xref: /linux/net/rxrpc/recvmsg.c (revision e5d3a64e650c721f9e9b1f76e5df8c62f16b734d)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* RxRPC recvmsg() implementation
3  *
4  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
5  * Written by David Howells (dhowells@redhat.com)
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/net.h>
11 #include <linux/skbuff.h>
12 #include <linux/export.h>
13 #include <linux/sched/signal.h>
14 
15 #include <net/sock.h>
16 #include <net/af_rxrpc.h>
17 #include "ar-internal.h"
18 
19 /*
20  * Post a call for attention by the socket or kernel service.  Further
21  * notifications are suppressed by putting recvmsg_link on a dummy queue.
22  */
23 void rxrpc_notify_socket(struct rxrpc_call *call)
24 {
25 	struct rxrpc_sock *rx;
26 	struct sock *sk;
27 
28 	_enter("%d", call->debug_id);
29 
30 	if (!list_empty(&call->recvmsg_link))
31 		return;
32 
33 	rcu_read_lock();
34 
35 	rx = rcu_dereference(call->socket);
36 	sk = &rx->sk;
37 	if (rx && sk->sk_state < RXRPC_CLOSE) {
38 		if (call->notify_rx) {
39 			spin_lock_bh(&call->notify_lock);
40 			call->notify_rx(sk, call, call->user_call_ID);
41 			spin_unlock_bh(&call->notify_lock);
42 		} else {
43 			write_lock_bh(&rx->recvmsg_lock);
44 			if (list_empty(&call->recvmsg_link)) {
45 				rxrpc_get_call(call, rxrpc_call_got);
46 				list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
47 			}
48 			write_unlock_bh(&rx->recvmsg_lock);
49 
50 			if (!sock_flag(sk, SOCK_DEAD)) {
51 				_debug("call %ps", sk->sk_data_ready);
52 				sk->sk_data_ready(sk);
53 			}
54 		}
55 	}
56 
57 	rcu_read_unlock();
58 	_leave("");
59 }
60 
61 /*
62  * Transition a call to the complete state.
63  */
64 bool __rxrpc_set_call_completion(struct rxrpc_call *call,
65 				 enum rxrpc_call_completion compl,
66 				 u32 abort_code,
67 				 int error)
68 {
69 	if (call->state < RXRPC_CALL_COMPLETE) {
70 		call->abort_code = abort_code;
71 		call->error = error;
72 		call->completion = compl;
73 		call->state = RXRPC_CALL_COMPLETE;
74 		trace_rxrpc_call_complete(call);
75 		wake_up(&call->waitq);
76 		rxrpc_notify_socket(call);
77 		return true;
78 	}
79 	return false;
80 }
81 
82 bool rxrpc_set_call_completion(struct rxrpc_call *call,
83 			       enum rxrpc_call_completion compl,
84 			       u32 abort_code,
85 			       int error)
86 {
87 	bool ret = false;
88 
89 	if (call->state < RXRPC_CALL_COMPLETE) {
90 		write_lock_bh(&call->state_lock);
91 		ret = __rxrpc_set_call_completion(call, compl, abort_code, error);
92 		write_unlock_bh(&call->state_lock);
93 	}
94 	return ret;
95 }
96 
97 /*
98  * Record that a call successfully completed.
99  */
100 bool __rxrpc_call_completed(struct rxrpc_call *call)
101 {
102 	return __rxrpc_set_call_completion(call, RXRPC_CALL_SUCCEEDED, 0, 0);
103 }
104 
105 bool rxrpc_call_completed(struct rxrpc_call *call)
106 {
107 	bool ret = false;
108 
109 	if (call->state < RXRPC_CALL_COMPLETE) {
110 		write_lock_bh(&call->state_lock);
111 		ret = __rxrpc_call_completed(call);
112 		write_unlock_bh(&call->state_lock);
113 	}
114 	return ret;
115 }
116 
117 /*
118  * Record that a call is locally aborted.
