xref: /openbmc/linux/net/rxrpc/recvmsg.c (revision d4295e12)
1 /* RxRPC recvmsg() implementation
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 
14 #include <linux/net.h>
15 #include <linux/skbuff.h>
16 #include <linux/export.h>
17 #include <linux/sched/signal.h>
18 
19 #include <net/sock.h>
20 #include <net/af_rxrpc.h>
21 #include "ar-internal.h"
22 
23 /*
24  * Post a call for attention by the socket or kernel service.  Further
25  * notifications are suppressed by putting recvmsg_link on a dummy queue.
26  */
27 void rxrpc_notify_socket(struct rxrpc_call *call)
28 {
29 	struct rxrpc_sock *rx;
30 	struct sock *sk;
31 
32 	_enter("%d", call->debug_id);
33 
34 	if (!list_empty(&call->recvmsg_link))
35 		return;
36 
37 	rcu_read_lock();
38 
39 	rx = rcu_dereference(call->socket);
40 	sk = &rx->sk;
41 	if (rx && sk->sk_state < RXRPC_CLOSE) {
42 		if (call->notify_rx) {
43 			spin_lock_bh(&call->notify_lock);
44 			call->notify_rx(sk, call, call->user_call_ID);
45 			spin_unlock_bh(&call->notify_lock);
46 		} else {
47 			write_lock_bh(&rx->recvmsg_lock);
48 			if (list_empty(&call->recvmsg_link)) {
49 				rxrpc_get_call(call, rxrpc_call_got);
50 				list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
51 			}
52 			write_unlock_bh(&rx->recvmsg_lock);
53 
54 			if (!sock_flag(sk, SOCK_DEAD)) {
55 				_debug("call %ps", sk->sk_data_ready);
56 				sk->sk_data_ready(sk);
57 			}
58 		}
59 	}
60 
61 	rcu_read_unlock();
62 	_leave("");
63 }
64 
65 /*
66  * Pass a call terminating message to userspace.
67  */
68 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
69 {
70 	u32 tmp = 0;
71 	int ret;
72 
73 	switch (call->completion) {
74 	case RXRPC_CALL_SUCCEEDED:
75 		ret = 0;
76 		if (rxrpc_is_service_call(call))
77 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
78 		break;
79 	case RXRPC_CALL_REMOTELY_ABORTED:
80 		tmp = call->abort_code;
81 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
82 		break;
83 	case RXRPC_CALL_LOCALLY_ABORTED:
84 		tmp = call->abort_code;
85 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
86 		break;
87 	case RXRPC_CALL_NETWORK_ERROR:
88 		tmp = -call->error;
89 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
90 		break;
91 	case RXRPC_CALL_LOCAL_ERROR:
92 		tmp = -call->error;
93 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
94 		break;
95 	default:
96 		pr_err("Invalid terminal call state %u\n", call->state);
97 		BUG();
98 		break;
99 	}
100 
101 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
102 			    call->rx_pkt_offset, call->rx_pkt_len, ret);
103 	return ret;
104 }
105 
106 /*
107  * Pass back notification of a new call.  The call is added to the
108  * to-be-accepted list.  This means that the next call to be accepted might not
109  * be the last call seen awaiting acceptance, but unless we leave this on the
110  * front of the queue and block all other messages until someone gives us a
111  * user_ID for it, there's not a lot we can do.
112  */
113 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
114 				  struct rxrpc_call *call,
115 				  struct msghdr *msg, int flags)
116 {
117 	int tmp = 0, ret;
118 
119 	ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
120 
121 	if (ret == 0 && !(flags & MSG_PEEK)) {
122 		_debug("to be accepted");
123 		write_lock_bh(&rx->recvmsg_lock);
124 		list_del_init(&call->recvmsg_link);
125 		write_unlock_bh(&rx->recvmsg_lock);
126 
127 		rxrpc_get_call(call, rxrpc_call_got);
128 		write_lock(&rx->call_lock);
129 		list_add_tail(&call->accept_link, &rx->to_be_accepted);
130 		write_unlock(&rx->call_lock);
131 	}
132 
133 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
134 	return ret;
135 }
136 
137 /*
138  * End the packet reception phase.
