug_id, rxrpc_acks(ack_reason), ntohl(serial),
103 	       immediate);
104 
105 	if (prior < rxrpc_ack_priority[call->ackr_reason]) {
106 		if (immediate)
107 			goto cancel_timer;
108 		return;
109 	}
110 
111 	/* update DELAY, IDLE, REQUESTED and PING_RESPONSE ACK serial
112 	 * numbers */
113 	if (prior == rxrpc_ack_priority[call->ackr_reason]) {
114 		if (prior <= 4)
115 			call->ackr_serial = serial;
116 		if (immediate)
117 			goto cancel_timer;
118 		return;
119 	}
120 
121 	call->ackr_reason = ack_reason;
122 	call->ackr_serial = serial;
123 
124 	switch (ack_reason) {
125 	case RXRPC_ACK_DELAY:
126 		_debug("run delay timer");
127 		expiry = rxrpc_soft_ack_delay;
128 		goto run_timer;
129 
130 	case RXRPC_ACK_IDLE:
131 		if (!immediate) {
132 			_debug("run defer timer");
133 			expiry = rxrpc_idle_ack_delay;
134 			goto run_timer;
135 		}
136 		goto cancel_timer;
137 
138 	case RXRPC_ACK_REQUESTED:
139 		expiry = rxrpc_requested_ack_delay;
140 		if (!expiry)
141 			goto cancel_timer;
142 		if (!immediate || serial == cpu_to_be32(1)) {
143 			_debug("run defer timer");
144 			goto run_timer;
145 		}
146 
147 	default:
148 		_debug("immediate ACK");
149 		goto cancel_timer;
150 	}
151 
152 run_timer:
153 	expiry += jiffies;
154 	if (!timer_pending(&call->ack_timer) ||
155 	    time_after(call->ack_timer.expires, expiry))
156 		mod_timer(&call->ack_timer, expiry);
157 	return;
158 
159 cancel_timer:
160 	_debug("cancel timer %%%u", ntohl(serial));
161 	try_to_del_timer_sync(&call->ack_timer);
162 	read_lock_bh(&call->state_lock);
163 	if (call->state <= RXRPC_CALL_COMPLETE &&
164 	    !test_and_set_bit(RXRPC_CALL_ACK, &call->events))
165 		rxrpc_queue_call(call);
166 	read_unlock_bh(&call->state_lock);
167 }
168 
169 /*
170  * propose an ACK be sent, locking the call structure
171  */
rxrpc_propose_ACK(struct rxrpc_call * call,u8 ack_reason,__be32 serial,bool immediate)172 void rxrpc_propose_ACK(struct rxrpc_call *call, u8 ack_reason,
173 		       __be32 serial, bool immediate)
174 {
175 	s8 prior = rxrpc_ack_priority[ack_reason];
176 
177 	if (prior > rxrpc_ack_priority[call->ackr_reason]) {
178 		spin_lock_bh(&call->lock);
179 		__rxrpc_propose_ACK(call, ack_reason, serial, immediate);
180 		spin_unlock_bh(&call->lock);
181 	}
182 }
183 
184 /*
185  * set the resend timer
186  */
rxrpc_set_resend(struct rxrpc_call * call,u8 resend,unsigned long resend_at)187 static void rxrpc_set_resend(struct rxrpc_call *call, u8 resend,
188 			     unsigned long resend_at)
189 {
190 	read_lock_bh(&call->state_lock);
191 	if (call->state >= RXRPC_CALL_COMPLETE)
192 		resend = 0;
193 
194 	if (resend & 1) {
195 		_debug("SET RESEND");
196 		set_bit(RXRPC_CALL_RESEND, &call->events);
197 	}
198 
199 	if (resend & 2) {
200 		_debug("MODIFY RESEND TIMER");
201 		set_bit(RXRPC_CALL_RUN_RTIMER, &call->flags);
202 		mod_timer(&call->resend_timer, resend_at);
203 	} else {
204 		_debug("KILL RESEND TIMER");
205 		del_timer_sync(&call->resend_timer);
206 		clear_bit(RXRPC_CALL_RESEND_TIMER, &call->events);
207 		clear_bit(RXRPC_CALL_RUN_RTIMER, &call->flags);
208 	}
209 	read_unlock_bh(&call->state_lock);
210 }
211 
212 /*
213  * resend packets
214  */
rxrpc_resend(struct rxrpc_call * call)215 static void rxrpc_resend(struct rxrpc_call *call)
216 {
217 	struct rxrpc_skb_priv *sp;
218 	struct rxrpc_header *hdr;
219 	struct sk_buff *txb;
220 	unsigned long *p_txb, resend_at;
221 	bool stop;
222 	int loop;
223 	u8 resend;
224 
225 	_enter("{%d,%d,%d,%d},",
226 	       call->acks_hard, call->acks_unacked,
227 	       atomic_read(&call->sequence),
228 	       CIRC_CNT(call->acks_head, call->acks_tail, call->acks_winsz));
229 
230 	stop = false;
231 	resend = 0;
232 	resend_at = 0;
233 
234 	for (loop = call->acks_tail;
235 	     loop != call->acks_head || stop;
236 	     loop = (loop + 1) &  (call->acks_winsz - 1)
237 	     ) {
238 		p_txb = call->acks_window + loop;
239 		smp_read_barrier_depends();
240 		if (*p_txb & 1)
241 			continue;
242 
243 		txb = (struct sk_buff *) *p_txb;
244 		sp = rxrpc_skb(txb);
245 
246 		if (sp->need_resend) {
247 			sp->need_resend = false;
248 
249 			/* each Tx packet has a new serial number */
250 			sp->hdr.serial =
251 				htonl(atomic_inc_return(&call->conn->serial));
252 
253 			hdr = (struct rxrpc_header *) txb->head;
254 			hdr->serial = sp->hdr.serial;
255 
256 			_proto("Tx DATA %%%u { #%d }",
257 			       ntohl(sp->hdr.serial), ntohl(sp->hdr.seq));
258 			if (rxrpc_send_packet(call->conn->trans, txb) < 0) {
259 				stop = true;
260 				sp->resend_at = jiffies + 3;
261 			} else {
262 				sp->resend_at =
263 					jiffies + rxrpc_resend_timeout;
264 			}
265 		}
266 
267 		if (time_after_eq(jiffies + 1, sp->resend_at)) {
268 			sp->need_resend = true;
269 			resend |= 1;