119  */
120 bool __rxrpc_abort_call(const char *why, struct rxrpc_call *call,
121 			rxrpc_seq_t seq, u32 abort_code, int error)
122 {
123 	trace_rxrpc_abort(call->debug_id, why, call->cid, call->call_id, seq,
124 			  abort_code, error);
125 	return __rxrpc_set_call_completion(call, RXRPC_CALL_LOCALLY_ABORTED,
126 					   abort_code, error);
127 }
128 
129 bool rxrpc_abort_call(const char *why, struct rxrpc_call *call,
130 		      rxrpc_seq_t seq, u32 abort_code, int error)
131 {
132 	bool ret;
133 
134 	write_lock_bh(&call->state_lock);
135 	ret = __rxrpc_abort_call(why, call, seq, abort_code, error);
136 	write_unlock_bh(&call->state_lock);
137 	return ret;
138 }
139 
140 /*
141  * Pass a call terminating message to userspace.
142  */
143 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
144 {
145 	u32 tmp = 0;
146 	int ret;
147 
148 	switch (call->completion) {
149 	case RXRPC_CALL_SUCCEEDED:
150 		ret = 0;
151 		if (rxrpc_is_service_call(call))
152 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
153 		break;
154 	case RXRPC_CALL_REMOTELY_ABORTED:
155 		tmp = call->abort_code;
156 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
157 		break;
158 	case RXRPC_CALL_LOCALLY_ABORTED:
159 		tmp = call->abort_code;
160 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
161 		break;
162 	case RXRPC_CALL_NETWORK_ERROR:
163 		tmp = -call->error;
164 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
165 		break;
166 	case RXRPC_CALL_LOCAL_ERROR:
167 		tmp = -call->error;
168 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
169 		break;
170 	default:
171 		pr_err("Invalid terminal call state %u\n", call->state);
172 		BUG();
173 		break;
174 	}
175 
176 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
177 			    call->rx_pkt_offset, call->rx_pkt_len, ret);
178 	return ret;
179 }
180 
181 /*
182  * End the packet reception phase.
183  */
184 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
185 {
186 	_enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
187 
188 	trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
189 	ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
190 
191 	if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
192 		rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, serial, false, true,
193 				  rxrpc_propose_ack_terminal_ack);
194 		//rxrpc_send_ack_packet(call, false, NULL);
195 	}
196 
197 	write_lock_bh(&call->state_lock);
198 
199 	switch (call->state) {
200 	case RXRPC_CALL_CLIENT_RECV_REPLY:
201 		__rxrpc_call_completed(call);
202 		write_unlock_bh(&call->state_lock);
203 		break;
204 
205 	case RXRPC_CALL_SERVER_RECV_REQUEST:
206 		call->tx_phase = true;
207 		call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
208 		call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
209 		write_unlock_bh(&call->state_lock);
210 		rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial, false, true,
211 				  rxrpc_propose_ack_processing_op);
212 		break;
213 	default:
214 		write_unlock_bh(&call->state_lock);
215 		break;
216 	}
217 }
218 
219 /*
220  * Discard a packet we've used up and advance the Rx window by one.
221  */
222 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
223 {
224 	struct rxrpc_skb_priv *sp;
225 	struct sk_buff *skb;
226 	rxrpc_serial_t serial;
227 	rxrpc_seq_t hard_ack, top;
228 	bool last = false;
229 	u8 subpacket;
230 	int ix;
231 
232 	_enter("%d", call->debug_id);
233 
234 	hard_ack = call->rx_hard_ack;
235 	top = smp_load_acquire(&call->rx_top);
236 	ASSERT(before(hard_ack, top));
237 
238 	hard_ack++;
239 	ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
240 	skb = call->rxtx_buffer[ix];
241 	rxrpc_see_skb(skb, rxrpc_skb_rotated);
242 	sp = rxrpc_skb(skb);
243 
244 	subpacket = call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
245 	serial = sp->hdr.serial + subpacket;
246 
247 	if (subpacket == sp->nr_subpackets - 1 &&
248 	    sp->rx_flags & RXRPC_SKB_INCL_LAST)
249 		last = true;
250 
251 	call->rxtx_buffer[ix] = NULL;
252 	call->rxtx_annotations[ix] = 0;
253 	/* Barrier against rxrpc_input_data(). */
254 	smp_store_release(&call->rx_hard_ack, hard_ack);
255 
256 	rxrpc_free_skb(skb, rxrpc_skb_freed);
257 
258 	trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
259 	if (last) {
260 		rxrpc_end_rx_phase(call, serial);
261 	} else {
262 		/* Check to see if there's an ACK that needs sending. */
263 		if (atomic_inc_return(&call->ackr_nr_consumed) > 2)
264 			rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, serial,
265 					  true, false,
266 					  rxrpc_propose_ack_rotate_rx);
267 		if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
268 			rxrpc_send_ack_packet(call, false, NULL);
269 	}
270 }
271 
272 /*
273  * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
274  * padding, but if this is the case, the packet length will be resident in the
275  * socket buffer.  Note that we can't modify the master skb info as the skb may
276  * be the home to multiple subpackets.