139  */
140 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
141 {
142 	_enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
143 
144 	trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
145 	ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
146 
147 	if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
148 		rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, serial, false, true,
149 				  rxrpc_propose_ack_terminal_ack);
150 		//rxrpc_send_ack_packet(call, false, NULL);
151 	}
152 
153 	write_lock_bh(&call->state_lock);
154 
155 	switch (call->state) {
156 	case RXRPC_CALL_CLIENT_RECV_REPLY:
157 		__rxrpc_call_completed(call);
158 		write_unlock_bh(&call->state_lock);
159 		break;
160 
161 	case RXRPC_CALL_SERVER_RECV_REQUEST:
162 		call->tx_phase = true;
163 		call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
164 		call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
165 		write_unlock_bh(&call->state_lock);
166 		rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial, false, true,
167 				  rxrpc_propose_ack_processing_op);
168 		break;
169 	default:
170 		write_unlock_bh(&call->state_lock);
171 		break;
172 	}
173 }
174 
175 /*
176  * Discard a packet we've used up and advance the Rx window by one.
177  */
178 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
179 {
180 	struct rxrpc_skb_priv *sp;
181 	struct sk_buff *skb;
182 	rxrpc_serial_t serial;
183 	rxrpc_seq_t hard_ack, top;
184 	u8 flags;
185 	int ix;
186 
187 	_enter("%d", call->debug_id);
188 
189 	hard_ack = call->rx_hard_ack;
190 	top = smp_load_acquire(&call->rx_top);
191 	ASSERT(before(hard_ack, top));
192 
193 	hard_ack++;
194 	ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
195 	skb = call->rxtx_buffer[ix];
196 	rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
197 	sp = rxrpc_skb(skb);
198 	flags = sp->hdr.flags;
199 	serial = sp->hdr.serial;
200 	if (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO)
201 		serial += (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO) - 1;
202 
203 	call->rxtx_buffer[ix] = NULL;
204 	call->rxtx_annotations[ix] = 0;
205 	/* Barrier against rxrpc_input_data(). */
206 	smp_store_release(&call->rx_hard_ack, hard_ack);
207 
208 	rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
209 
210 	_debug("%u,%u,%02x", hard_ack, top, flags);
211 	trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
212 	if (flags & RXRPC_LAST_PACKET) {
213 		rxrpc_end_rx_phase(call, serial);
214 	} else {
215 		/* Check to see if there's an ACK that needs sending. */
216 		if (after_eq(hard_ack, call->ackr_consumed + 2) ||
217 		    after_eq(top, call->ackr_seen + 2) ||
218 		    (hard_ack == top && after(hard_ack, call->ackr_consumed)))
219 			rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial,
220 					  true, true,
221 					  rxrpc_propose_ack_rotate_rx);
222 		if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
223 			rxrpc_send_ack_packet(call, false, NULL);
224 	}
225 }
226 
227 /*
228  * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
229  * padding, but if this is the case, the packet length will be resident in the
230  * socket buffer.  Note that we can't modify the master skb info as the skb may
231  * be the home to multiple subpackets.
232  */
233 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
234 			       u8 annotation,
235 			       unsigned int offset, unsigned int len)
236 {
237 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
238 	rxrpc_seq_t seq = sp->hdr.seq;
239 	u16 cksum = sp->hdr.cksum;
240 
241 	_enter("");
242 
243 	/* For all but the head jumbo subpacket, the security checksum is in a
244 	 * jumbo header immediately prior to the data.
245 	 */
246 	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
247 		__be16 tmp;
248 		if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
249 			BUG();
250 		cksum = ntohs(tmp);
251 		seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
252 	}
253 
254 	return call->conn->security->verify_packet(call, skb, offset, len,
255 						   seq, cksum);
256 }
257 
258 /*
259  * Locate the data within a packet.  This is complicated by:
260  *
261  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
262  *     subpacket.
263  *
264  * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
265  *     contains an extra header which includes the true length of the data,
266  *     excluding any encrypted padding.