270 		} else if (resend & 2) {
271 			if (time_before(sp->resend_at, resend_at))
272 				resend_at = sp->resend_at;
273 		} else {
274 			resend_at = sp->resend_at;
275 			resend |= 2;
276 		}
277 	}
278 
279 	rxrpc_set_resend(call, resend, resend_at);
280 	_leave("");
281 }
282 
283 /*
284  * handle resend timer expiry
285  */
rxrpc_resend_timer(struct rxrpc_call * call)286 static void rxrpc_resend_timer(struct rxrpc_call *call)
287 {
288 	struct rxrpc_skb_priv *sp;
289 	struct sk_buff *txb;
290 	unsigned long *p_txb, resend_at;
291 	int loop;
292 	u8 resend;
293 
294 	_enter("%d,%d,%d",
295 	       call->acks_tail, call->acks_unacked, call->acks_head);
296 
297 	if (call->state >= RXRPC_CALL_COMPLETE)
298 		return;
299 
300 	resend = 0;
301 	resend_at = 0;
302 
303 	for (loop = call->acks_unacked;
304 	     loop != call->acks_head;
305 	     loop = (loop + 1) &  (call->acks_winsz - 1)
306 	     ) {
307 		p_txb = call->acks_window + loop;
308 		smp_read_barrier_depends();
309 		txb = (struct sk_buff *) (*p_txb & ~1);
310 		sp = rxrpc_skb(txb);
311 
312 		ASSERT(!(*p_txb & 1));
313 
314 		if (sp->need_resend) {
315 			;
316 		} else if (time_after_eq(jiffies + 1, sp->resend_at)) {
317 			sp->need_resend = true;
318 			resend |= 1;
319 		} else if (resend & 2) {
320 			if (time_before(sp->resend_at, resend_at))
321 				resend_at = sp->resend_at;
322 		} else {
323 			resend_at = sp->resend_at;
324 			resend |= 2;
325 		}
326 	}
327 
328 	rxrpc_set_resend(call, resend, resend_at);
329 	_leave("");
330 }
331 
332 /*
333  * process soft ACKs of our transmitted packets
334  * - these indicate packets the peer has or has not received, but hasn't yet
335  *   given to the consumer, and so can still be discarded and re-requested
336  */
rxrpc_process_soft_ACKs(struct rxrpc_call * call,struct rxrpc_ackpacket * ack,struct sk_buff * skb)337 static int rxrpc_process_soft_ACKs(struct rxrpc_call *call,
338 				   struct rxrpc_ackpacket *ack,
339 				   struct sk_buff *skb)
340 {
341 	struct rxrpc_skb_priv *sp;
342 	struct sk_buff *txb;
343 	unsigned long *p_txb, resend_at;
344 	int loop;
345 	u8 sacks[RXRPC_MAXACKS], resend;
346 
347 	_enter("{%d,%d},{%d},",
348 	       call->acks_hard,
349 	       CIRC_CNT(call->acks_head, call->acks_tail, call->acks_winsz),
350 	       ack->nAcks);
351 
352 	if (skb_copy_bits(skb, 0, sacks, ack->nAcks) < 0)
353 		goto protocol_error;
354 
355 	resend = 0;
356 	resend_at = 0;
357 	for (loop = 0; loop < ack->nAcks; loop++) {
358 		p_txb = call->acks_window;
359 		p_txb += (call->acks_tail + loop) & (call->acks_winsz - 1);
360 		smp_read_barrier_depends();
361 		txb = (struct sk_buff *) (*p_txb & ~1);
362 		sp = rxrpc_skb(txb);
363 
364 		switch (sacks[loop]) {
365 		case RXRPC_ACK_TYPE_ACK:
366 			sp->need_resend = false;
367 			*p_txb |= 1;
368 			break;
369 		case RXRPC_ACK_TYPE_NACK:
370 			sp->need_resend = true;
371 			*p_txb &= ~1;
372 			resend = 1;
373 			break;
374 		default:
375 			_debug("Unsupported ACK type %d", sacks[loop]);
376 			goto protocol_error;
377 		}
378 	}
379 
380 	smp_mb();
381 	call->acks_unacked = (call->acks_tail + loop) & (call->acks_winsz - 1);
382 
383 	/* anything not explicitly ACK'd is implicitly NACK'd, but may just not
384 	 * have been received or processed yet by the far end */
385 	for (loop = call->acks_unacked;
386 	     loop != call->acks_head;
387 	     loop = (loop + 1) &  (call->acks_winsz - 1)
388 	     ) {
389 		p_txb = call->acks_window + loop;
390 		smp_read_barrier_depends();
391 		txb = (struct sk_buff *) (*p_txb & ~1);
392 		sp = rxrpc_skb(txb);
393 
394 		if (*p_txb & 1) {
395 			/* packet must have been discarded */
396 			sp->need_resend = true;
397 			*p_txb &= ~1;
398 			resend |= 1;
399 		} else if (sp->need_resend) {
400 			;
401 		} else if (time_after_eq(jiffies + 1, sp->resend_at)) {
402 			sp->need_resend = true;
403 			resend |= 1;
404 		} else if (resend & 2) {
405 			if (time_before(sp->resend_at, resend_at))
406 				resend_at = sp->resend_at;
407 		} else {
408 			resend_at = sp->resend_at;
409 			resend |= 2;
410 		}
411 	}
412 
413 	rxrpc_set_resend(call, resend, resend_at);
414 	_leave(" = 0");
415 	return 0;
416 
417 protocol_error:
418 	_leave(" = -EPROTO");
419 	return -EPROTO;
420 }
421 
422 /*
423  * discard hard-ACK'd packets from the Tx window
424  */
rxrpc_rotate_tx_window(struct rxrpc_call * call,u32 hard)425 static void rxrpc_rotate_tx_window(struct rxrpc_call *call, u32 hard)
426 {
427 	unsigned long _skb;
428 	int tail = call->acks_tail, old_tail;
429 	int win = CIRC_CNT(call->acks_head, tail, call->acks_winsz);
430 
431 	_enter("{%u,%u},%u", call->acks_hard, win, hard);
432 
433 	ASSERTCMP(hard - call->acks_hard, <=, win);
434 
435 	while (call->acks_hard < hard) {