277  */
278 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
279 			       u8 annotation,
280 			       unsigned int offset, unsigned int len)
281 {
282 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
283 	rxrpc_seq_t seq = sp->hdr.seq;
284 	u16 cksum = sp->hdr.cksum;
285 	u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
286 
287 	_enter("");
288 
289 	/* For all but the head jumbo subpacket, the security checksum is in a
290 	 * jumbo header immediately prior to the data.
291 	 */
292 	if (subpacket > 0) {
293 		__be16 tmp;
294 		if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
295 			BUG();
296 		cksum = ntohs(tmp);
297 		seq += subpacket;
298 	}
299 
300 	return call->security->verify_packet(call, skb, offset, len,
301 					     seq, cksum);
302 }
303 
304 /*
305  * Locate the data within a packet.  This is complicated by:
306  *
307  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
308  *     subpacket.
309  *
310  * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
311  *     contains an extra header which includes the true length of the data,
312  *     excluding any encrypted padding.
313  */
314 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
315 			     u8 *_annotation,
316 			     unsigned int *_offset, unsigned int *_len,
317 			     bool *_last)
318 {
319 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
320 	unsigned int offset = sizeof(struct rxrpc_wire_header);
321 	unsigned int len;
322 	bool last = false;
323 	int ret;
324 	u8 annotation = *_annotation;
325 	u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
326 
327 	/* Locate the subpacket */
328 	offset += subpacket * RXRPC_JUMBO_SUBPKTLEN;
329 	len = skb->len - offset;
330 	if (subpacket < sp->nr_subpackets - 1)
331 		len = RXRPC_JUMBO_DATALEN;
332 	else if (sp->rx_flags & RXRPC_SKB_INCL_LAST)
333 		last = true;
334 
335 	if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
336 		ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
337 		if (ret < 0)
338 			return ret;
339 		*_annotation |= RXRPC_RX_ANNO_VERIFIED;
340 	}
341 
342 	*_offset = offset;
343 	*_len = len;
344 	*_last = last;
345 	call->security->locate_data(call, skb, _offset, _len);
346 	return 0;
347 }
348 
349 /*
350  * Deliver messages to a call.  This keeps processing packets until the buffer
351  * is filled and we find either more DATA (returns 0) or the end of the DATA
352  * (returns 1).  If more packets are required, it returns -EAGAIN.