267  */
268 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
269 			     u8 *_annotation,
270 			     unsigned int *_offset, unsigned int *_len)
271 {
272 	unsigned int offset = sizeof(struct rxrpc_wire_header);
273 	unsigned int len;
274 	int ret;
275 	u8 annotation = *_annotation;
276 
277 	/* Locate the subpacket */
278 	len = skb->len - offset;
279 	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
280 		offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
281 			   RXRPC_JUMBO_SUBPKTLEN);
282 		len = (annotation & RXRPC_RX_ANNO_JLAST) ?
283 			skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
284 	}
285 
286 	if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
287 		ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
288 		if (ret < 0)
289 			return ret;
290 		*_annotation |= RXRPC_RX_ANNO_VERIFIED;
291 	}
292 
293 	*_offset = offset;
294 	*_len = len;
295 	call->conn->security->locate_data(call, skb, _offset, _len);
296 	return 0;
297 }
298 
299 /*
300  * Deliver messages to a call.  This keeps processing packets until the buffer
301  * is filled and we find either more DATA (returns 0) or the end of the DATA
302  * (returns 1).  If more packets are required, it returns -EAGAIN.
303  */
304 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
305 			      struct msghdr *msg, struct iov_iter *iter,
306 			      size_t len, int flags, size_t *_offset)
307 {
308 	struct rxrpc_skb_priv *sp;
309 	struct sk_buff *skb;
310 	rxrpc_seq_t hard_ack, top, seq;
311 	size_t remain;
312 	bool last;
313 	unsigned int rx_pkt_offset, rx_pkt_len;
314 	int ix, copy, ret = -EAGAIN, ret2;
315 
316 	if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
317 	    call->ackr_reason)
318 		rxrpc_send_ack_packet(call, false, NULL);
319 
320 	rx_pkt_offset = call->rx_pkt_offset;
321 	rx_pkt_len = call->rx_pkt_len;
322 
323 	if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
324 		seq = call->rx_hard_ack;
325 		ret = 1;
326 		goto done;
327 	}
328 
329 	/* Barriers against rxrpc_input_data(). */
330 	hard_ack = call->rx_hard_ack;
331 	seq = hard_ack + 1;
332 	while (top = smp_load_acquire(&call->rx_top),
333 	       before_eq(seq, top)
334 	       ) {
335 		ix = seq & RXRPC_RXTX_BUFF_MASK;
336 		skb = call->rxtx_buffer[ix];
337 		if (!skb) {
338 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
339 					    rx_pkt_offset, rx_pkt_len, 0);
340 			break;
341 		}
342 		smp_rmb();
343 		rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
344 		sp = rxrpc_skb(skb);
345 
346 		if (!(flags & MSG_PEEK))
347 			trace_rxrpc_receive(call, rxrpc_receive_front,
348 					    sp->hdr.serial, seq);
349 
350 		if (msg)
351 			sock_recv_timestamp(msg, sock->sk, skb);
352 
353 		if (rx_pkt_offset == 0) {
354 			ret2 = rxrpc_locate_data(call, skb,
355 						 &call->rxtx_annotations[ix],
356 						 &rx_pkt_offset, &rx_pkt_len);
357 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
358 					    rx_pkt_offset, rx_pkt_len, ret2);
359 			if (ret2 < 0) {
360 				ret = ret2;
361 				goto out;
362 			}
363 		} else {
364 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
365 					    rx_pkt_offset, rx_pkt_len, 0);
366 		}
367 
368 		/* We have to handle short, empty and used-up DATA packets. */
369 		remain = len - *_offset;
370 		copy = rx_pkt_len;
371 		if (copy > remain)
372 			copy = remain;
373 		if (copy > 0) {
374 			ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
375 						      copy);
376 			if (ret2 < 0) {