436 		smp_read_barrier_depends();
437 		_skb = call->acks_window[tail] & ~1;
438 		rxrpc_free_skb((struct sk_buff *) _skb);
439 		old_tail = tail;
440 		tail = (tail + 1) & (call->acks_winsz - 1);
441 		call->acks_tail = tail;
442 		if (call->acks_unacked == old_tail)
443 			call->acks_unacked = tail;
444 		call->acks_hard++;
445 	}
446 
447 	wake_up(&call->tx_waitq);
448 }
449 
450 /*
451  * clear the Tx window in the event of a failure
452  */
rxrpc_clear_tx_window(struct rxrpc_call * call)453 static void rxrpc_clear_tx_window(struct rxrpc_call *call)
454 {
455 	rxrpc_rotate_tx_window(call, atomic_read(&call->sequence));
456 }
457 
458 /*
459  * drain the out of sequence received packet queue into the packet Rx queue
460  */
rxrpc_drain_rx_oos_queue(struct rxrpc_call * call)461 static int rxrpc_drain_rx_oos_queue(struct rxrpc_call *call)
462 {
463 	struct rxrpc_skb_priv *sp;
464 	struct sk_buff *skb;
465 	bool terminal;
466 	int ret;
467 
468 	_enter("{%d,%d}", call->rx_data_post, call->rx_first_oos);
469 
470 	spin_lock_bh(&call->lock);
471 
472 	ret = -ECONNRESET;
473 	if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
474 		goto socket_unavailable;
475 
476 	skb = skb_dequeue(&call->rx_oos_queue);
477 	if (skb) {
478 		sp = rxrpc_skb(skb);
479 
480 		_debug("drain OOS packet %d [%d]",
481 		       ntohl(sp->hdr.seq), call->rx_first_oos);
482 
483 		if (ntohl(sp->hdr.seq) != call->rx_first_oos) {
484 			skb_queue_head(&call->rx_oos_queue, skb);
485 			call->rx_first_oos = ntohl(rxrpc_skb(skb)->hdr.seq);
486 			_debug("requeue %p {%u}", skb, call->rx_first_oos);
487 		} else {
488 			skb->mark = RXRPC_SKB_MARK_DATA;
489 			terminal = ((sp->hdr.flags & RXRPC_LAST_PACKET) &&
490 				!(sp->hdr.flags & RXRPC_CLIENT_INITIATED));
491 			ret = rxrpc_queue_rcv_skb(call, skb, true, terminal);
492 			BUG_ON(ret < 0);
493 			_debug("drain #%u", call->rx_data_post);
494 			call->rx_data_post++;
495 
496 			/* find out what the next packet is */
497 			skb = skb_peek(&call->rx_oos_queue);
498 			if (skb)
499 				call->rx_first_oos =
500 					ntohl(rxrpc_skb(skb)->hdr.seq);
501 			else
502 				call->rx_first_oos = 0;
503 			_debug("peek %p {%u}", skb, call->rx_first_oos);
504 		}
505 	}
506 
507 	ret = 0;
508 socket_unavailable:
509 	spin_unlock_bh(&call->lock);
510 	_leave(" = %d", ret);
511 	return ret;
512 }
513 
514 /*
515  * insert an out of sequence packet into the buffer
516  */
rxrpc_insert_oos_packet(struct rxrpc_call * call,struct sk_buff * skb)517 static void rxrpc_insert_oos_packet(struct rxrpc_call *call,
518 				    struct sk_buff *skb)
519 {
520 	struct rxrpc_skb_priv *sp, *psp;
521 	struct sk_buff *p;
522 	u32 seq;
523 
524 	sp = rxrpc_skb(skb);
525 	seq = ntohl(sp->hdr.seq);
526 	_enter(",,{%u}", seq);
527 
528 	skb->destructor = rxrpc_packet_destructor;
529 	ASSERTCMP(sp->call, ==, NULL);
530 	sp->call = call;
531 	rxrpc_get_call(call);
532 
533 	/* insert into the buffer in sequence order */
534 	spin_lock_bh(&call->lock);
535 
536 	skb_queue_walk(&call->rx_oos_queue, p) {
537 		psp = rxrpc_skb(p);
538 		if (ntohl(psp->hdr.seq) > seq) {
539 			_debug("insert oos #%u before #%u",
540 			       seq, ntohl(psp->hdr.seq));
541 			skb_insert(p, skb, &call->rx_oos_queue);
542 			goto inserted;
543 		}
544 	}
545 
546 	_debug("append oos #%u", seq);
547 	skb_queue_tail(&call->rx_oos_queue, skb);
548 inserted:
549 
550 	/* we might now have a new front to the queue */
551 	if (call->rx_first_oos == 0 || seq < call->rx_first_oos)
552 		call->rx_first_oos = seq;
553 
554 	read_lock(&call->state_lock);
555 	if (call->state < RXRPC_CALL_COMPLETE &&
556 	    call->rx_data_post == call->rx_first_oos) {
557 		_debug("drain rx oos now");
558 		set_bit(RXRPC_CALL_DRAIN_RX_OOS, &call->events);
559 	}
560 	read_unlock(&call->state_lock);
561 
562 	spin_unlock_bh(&call->lock);
563 	_leave(" [stored #%u]", call->rx_first_oos);
564 }
565 
566 /*
567  * clear the Tx window on final ACK reception
568  */
rxrpc_zap_tx_window(struct rxrpc_call * call)569 static void rxrpc_zap_tx_window(struct rxrpc_call *call)
570 {
571 	struct rxrpc_skb_priv *sp;
572 	struct sk_buff *skb;
573 	unsigned long _skb, *acks_window;
574 	u8 winsz = call->acks_winsz;
575 	int tail;
576 
577 	acks_window = call->acks_window;
578 	call->acks_window = NULL;
579 
580 	while (CIRC_CNT(call->acks_head, call->acks_tail, winsz) > 0) {
581 		tail = call->acks_tail;
582 		smp_read_barrier_depends();
583 		_skb = acks_window[tail] & ~1;
584 		smp_mb();
585 		call->acks_tail = (call->acks_tail + 1) & (winsz - 1);
586 
587 		skb = (struct sk_buff *) _skb;
588 		sp = rxrpc_skb(skb);
589 		_debug("+++ clear Tx %u", ntohl(sp->hdr.seq));
590 		rxrpc_free_skb(skb);
591 	}
592 
593 	kfree(acks_window);
594 }
595 
596 /*