353  */
354 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
355 			      struct msghdr *msg, struct iov_iter *iter,
356 			      size_t len, int flags, size_t *_offset)
357 {
358 	struct rxrpc_skb_priv *sp;
359 	struct sk_buff *skb;
360 	rxrpc_serial_t serial;
361 	rxrpc_seq_t hard_ack, top, seq;
362 	size_t remain;
363 	bool rx_pkt_last;
364 	unsigned int rx_pkt_offset, rx_pkt_len;
365 	int ix, copy, ret = -EAGAIN, ret2;
366 
367 	if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
368 	    call->ackr_reason)
369 		rxrpc_send_ack_packet(call, false, NULL);
370 
371 	rx_pkt_offset = call->rx_pkt_offset;
372 	rx_pkt_len = call->rx_pkt_len;
373 	rx_pkt_last = call->rx_pkt_last;
374 
375 	if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
376 		seq = call->rx_hard_ack;
377 		ret = 1;
378 		goto done;
379 	}
380 
381 	/* Barriers against rxrpc_input_data(). */
382 	hard_ack = call->rx_hard_ack;
383 	seq = hard_ack + 1;
384 
385 	while (top = smp_load_acquire(&call->rx_top),
386 	       before_eq(seq, top)
387 	       ) {
388 		ix = seq & RXRPC_RXTX_BUFF_MASK;
389 		skb = call->rxtx_buffer[ix];
390 		if (!skb) {
391 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
392 					    rx_pkt_offset, rx_pkt_len, 0);
393 			break;
394 		}
395 		smp_rmb();
396 		rxrpc_see_skb(skb, rxrpc_skb_seen);
397 		sp = rxrpc_skb(skb);
398 
399 		if (!(flags & MSG_PEEK)) {
400 			serial = sp->hdr.serial;
401 			serial += call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
402 			trace_rxrpc_receive(call, rxrpc_receive_front,
403 					    serial, seq);
404 		}
405 
406 		if (msg)
407 			sock_recv_timestamp(msg, sock->sk, skb);
408 
409 		if (rx_pkt_offset == 0) {
410 			ret2 = rxrpc_locate_data(call, skb,
411 						 &call->rxtx_annotations[ix],
412 						 &rx_pkt_offset, &rx_pkt_len,
413 						 &rx_pkt_last);
414 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
415 					    rx_pkt_offset, rx_pkt_len, ret2);
416 			if (ret2 < 0) {
417 				ret = ret2;
418 				goto out;
419 			}
420 		} else {
421 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
422 					    rx_pkt_offset, rx_pkt_len, 0);
423 		}
424 
425 		/* We have to handle short, empty and used-up DATA packets. */
426 		remain = len - *_offset;
427 		copy = rx_pkt_len;
428 		if (copy > remain)
429 			copy = remain;
430 		if (copy > 0) {
431 			ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
432 						      copy);
433 			if (ret2 < 0) {
434 				ret = ret2;
435 				goto out;
436 			}
437 
438 			/* handle piecemeal consumption of data packets */
439 			rx_pkt_offset += copy;
440 			rx_pkt_len -= copy;
441 			*_offset += copy;
442 		}
443 
444 		if (rx_pkt_len > 0) {
445 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
446 					    rx_pkt_offset, rx_pkt_len, 0);
447 			ASSERTCMP(*_offset, ==, len);
448 			ret = 0;
449 			break;
450 		}
451 
452 		/* The whole packet has been transferred. */
453 		if (!(flags & MSG_PEEK))
454 			rxrpc_rotate_rx_window(call);
455 		rx_pkt_offset = 0;
456 		rx_pkt_len = 0;
457 
458 		if (rx_pkt_last) {
459 			ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
460 			ret = 1;
461 			goto out;
462 		}
463 
464 		seq++;
465 	}
466 
467 out:
468 	if (!(flags & MSG_PEEK)) {
469 		call->rx_pkt_offset = rx_pkt_offset;
470 		call->rx_pkt_len = rx_pkt_len;
471 		call->rx_pkt_last = rx_pkt_last;
472 	}
473 done:
474 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
475 			    rx_pkt_offset, rx_pkt_len, ret);
476 	if (ret == -EAGAIN)
477 		set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
478 	return ret;
479 }
480 
481 /*
482  * Receive a message from an RxRPC socket
483  * - we need to be careful about two or more threads calling recvmsg
484  *   simultaneously
485  */
486 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
487 		  int flags)
488 {
489 	struct rxrpc_call *call;
490 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
491 	struct list_head *l;
492 	size_t copied = 0;
493 	long timeo;
494 	int ret;
495 
496 	DEFINE_WAIT(wait);
497 
498 	trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
499 
500 	if (flags & (MSG_OOB | MSG_TRUNC))
501 		return -EOPNOTSUPP;
502 
503 	timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
504 
505 try_again:
506 	lock_sock(&rx->sk);
507 
508 	/* Return immediately if a client socket has no outstanding calls */
509 	if (RB_EMPTY_ROOT(&rx->calls) &&
510 	    list_empty(&rx->recvmsg_q) &&
511 	    rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
512 		release_sock(&rx->sk);
513 		return -EAGAIN;
514 	}
515 
516 	if (list_empty(&rx->recvmsg_q)) {
517 		ret = -EWOULDBLOCK;
518 		if (timeo == 0) {
519 			call = NULL;
520 			goto error_no_call;
521 		}
522 
523 		release_sock(&rx->sk);
524 
525 		/* Wait for something to happen */
526 		prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
527 					  TASK_INTERRUPTIBLE);
528 		ret = sock_error(&rx->sk);
529 		if (ret)
530 			goto wait_error;
531 
532 		if (list_empty(&rx->recvmsg_q)) {
533 			if (signal_pending(current))
534 				goto wait_interrupted;
535 			trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
536 					    0, 0, 0, 0);
537 			timeo = schedule_timeout(timeo);
538 		}
539 		finish_wait(sk_sleep(&rx->sk), &wait);
540 		goto try_again;
541 	}
542 
543 	/* Find the next call and dequeue it if we're not just peeking.  If we
544 	 * do dequeue it, that comes with a ref that we will need to release.