377 				ret = ret2;
378 				goto out;
379 			}
380 
381 			/* handle piecemeal consumption of data packets */
382 			rx_pkt_offset += copy;
383 			rx_pkt_len -= copy;
384 			*_offset += copy;
385 		}
386 
387 		if (rx_pkt_len > 0) {
388 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
389 					    rx_pkt_offset, rx_pkt_len, 0);
390 			ASSERTCMP(*_offset, ==, len);
391 			ret = 0;
392 			break;
393 		}
394 
395 		/* The whole packet has been transferred. */
396 		last = sp->hdr.flags & RXRPC_LAST_PACKET;
397 		if (!(flags & MSG_PEEK))
398 			rxrpc_rotate_rx_window(call);
399 		rx_pkt_offset = 0;
400 		rx_pkt_len = 0;
401 
402 		if (last) {
403 			ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
404 			ret = 1;
405 			goto out;
406 		}
407 
408 		seq++;
409 	}
410 
411 out:
412 	if (!(flags & MSG_PEEK)) {
413 		call->rx_pkt_offset = rx_pkt_offset;
414 		call->rx_pkt_len = rx_pkt_len;
415 	}
416 done:
417 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
418 			    rx_pkt_offset, rx_pkt_len, ret);
419 	if (ret == -EAGAIN)
420 		set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
421 	return ret;
422 }
423 
424 /*
425  * Receive a message from an RxRPC socket
426  * - we need to be careful about two or more threads calling recvmsg
427  *   simultaneously
428  */
429 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
430 		  int flags)
431 {
432 	struct rxrpc_call *call;
433 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
434 	struct list_head *l;
435 	size_t copied = 0;
436 	long timeo;
437 	int ret;
438 
439 	DEFINE_WAIT(wait);
440 
441 	trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
442 
443 	if (flags & (MSG_OOB | MSG_TRUNC))
444 		return -EOPNOTSUPP;
445 
446 	timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
447 
448 try_again:
449 	lock_sock(&rx->sk);
450 
451 	/* Return immediately if a client socket has no outstanding calls */
452 	if (RB_EMPTY_ROOT(&rx->calls) &&
453 	    list_empty(&rx->recvmsg_q) &&
454 	    rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
455 		release_sock(&rx->sk);
456 		return -ENODATA;
457 	}
458 
459 	if (list_empty(&rx->recvmsg_q)) {
460 		ret = -EWOULDBLOCK;
461 		if (timeo == 0) {
462 			call = NULL;
463 			goto error_no_call;
464 		}
465 
466 		release_sock(&rx->sk);
467 
468 		/* Wait for something to happen */
469 		prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
470 					  TASK_INTERRUPTIBLE);
471 		ret = sock_error(&rx->sk);
472 		if (ret)
473 			goto wait_error;
474 
475 		if (list_empty(&rx->recvmsg_q)) {
476 			if (signal_pending(current))
477 				goto wait_interrupted;
478 			trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
479 					    0, 0, 0, 0);
480 			timeo = schedule_timeout(timeo);
481 		}
482 		finish_wait(sk_sleep(&rx->sk), &wait);
483 		goto try_again;
484 	}
485 
486 	/* Find the next call and dequeue it if we're not just peeking.  If we
487 	 * do dequeue it, that comes with a ref that we will need to release.
488 	 */
489 	write_lock_bh(&rx->recvmsg_lock);
490 	l = rx->recvmsg_q.next;
491 	call = list_entry(l, struct rxrpc_call, recvmsg_link);
492 	if (!(flags & MSG_PEEK))
493 		list_del_init(&call->recvmsg_link);
494 	else
495 		rxrpc_get_call(call, rxrpc_call_got);
496 	write_unlock_bh(&rx->recvmsg_lock);
497 
498 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
499 
500 	/* We're going to drop the socket lock, so we need to lock the call
501 	 * against interference by sendmsg.