597  * process the extra information that may be appended to an ACK packet
598  */
rxrpc_extract_ackinfo(struct rxrpc_call * call,struct sk_buff * skb,unsigned int latest,int nAcks)599 static void rxrpc_extract_ackinfo(struct rxrpc_call *call, struct sk_buff *skb,
600 				  unsigned int latest, int nAcks)
601 {
602 	struct rxrpc_ackinfo ackinfo;
603 	struct rxrpc_peer *peer;
604 	unsigned int mtu;
605 
606 	if (skb_copy_bits(skb, nAcks + 3, &ackinfo, sizeof(ackinfo)) < 0) {
607 		_leave(" [no ackinfo]");
608 		return;
609 	}
610 
611 	_proto("Rx ACK %%%u Info { rx=%u max=%u rwin=%u jm=%u }",
612 	       latest,
613 	       ntohl(ackinfo.rxMTU), ntohl(ackinfo.maxMTU),
614 	       ntohl(ackinfo.rwind), ntohl(ackinfo.jumbo_max));
615 
616 	mtu = min(ntohl(ackinfo.rxMTU), ntohl(ackinfo.maxMTU));
617 
618 	peer = call->conn->trans->peer;
619 	if (mtu < peer->maxdata) {
620 		spin_lock_bh(&peer->lock);
621 		peer->maxdata = mtu;
622 		peer->mtu = mtu + peer->hdrsize;
623 		spin_unlock_bh(&peer->lock);
624 		_net("Net MTU %u (maxdata %u)", peer->mtu, peer->maxdata);
625 	}
626 }
627 
628 /*
629  * process packets in the reception queue
630  */
rxrpc_process_rx_queue(struct rxrpc_call * call,u32 * _abort_code)631 static int rxrpc_process_rx_queue(struct rxrpc_call *call,
632 				  u32 *_abort_code)
633 {
634 	struct rxrpc_ackpacket ack;
635 	struct rxrpc_skb_priv *sp;
636 	struct sk_buff *skb;
637 	bool post_ACK;
638 	int latest;
639 	u32 hard, tx;
640 
641 	_enter("");
642 
643 process_further:
644 	skb = skb_dequeue(&call->rx_queue);
645 	if (!skb)
646 		return -EAGAIN;
647 
648 	_net("deferred skb %p", skb);
649 
650 	sp = rxrpc_skb(skb);
651 
652 	_debug("process %s [st %d]", rxrpc_pkts[sp->hdr.type], call->state);
653 
654 	post_ACK = false;
655 
656 	switch (sp->hdr.type) {
657 		/* data packets that wind up here have been received out of
658 		 * order, need security processing or are jumbo packets */
659 	case RXRPC_PACKET_TYPE_DATA:
660 		_proto("OOSQ DATA %%%u { #%u }",
661 		       ntohl(sp->hdr.serial), ntohl(sp->hdr.seq));
662 
663 		/* secured packets must be verified and possibly decrypted */
664 		if (rxrpc_verify_packet(call, skb, _abort_code) < 0)
665 			goto protocol_error;
666 
667 		rxrpc_insert_oos_packet(call, skb);
668 		goto process_further;
669 
670 		/* partial ACK to process */
671 	case RXRPC_PACKET_TYPE_ACK:
672 		if (skb_copy_bits(skb, 0, &ack, sizeof(ack)) < 0) {
673 			_debug("extraction failure");
674 			goto protocol_error;
675 		}
676 		if (!skb_pull(skb, sizeof(ack)))
677 			BUG();
678 
679 		latest = ntohl(sp->hdr.serial);
680 		hard = ntohl(ack.firstPacket);
681 		tx = atomic_read(&call->sequence);
682 
683 		_proto("Rx ACK %%%u { m=%hu f=#%u p=#%u s=%%%u r=%s n=%u }",
684 		       latest,
685 		       ntohs(ack.maxSkew),
686 		       hard,
687 		       ntohl(ack.previousPacket),
688 		       ntohl(ack.serial),
689 		       rxrpc_acks(ack.reason),
690 		       ack.nAcks);
691 
692 		rxrpc_extract_ackinfo(call, skb, latest, ack.nAcks);
693 
694 		if (ack.reason == RXRPC_ACK_PING) {
695 			_proto("Rx ACK %%%u PING Request", latest);
696 			rxrpc_propose_ACK(call, RXRPC_ACK_PING_RESPONSE,
697 					  sp->hdr.serial, true);
698 		}
699 
700 		/* discard any out-of-order or duplicate ACKs */
701 		if (latest - call->acks_latest <= 0) {
702 			_debug("discard ACK %d <= %d",
703 			       latest, call->acks_latest);
704 			goto discard;
705 		}
706 		call->acks_latest = latest;
707 
708 		if (call->state != RXRPC_CALL_CLIENT_SEND_REQUEST &&
709 		    call->state != RXRPC_CALL_CLIENT_AWAIT_REPLY &&
710 		    call->state != RXRPC_CALL_SERVER_SEND_REPLY &&
711 		    call->state != RXRPC_CALL_SERVER_AWAIT_ACK)
712 			goto discard;
713 
714 		_debug("Tx=%d H=%u S=%d", tx, call->acks_hard, call->state);
715 
716 		if (hard > 0) {
717 			if (hard - 1 > tx) {
718 				_debug("hard-ACK'd packet %d not transmitted"
719 				       " (%d top)",
720 				       hard - 1, tx);
721 				goto protocol_error;
722 			}
723 
724 			if ((call->state == RXRPC_CALL_CLIENT_AWAIT_REPLY ||
725 			     call->state == RXRPC_CALL_SERVER_AWAIT_ACK) &&
726 			    hard > tx)
727 				goto all_acked;
728 
729 			smp_rmb();
730 			rxrpc_rotate_tx_window(call, hard - 1);
731 		}
732 
733 		if (ack.nAcks > 0) {
734 			if (hard - 1 + ack.nAcks > tx) {
735 				_debug("soft-ACK'd packet %d+%d not"
736 				       " transmitted (%d top)",
737 				       hard - 1, ack.nAcks, tx);
738 				goto protocol_error;
739 			}
740 
741 			if (rxrpc_process_soft_ACKs(call, &ack, skb) < 0)
742 				goto protocol_error;
743 		}
744 		goto discard;
745 
746 		/* complete ACK to process */
747 	case RXRPC_PACKET_TYPE_ACKALL:
748 		goto all_acked;
749 
750 		/* abort and busy are handled elsewhere */
751 	case RXRPC_PACKET_TYPE_BUSY:
752 	case RXRPC_PACKET_TYPE_ABORT:
753 		BUG();
754 
755 		/* connection level events - also handled elsewhere */
756 	case RXRPC_PACKET_TYPE_CHALLENGE:
757 	case RXRPC_PACKET_TYPE_RESPONSE:
758 	case RXRPC_PACKET_TYPE_DEBUG:
759 		BUG();
760 	}
761 
762 	/* if we've had a hard ACK that covers all the packets we've sent, then
763 	 * that ends that phase of the operation */
764 all_acked:
765 	write_lock_bh(&call->state_lock);
766 	_debug("ack all %d", call->state);
767 
768 	switch (call->state) {
769 	case RXRPC_CALL_CLIENT_AWAIT_REPLY:
770 		call->state = RXRPC_CALL_CLIENT_RECV_REPLY;
771 		break;
772 	case RXRPC_CALL_SERVER_AWAIT_ACK:
773 		_debug("srv complete");
774 		call->state = RXRPC_CALL_COMPLETE;
775 		post_ACK = true;
776 		break;
777 	case RXRPC_CALL_CLIENT_SEND_REQUEST:
778 	case RXRPC_CALL_SERVER_RECV_REQUEST:
779 		goto protocol_error_unlock; /* can't occur yet */
780 	default:
781 		write_unlock_bh(&call->state_lock);
782 		goto discard; /* assume packet left over from earlier phase */
783 	}
784 
785 	write_unlock_bh(&call->state_lock);
786 
787 	/* if all the packets we sent are hard-ACK'd, then we can discard
788 	 * whatever we've got left */
789 	_debug("clear Tx %d",
790 	       CIRC_CNT(call->acks_head, call->acks_tail, call->acks_winsz));
791 
792 	del_timer_sync(&call->resend_timer);
793 	clear_bit(RXRPC_CALL_RUN_RTIMER, &call->flags);
794 	clear_bit(RXRPC_CALL_RESEND_TIMER, &call->events);
795 
796 	if (call->acks_window)
797 		rxrpc_zap_tx_window(call);
798 
799 	if (post_ACK) {
800 		/* post the final ACK message for userspace to pick up */
801 		_debug("post ACK");
802 		skb->mark = RXRPC_SKB_MARK_FINAL_ACK;
803 		sp->call = call;
804 		rxrpc_get_call(call);
805 		spin_lock_bh(&call->lock);
806 		if (rxrpc_queue_rcv_skb(call, skb, true, true) < 0)
807 			BUG();
808 		spin_unlock_bh(&call->lock);
809 		goto process_further;
810 	}
811 
812 discard:
813 	rxrpc_free_skb(skb);
814 	goto process_further;
815 
816 protocol_error_unlock:
817 	write_unlock_bh(&call->state_lock);
818 protocol_error:
819 	rxrpc_free_skb(skb);
820 	_leave(" = -EPROTO");
821 	return -EPROTO;
822 }
823 
824 /*
825  * post a message to the socket Rx queue for recvmsg() to pick up
826  */
rxrpc_post_message(struct rxrpc_call * call,u32 mark,u32 error,bool fatal)827 static int rxrpc_post_message(struct rxrpc_call *call, u32 mark, u32 error,
828 			      bool fatal)
829 {
830 	struct rxrpc_skb_priv *sp;
831 	struct sk_buff *skb;
832 	int ret;
833 
834 	_enter("{%d,%lx},%u,%u,%d",
835 	       call->debug_id, call->flags, mark, error, fatal);
836 
837 	/* remove timers and things for fatal messages */
838 	if (fatal) {
839 		del_timer_sync(&call->resend_timer);
840 		del_timer_sync(&call->ack_timer);
841 		clear_bit(RXRPC_CALL_RUN_RTIMER, &call->flags);
842 	}
843 
844 	if (mark != RXRPC_SKB_MARK_NEW_CALL &&
845 	    !test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
846 		_leave("[no userid]");
847 		return 0;
848 	}
849 
850 	if (!test_bit(RXRPC_CALL_TERMINAL_MSG, &call->flags)) {
851 		skb = alloc_skb(0, GFP_NOFS);
852 		if (!skb)
853 			return -ENOMEM;
854 
855 		rxrpc_new_skb(skb);
856 
857 		skb->mark = mark;
858 
859 		sp = rxrpc_skb(skb);
860 		memset(sp, 0, sizeof(*sp));
861 		sp->error = error;
862 		sp->call = call;
863 		rxrpc_get_call(call);
864 
865 		spin_lock_bh(&call->lock);
866 		ret = rxrpc_queue_rcv_skb(call, skb, true, fatal);
867 		spin_unlock_bh(&call->lock);
868 		BUG_ON(ret < 0);
869 	}
870 
871 	return 0;
872 }
873 
874 /*
875  * handle background processing of incoming call packets and ACK / abort
876  * generation
877  */
rxrpc_process_call(struct work_struct * work)878 void rxrpc_process_call(struct work_struct *work)
879 {
880 	struct rxrpc_call *call =
881 		container_of(work, struct rxrpc_call, processor);
882 	struct rxrpc_ackpacket ack;
883 	struct rxrpc_ackinfo ackinfo;
884 	struct rxrpc_header hdr;
885 	struct msghdr msg;
886 	struct kvec iov[5];
887 	unsigned long bits;
888 	__be32 data, pad;
889 	size_t len;
890 	int genbit, loop, nbit, ioc, ret, mtu;
891 	u32 abort_code = RX_PROTOCOL_ERROR;
892 	u8 *acks = NULL;
893 
894 	//printk("\n--------------------\n");
895 	_enter("{%d,%s,%lx} [%lu]",
896 	       call->debug_id, rxrpc_call_states[call->state], call->events,
897 	       (jiffies - call->creation_jif) / (HZ / 10));
898 
899 	if (test_and_set_bit(RXRPC_CALL_PROC_BUSY, &call->flags)) {
900 		_debug("XXXXXXXXXXXXX RUNNING ON MULTIPLE CPUS XXXXXXXXXXXXX");
901 		return;
902 	}
903 
904 	/* there's a good chance we're going to have to send a message, so set
905 	 * one up in advance */
906 	msg.msg_name	= &call->conn->trans->peer->srx.transport.sin;
907 	msg.msg_namelen	= sizeof(call->conn->trans->peer->srx.transport.sin);
908 	msg.msg_control	= NULL;
909 	msg.msg_controllen = 0;