545 	 */
546 	write_lock_bh(&rx->recvmsg_lock);
547 	l = rx->recvmsg_q.next;
548 	call = list_entry(l, struct rxrpc_call, recvmsg_link);
549 	if (!(flags & MSG_PEEK))
550 		list_del_init(&call->recvmsg_link);
551 	else
552 		rxrpc_get_call(call, rxrpc_call_got);
553 	write_unlock_bh(&rx->recvmsg_lock);
554 
555 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
556 
557 	/* We're going to drop the socket lock, so we need to lock the call
558 	 * against interference by sendmsg.
559 	 */
560 	if (!mutex_trylock(&call->user_mutex)) {
561 		ret = -EWOULDBLOCK;
562 		if (flags & MSG_DONTWAIT)
563 			goto error_requeue_call;
564 		ret = -ERESTARTSYS;
565 		if (mutex_lock_interruptible(&call->user_mutex) < 0)
566 			goto error_requeue_call;
567 	}
568 
569 	release_sock(&rx->sk);
570 
571 	if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
572 		BUG();
573 
574 	if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
575 		if (flags & MSG_CMSG_COMPAT) {
576 			unsigned int id32 = call->user_call_ID;
577 
578 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
579 				       sizeof(unsigned int), &id32);
580 		} else {
581 			unsigned long idl = call->user_call_ID;
582 
583 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
584 				       sizeof(unsigned long), &idl);
585 		}
586 		if (ret < 0)
587 			goto error_unlock_call;
588 	}
589 
590 	if (msg->msg_name && call->peer) {
591 		struct sockaddr_rxrpc *srx = msg->msg_name;
592 		size_t len = sizeof(call->peer->srx);
593 
594 		memcpy(msg->msg_name, &call->peer->srx, len);
595 		srx->srx_service = call->service_id;
596 		msg->msg_namelen = len;
597 	}
598 
599 	switch (READ_ONCE(call->state)) {
600 	case RXRPC_CALL_CLIENT_RECV_REPLY:
601 	case RXRPC_CALL_SERVER_RECV_REQUEST:
602 	case RXRPC_CALL_SERVER_ACK_REQUEST:
603 		ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
604 					 flags, &copied);
605 		if (ret == -EAGAIN)
606 			ret = 0;
607 
608 		if (after(call->rx_top, call->rx_hard_ack) &&
609 		    call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
610 			rxrpc_notify_socket(call);
611 		break;
612 	default:
613 		ret = 0;
614 		break;
615 	}
616 
617 	if (ret < 0)
618 		goto error_unlock_call;
619 
620 	if (call->state == RXRPC_CALL_COMPLETE) {
621 		ret = rxrpc_recvmsg_term(call, msg);
622 		if (ret < 0)
623 			goto error_unlock_call;
624 		if (!(flags & MSG_PEEK))
625 			rxrpc_release_call(rx, call);
626 		msg->msg_flags |= MSG_EOR;
627 		ret = 1;
628 	}
629 
630 	if (ret == 0)
631 		msg->msg_flags |= MSG_MORE;
632 	else
633 		msg->msg_flags &= ~MSG_MORE;
634 	ret = copied;
635 
636 error_unlock_call:
637 	mutex_unlock(&call->user_mutex);
638 	rxrpc_put_call(call, rxrpc_call_put);
639 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
640 	return ret;
641 
642 error_requeue_call:
643 	if (!(flags & MSG_PEEK)) {
644 		write_lock_bh(&rx->recvmsg_lock);
645 		list_add(&call->recvmsg_link, &rx->recvmsg_q);
646 		write_unlock_bh(&rx->recvmsg_lock);
647 		trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
648 	} else {
649 		rxrpc_put_call(call, rxrpc_call_put);
650 	}
651 error_no_call:
652 	release_sock(&rx->sk);
653 error_trace:
654 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
655 	return ret;
656 
657 wait_interrupted:
658 	ret = sock_intr_errno(timeo);
659 wait_error:
660 	finish_wait(sk_sleep(&rx->sk), &wait);