910 	msg.msg_flags	= 0;
911 
912 	hdr.epoch	= call->conn->epoch;
913 	hdr.cid		= call->cid;
914 	hdr.callNumber	= call->call_id;
915 	hdr.seq		= 0;
916 	hdr.type	= RXRPC_PACKET_TYPE_ACK;
917 	hdr.flags	= call->conn->out_clientflag;
918 	hdr.userStatus	= 0;
919 	hdr.securityIndex = call->conn->security_ix;
920 	hdr._rsvd	= 0;
921 	hdr.serviceId	= call->conn->service_id;
922 
923 	memset(iov, 0, sizeof(iov));
924 	iov[0].iov_base	= &hdr;
925 	iov[0].iov_len	= sizeof(hdr);
926 
927 	/* deal with events of a final nature */
928 	if (test_bit(RXRPC_CALL_RELEASE, &call->events)) {
929 		rxrpc_release_call(call);
930 		clear_bit(RXRPC_CALL_RELEASE, &call->events);
931 	}
932 
933 	if (test_bit(RXRPC_CALL_RCVD_ERROR, &call->events)) {
934 		int error;
935 
936 		clear_bit(RXRPC_CALL_CONN_ABORT, &call->events);
937 		clear_bit(RXRPC_CALL_REJECT_BUSY, &call->events);
938 		clear_bit(RXRPC_CALL_ABORT, &call->events);
939 
940 		error = call->conn->trans->peer->net_error;
941 		_debug("post net error %d", error);
942 
943 		if (rxrpc_post_message(call, RXRPC_SKB_MARK_NET_ERROR,
944 				       error, true) < 0)
945 			goto no_mem;
946 		clear_bit(RXRPC_CALL_RCVD_ERROR, &call->events);
947 		goto kill_ACKs;
948 	}
949 
950 	if (test_bit(RXRPC_CALL_CONN_ABORT, &call->events)) {
951 		ASSERTCMP(call->state, >, RXRPC_CALL_COMPLETE);
952 
953 		clear_bit(RXRPC_CALL_REJECT_BUSY, &call->events);
954 		clear_bit(RXRPC_CALL_ABORT, &call->events);
955 
956 		_debug("post conn abort");
957 
958 		if (rxrpc_post_message(call, RXRPC_SKB_MARK_LOCAL_ERROR,
959 				       call->conn->error, true) < 0)
960 			goto no_mem;
961 		clear_bit(RXRPC_CALL_CONN_ABORT, &call->events);
962 		goto kill_ACKs;
963 	}
964 
965 	if (test_bit(RXRPC_CALL_REJECT_BUSY, &call->events)) {
966 		hdr.type = RXRPC_PACKET_TYPE_BUSY;
967 		genbit = RXRPC_CALL_REJECT_BUSY;
968 		goto send_message;
969 	}
970 
971 	if (test_bit(RXRPC_CALL_ABORT, &call->events)) {
972 		ASSERTCMP(call->state, >, RXRPC_CALL_COMPLETE);
973 
974 		if (rxrpc_post_message(call, RXRPC_SKB_MARK_LOCAL_ERROR,
975 				       ECONNABORTED, true) < 0)
976 			goto no_mem;
977 		hdr.type = RXRPC_PACKET_TYPE_ABORT;
978 		data = htonl(call->abort_code);
979 		iov[1].iov_base = &data;
980 		iov[1].iov_len = sizeof(data);
981 		genbit = RXRPC_CALL_ABORT;
982 		goto send_message;
983 	}
984 
985 	if (test_bit(RXRPC_CALL_ACK_FINAL, &call->events)) {
986 		genbit = RXRPC_CALL_ACK_FINAL;
987 
988 		ack.bufferSpace	= htons(8);
989 		ack.maxSkew	= 0;
990 		ack.serial	= 0;
991 		ack.reason	= RXRPC_ACK_IDLE;
992 		ack.nAcks	= 0;
993 		call->ackr_reason = 0;
994 
995 		spin_lock_bh(&call->lock);
996 		ack.serial = call->ackr_serial;
997 		ack.previousPacket = call->ackr_prev_seq;
998 		ack.firstPacket = htonl(call->rx_data_eaten + 1);
999 		spin_unlock_bh(&call->lock);
1000 
1001 		pad = 0;
1002 
1003 		iov[1].iov_base = &ack;
1004 		iov[1].iov_len	= sizeof(ack);
1005 		iov[2].iov_base = &pad;
1006 		iov[2].iov_len	= 3;
1007 		iov[3].iov_base = &ackinfo;
1008 		iov[3].iov_len	= sizeof(ackinfo);
1009 		goto send_ACK;
1010 	}
1011 
1012 	if (call->events & ((1 << RXRPC_CALL_RCVD_BUSY) |
1013 			    (1 << RXRPC_CALL_RCVD_ABORT))
1014 	    ) {
1015 		u32 mark;
1016 
1017 		if (test_bit(RXRPC_CALL_RCVD_ABORT, &call->events))
1018 			mark = RXRPC_SKB_MARK_REMOTE_ABORT;
1019 		else
1020 			mark = RXRPC_SKB_MARK_BUSY;
1021 
1022 		_debug("post abort/busy");
1023 		rxrpc_clear_tx_window(call);
1024 		if (rxrpc_post_message(call, mark, ECONNABORTED, true) < 0)
1025 			goto no_mem;
1026 
1027 		clear_bit(RXRPC_CALL_RCVD_BUSY, &call->events);
1028 		clear_bit(RXRPC_CALL_RCVD_ABORT, &call->events);
1029 		goto kill_ACKs;
1030 	}
1031 
1032 	if (test_and_clear_bit(RXRPC_CALL_RCVD_ACKALL, &call->events)) {
1033 		_debug("do implicit ackall");
1034 		rxrpc_clear_tx_window(call);
1035 	}
1036 
1037 	if (test_bit(RXRPC_CALL_LIFE_TIMER, &call->events)) {
1038 		write_lock_bh(&call->state_lock);
1039 		if (call->state <= RXRPC_CALL_COMPLETE) {
1040 			call->state = RXRPC_CALL_LOCALLY_ABORTED;
1041 			call->abort_code = RX_CALL_TIMEOUT;
1042 			set_bit(RXRPC_CALL_ABORT, &call->events);
1043 		}
1044 		write_unlock_bh(&call->state_lock);
1045 
1046 		_debug("post timeout");
1047 		if (rxrpc_post_message(call, RXRPC_SKB_MARK_LOCAL_ERROR,
1048 				       ETIME, true) < 0)
1049 			goto no_mem;
1050 
1051 		clear_bit(RXRPC_CALL_LIFE_TIMER, &call->events);
1052 		goto kill_ACKs;
1053 	}
1054 
1055 	/* deal with assorted inbound messages */
1056 	if (!skb_queue_empty(&call->rx_queue)) {
1057 		switch (rxrpc_process_rx_queue(call, &abort_code)) {
1058 		case 0:
1059 		case -EAGAIN:
1060 			break;
1061 		case -ENOMEM:
1062 			goto no_mem;
1063 		case -EKEYEXPIRED:
1064 		case -EKEYREJECTED:
1065 		case -EPROTO:
1066 			rxrpc_abort_call(call, abort_code);
1067 			goto kill_ACKs;
1068 		}
1069 	}
1070 
1071 	/* handle resending */
1072 	if (test_and_clear_bit(RXRPC_CALL_RESEND_TIMER, &call->events))
1073 		rxrpc_resend_timer(call);
1074 	if (test_and_clear_bit(RXRPC_CALL_RESEND, &call->events))
1075 		rxrpc_resend(call);
1076 
1077 	/* consider sending an ordinary ACK */
1078 	if (test_bit(RXRPC_CALL_ACK, &call->events)) {
1079 		_debug("send ACK: window: %d - %d { %lx }",
1080 		       call->rx_data_eaten, call->ackr_win_top,
1081 		       call->ackr_window[0]);
1082 
1083 		if (call->state > RXRPC_CALL_SERVER_ACK_REQUEST &&
1084 		    call->ackr_reason != RXRPC_ACK_PING_RESPONSE) {
1085 			/* ACK by sending reply DATA packet in this state */
1086 			clear_bit(RXRPC_CALL_ACK, &call->events);
1087 			goto maybe_reschedule;
1088 		}
1089 
1090 		genbit = RXRPC_CALL_ACK;
1091 
1092 		acks = kzalloc(call->ackr_win_top - call->rx_data_eaten,
1093 			       GFP_NOFS);
1094 		if (!acks)
1095 			goto no_mem;
1096 
1097 		//hdr.flags	= RXRPC_SLOW_START_OK;
1098 		ack.bufferSpace	= htons(8);
1099 		ack.maxSkew	= 0;
1100 		ack.serial	= 0;
1101 		ack.reason	= 0;
1102 
1103 		spin_lock_bh(&call->lock);
1104 		ack.reason = call->ackr_reason;
1105 		ack.serial = call->ackr_serial;
1106 		ack.previousPacket = call->ackr_prev_seq;
1107 		ack.firstPacket = htonl(call->rx_data_eaten + 1);
1108 
1109 		ack.nAcks = 0;
1110 		for (loop = 0; loop < RXRPC_ACKR_WINDOW_ASZ; loop++) {
1111 			nbit = loop * BITS_PER_LONG;
1112 			for (bits = call->ackr_window[loop]; bits; bits >>= 1
1113 			     ) {
1114 				_debug("- l=%d n=%d b=%lx", loop, nbit, bits);
1115 				if (bits & 1) {
1116 					acks[nbit] = RXRPC_ACK_TYPE_ACK;
1117 					ack.nAcks = nbit + 1;
1118 				}
1119 				nbit++;
1120 			}
1121 		}
1122 		call->ackr_reason = 0;
1123 		spin_unlock_bh(&call->lock);
1124 
1125 		pad = 0;
1126 
1127 		iov[1].iov_base = &ack;
1128 		iov[1].iov_len	= sizeof(ack);
1129 		iov[2].iov_base = acks;
1130 		iov[2].iov_len	= ack.nAcks;
1131 		iov[3].iov_base = &pad;
1132 		iov[3].iov_len	= 3;
1133 		iov[4].iov_base = &ackinfo;
1134 		iov[4].iov_len	= sizeof(ackinfo);
1135 
1136 		switch (ack.reason) {
1137 		case RXRPC_ACK_REQUESTED:
1138 		case RXRPC_ACK_DUPLICATE:
1139 		case RXRPC_ACK_OUT_OF_SEQUENCE:
1140 		case RXRPC_ACK_EXCEEDS_WINDOW:
1141 		case RXRPC_ACK_NOSPACE:
1142 		case RXRPC_ACK_PING:
1143 		case RXRPC_ACK_PING_RESPONSE:
1144 			goto send_ACK_with_skew;
1145 		case RXRPC_ACK_DELAY:
1146 		case RXRPC_ACK_IDLE:
1147 			goto send_ACK;
1148 		}
1149 	}
1150 
1151 	/* handle completion of security negotiations on an incoming
1152 	 * connection */
1153 	if (test_and_clear_bit(RXRPC_CALL_SECURED, &call->events)) {
1154 		_debug("secured");
1155 		spin_lock_bh(&call->lock);
1156 
1157 		if (call->state == RXRPC_CALL_SERVER_SECURING) {
1158 			_debug("securing");
1159 			write_lock(&call->conn->lock);
1160 			if (!test_bit(RXRPC_CALL_RELEASED, &call->flags) &&
1161 			    !test_bit(RXRPC_CALL_RELEASE, &call->events)) {
1162 				_debug("not released");
1163 				call->state = RXRPC_CALL_SERVER_ACCEPTING;
1164 				list_move_tail(&call->accept_link,
1165 					       &call->socket->acceptq);
1166 			}
1167 			write_unlock(&call->conn->lock);
1168 			read_lock(&call->state_lock);
1169 			if (call->state < RXRPC_CALL_COMPLETE)
1170 				set_bit(RXRPC_CALL_POST_ACCEPT, &call->events);
1171 			read_unlock(&call->state_lock);
1172 		}
1173 
1174 		spin_unlock_bh(&call->lock);
1175 		if (!test_bit(RXRPC_CALL_POST_ACCEPT, &call->events))
1176 			goto maybe_reschedule;
1177 	}
1178 
1179 	/* post a notification of an acceptable connection to the app */
1180 	if (test_bit(RXRPC_CALL_POST_ACCEPT, &call->events)) {
1181 		_debug("post accept");
1182 		if (rxrpc_post_message(call, RXRPC_SKB_MARK_NEW_CALL,
1183 				       0, false) < 0)
1184 			goto no_mem;
1185 		clear_bit(RXRPC_CALL_POST_ACCEPT, &call->events);
1186 		goto maybe_reschedule;
1187 	}
1188 
1189 	/* handle incoming call acceptance */
1190 	if (test_and_clear_bit(RXRPC_CALL_ACCEPTED, &call->events)) {
1191 		_debug("accepted");
1192 		ASSERTCMP(call->rx_data_post, ==, 0);
1193 		call->rx_data_post = 1;
1194 		read_lock_bh(&call->state_lock);
1195 		if (call->state < RXRPC_CALL_COMPLETE)
1196 			set_bit(RXRPC_CALL_DRAIN_RX_OOS, &call->events);
1197 		read_unlock_bh(&call->state_lock);
1198 	}
1199 
1200 	/* drain the out of sequence received packet queue into the packet Rx
1201 	 * queue */
1202 	if (test_and_clear_bit(RXRPC_CALL_DRAIN_RX_OOS, &call->events)) {
1203 		while (call->rx_data_post == call->rx_first_oos)
1204 			if (rxrpc_drain_rx_oos_queue(call) < 0)
1205 				break;
1206 		goto maybe_reschedule;
1207 	}
1208 
1209 	/* other events may have been raised since we started checking */