661 	call = NULL;
662 	goto error_trace;
663 }
664 
665 /**
666  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
667  * @sock: The socket that the call exists on
668  * @call: The call to send data through
669  * @iter: The buffer to receive into
670  * @_len: The amount of data we want to receive (decreased on return)
671  * @want_more: True if more data is expected to be read
672  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
673  * @_service: Where to store the actual service ID (may be upgraded)
674  *
675  * Allow a kernel service to receive data and pick up information about the
676  * state of a call.  Returns 0 if got what was asked for and there's more
677  * available, 1 if we got what was asked for and we're at the end of the data
678  * and -EAGAIN if we need more data.
679  *
680  * Note that we may return -EAGAIN to drain empty packets at the end of the
681  * data, even if we've already copied over the requested data.
682  *
683  * *_abort should also be initialised to 0.
684  */
685 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
686 			   struct iov_iter *iter, size_t *_len,
687 			   bool want_more, u32 *_abort, u16 *_service)
688 {
689 	size_t offset = 0;
690 	int ret;
691 
692 	_enter("{%d,%s},%zu,%d",
693 	       call->debug_id, rxrpc_call_states[call->state],
694 	       *_len, want_more);
695 
696 	ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_SECURING);
697 
698 	mutex_lock(&call->user_mutex);
699 
700 	switch (READ_ONCE(call->state)) {
701 	case RXRPC_CALL_CLIENT_RECV_REPLY:
702 	case RXRPC_CALL_SERVER_RECV_REQUEST:
703 	case RXRPC_CALL_SERVER_ACK_REQUEST:
704 		ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
705 					 *_len, 0, &offset);
706 		*_len -= offset;
707 		if (ret < 0)
708 			goto out;
709 
710 		/* We can only reach here with a partially full buffer if we
711 		 * have reached the end of the data.  We must otherwise have a
712 		 * full buffer or have been given -EAGAIN.
713 		 */
714 		if (ret == 1) {
715 			if (iov_iter_count(iter) > 0)
716 				goto short_data;
717 			if (!want_more)
718 				goto read_phase_complete;
719 			ret = 0;
720 			goto out;
721 		}
722 
723 		if (!want_more)
724 			goto excess_data;
725 		goto out;
726 
727 	case RXRPC_CALL_COMPLETE:
728 		goto call_complete;
729 
730 	default:
731 		ret = -EINPROGRESS;
732 		goto out;
733 	}
734 
735 read_phase_complete:
736 	ret = 1;
737 out:
738 	switch (call->ackr_reason) {
739 	case RXRPC_ACK_IDLE:
740 		break;
741 	case RXRPC_ACK_DELAY:
742 		if (ret != -EAGAIN)
743 			break;
744 		fallthrough;
745 	default:
746 		rxrpc_send_ack_packet(call, false, NULL);
747 	}
748 
749 	if (_service)
750 		*_service = call->service_id;
751 	mutex_unlock(&call->user_mutex);
752 	_leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
753 	return ret;
754 
755 short_data:
756 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
757 	ret = -EBADMSG;
758 	goto out;
759 excess_data:
760 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
761 	ret = -EMSGSIZE;
762 	goto out;
763 call_complete:
764 	*_abort = call->abort_code;
765 	ret = call->error;
766 	if (call->completion == RXRPC_CALL_SUCCEEDED) {
767 		ret = 1;
768 		if (iov_iter_count(iter) > 0)
769 			ret = -ECONNRESET;
770 	}
771 	goto out;
772 }
773 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
774