1210 	goto maybe_reschedule;
1211 
1212 send_ACK_with_skew:
1213 	ack.maxSkew = htons(atomic_read(&call->conn->hi_serial) -
1214 			    ntohl(ack.serial));
1215 send_ACK:
1216 	mtu = call->conn->trans->peer->if_mtu;
1217 	mtu -= call->conn->trans->peer->hdrsize;
1218 	ackinfo.maxMTU	= htonl(mtu);
1219 	ackinfo.rwind	= htonl(rxrpc_rx_window_size);
1220 
1221 	/* permit the peer to send us jumbo packets if it wants to */
1222 	ackinfo.rxMTU	= htonl(rxrpc_rx_mtu);
1223 	ackinfo.jumbo_max = htonl(rxrpc_rx_jumbo_max);
1224 
1225 	hdr.serial = htonl(atomic_inc_return(&call->conn->serial));
1226 	_proto("Tx ACK %%%u { m=%hu f=#%u p=#%u s=%%%u r=%s n=%u }",
1227 	       ntohl(hdr.serial),
1228 	       ntohs(ack.maxSkew),
1229 	       ntohl(ack.firstPacket),
1230 	       ntohl(ack.previousPacket),
1231 	       ntohl(ack.serial),
1232 	       rxrpc_acks(ack.reason),
1233 	       ack.nAcks);
1234 
1235 	del_timer_sync(&call->ack_timer);
1236 	if (ack.nAcks > 0)
1237 		set_bit(RXRPC_CALL_TX_SOFT_ACK, &call->flags);
1238 	goto send_message_2;
1239 
1240 send_message:
1241 	_debug("send message");
1242 
1243 	hdr.serial = htonl(atomic_inc_return(&call->conn->serial));
1244 	_proto("Tx %s %%%u", rxrpc_pkts[hdr.type], ntohl(hdr.serial));
1245 send_message_2:
1246 
1247 	len = iov[0].iov_len;
1248 	ioc = 1;
1249 	if (iov[4].iov_len) {
1250 		ioc = 5;
1251 		len += iov[4].iov_len;
1252 		len += iov[3].iov_len;
1253 		len += iov[2].iov_len;
1254 		len += iov[1].iov_len;
1255 	} else if (iov[3].iov_len) {
1256 		ioc = 4;
1257 		len += iov[3].iov_len;
1258 		len += iov[2].iov_len;
1259 		len += iov[1].iov_len;
1260 	} else if (iov[2].iov_len) {
1261 		ioc = 3;
1262 		len += iov[2].iov_len;
1263 		len += iov[1].iov_len;
1264 	} else if (iov[1].iov_len) {
1265 		ioc = 2;
1266 		len += iov[1].iov_len;
1267 	}
1268 
1269 	ret = kernel_sendmsg(call->conn->trans->local->socket,
1270 			     &msg, iov, ioc, len);
1271 	if (ret < 0) {
1272 		_debug("sendmsg failed: %d", ret);
1273 		read_lock_bh(&call->state_lock);
1274 		if (call->state < RXRPC_CALL_DEAD)
1275 			rxrpc_queue_call(call);
1276 		read_unlock_bh(&call->state_lock);
1277 		goto error;
1278 	}
1279 
1280 	switch (genbit) {
1281 	case RXRPC_CALL_ABORT:
1282 		clear_bit(genbit, &call->events);
1283 		clear_bit(RXRPC_CALL_RCVD_ABORT, &call->events);
1284 		goto kill_ACKs;
1285 
1286 	case RXRPC_CALL_ACK_FINAL:
1287 		write_lock_bh(&call->state_lock);
1288 		if (call->state == RXRPC_CALL_CLIENT_FINAL_ACK)
1289 			call->state = RXRPC_CALL_COMPLETE;
1290 		write_unlock_bh(&call->state_lock);
1291 		goto kill_ACKs;
1292 
1293 	default:
1294 		clear_bit(genbit, &call->events);
1295 		switch (call->state) {
1296 		case RXRPC_CALL_CLIENT_AWAIT_REPLY:
1297 		case RXRPC_CALL_CLIENT_RECV_REPLY:
1298 		case RXRPC_CALL_SERVER_RECV_REQUEST:
1299 		case RXRPC_CALL_SERVER_ACK_REQUEST:
1300 			_debug("start ACK timer");
1301 			rxrpc_propose_ACK(call, RXRPC_ACK_DELAY,
1302 					  call->ackr_serial, false);
1303 		default:
1304 			break;
1305 		}
1306 		goto maybe_reschedule;
1307 	}
1308 
1309 kill_ACKs:
1310 	del_timer_sync(&call->ack_timer);
1311 	if (test_and_clear_bit(RXRPC_CALL_ACK_FINAL, &call->events))
1312 		rxrpc_put_call(call);
1313 	clear_bit(RXRPC_CALL_ACK, &call->events);
1314 
1315 maybe_reschedule:
1316 	if (call->events || !skb_queue_empty(&call->rx_queue)) {
1317 		read_lock_bh(&call->state_lock);
1318 		if (call->state < RXRPC_CALL_DEAD)
1319 			rxrpc_queue_call(call);
1320 		read_unlock_bh(&call->state_lock);
1321 	}
1322 
1323 	/* don't leave aborted connections on the accept queue */
1324 	if (call->state >= RXRPC_CALL_COMPLETE &&
1325 	    !list_empty(&call->accept_link)) {
1326 		_debug("X unlinking once-pending call %p { e=%lx f=%lx c=%x }",
1327 		       call, call->events, call->flags,
1328 		       ntohl(call->conn->cid));
1329 
1330 		read_lock_bh(&call->state_lock);
1331 		if (!test_bit(RXRPC_CALL_RELEASED, &call->flags) &&
1332 		    !test_and_set_bit(RXRPC_CALL_RELEASE, &call->events))
1333 			rxrpc_queue_call(call);
1334 		read_unlock_bh(&call->state_lock);
1335 	}
1336 
1337 error:
1338 	clear_bit(RXRPC_CALL_PROC_BUSY, &call->flags);
1339 	kfree(acks);
1340 
1341 	/* because we don't want two CPUs both processing the work item for one
1342 	 * call at the same time, we use a flag to note when it's busy; however
1343 	 * this means there's a race between clearing the flag and setting the
1344 	 * work pending bit and the work item being processed again */
1345 	if (call->events && !work_pending(&call->processor)) {
1346 		_debug("jumpstart %x", ntohl(call->conn->cid));
1347 		rxrpc_queue_call(call);
1348 	}
1349 
1350 	_leave("");
1351 	return;
1352 
1353 no_mem:
1354 	_debug("out of memory");
1355 	goto maybe_reschedule;
1356 }
1357