502 	 */
503 	if (!mutex_trylock(&call->user_mutex)) {
504 		ret = -EWOULDBLOCK;
505 		if (flags & MSG_DONTWAIT)
506 			goto error_requeue_call;
507 		ret = -ERESTARTSYS;
508 		if (mutex_lock_interruptible(&call->user_mutex) < 0)
509 			goto error_requeue_call;
510 	}
511 
512 	release_sock(&rx->sk);
513 
514 	if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
515 		BUG();
516 
517 	if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
518 		if (flags & MSG_CMSG_COMPAT) {
519 			unsigned int id32 = call->user_call_ID;
520 
521 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
522 				       sizeof(unsigned int), &id32);
523 		} else {
524 			unsigned long idl = call->user_call_ID;
525 
526 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
527 				       sizeof(unsigned long), &idl);
528 		}
529 		if (ret < 0)
530 			goto error_unlock_call;
531 	}
532 
533 	if (msg->msg_name) {
534 		struct sockaddr_rxrpc *srx = msg->msg_name;
535 		size_t len = sizeof(call->peer->srx);
536 
537 		memcpy(msg->msg_name, &call->peer->srx, len);
538 		srx->srx_service = call->service_id;
539 		msg->msg_namelen = len;
540 	}
541 
542 	switch (READ_ONCE(call->state)) {
543 	case RXRPC_CALL_SERVER_ACCEPTING:
544 		ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
545 		break;
546 	case RXRPC_CALL_CLIENT_RECV_REPLY:
547 	case RXRPC_CALL_SERVER_RECV_REQUEST:
548 	case RXRPC_CALL_SERVER_ACK_REQUEST:
549 		ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
550 					 flags, &copied);
551 		if (ret == -EAGAIN)
552 			ret = 0;
553 
554 		if (after(call->rx_top, call->rx_hard_ack) &&
555 		    call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
556 			rxrpc_notify_socket(call);
557 		break;
558 	default:
559 		ret = 0;
560 		break;
561 	}
562 
563 	if (ret < 0)
564 		goto error_unlock_call;
565 
566 	if (call->state == RXRPC_CALL_COMPLETE) {
567 		ret = rxrpc_recvmsg_term(call, msg);
568 		if (ret < 0)
569 			goto error_unlock_call;
570 		if (!(flags & MSG_PEEK))
571 			rxrpc_release_call(rx, call);
572 		msg->msg_flags |= MSG_EOR;
573 		ret = 1;
574 	}
575 
576 	if (ret == 0)
577 		msg->msg_flags |= MSG_MORE;
578 	else
579 		msg->msg_flags &= ~MSG_MORE;
580 	ret = copied;
581 
582 error_unlock_call:
583 	mutex_unlock(&call->user_mutex);
584 	rxrpc_put_call(call, rxrpc_call_put);
585 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
586 	return ret;
587 
588 error_requeue_call:
589 	if (!(flags & MSG_PEEK)) {
590 		write_lock_bh(&rx->recvmsg_lock);
591 		list_add(&call->recvmsg_link, &rx->recvmsg_q);
592 		write_unlock_bh(&rx->recvmsg_lock);
593 		trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
594 	} else {
595 		rxrpc_put_call(call, rxrpc_call_put);
596 	}
597 error_no_call:
598 	release_sock(&rx->sk);
599 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
600 	return ret;
601 
602 wait_interrupted:
603 	ret = sock_intr_errno(timeo);
604 wait_error:
605 	finish_wait(sk_sleep(&rx->sk), &wait);
606 	call = NULL;
607 	goto error_no_call;
608 }
609 
610 /**
611  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
612  * @sock: The socket that the call exists on
613  * @call: The call to send data through
614  * @iter: The buffer to receive into
615  * @want_more: True if more data is expected to be read
616  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
617  * @_service: Where to store the actual service ID (may be upgraded)
618  *
619  * Allow a kernel service to receive data and pick up information about the
620  * state of a call.  Returns 0 if got what was asked for and there's more
621  * available, 1 if we got what was asked for and we're at the end of the data
622  * and -EAGAIN if we need more data.
623  *
624  * Note that we may return -EAGAIN to drain empty packets at the end of the
625  * data, even if we've already copied over the requested data.
626  *
627  * *_abort should also be initialised to 0.
628  */
629 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
630 			   struct iov_iter *iter,
631 			   bool want_more, u32 *_abort, u16 *_service)
632 {
633 	size_t offset = 0;
634 	int ret;
635 
636 	_enter("{%d,%s},%zu,%d",
637 	       call->debug_id, rxrpc_call_states[call->state],
638 	       iov_iter_count(iter), want_more);
639 
640 	ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
641 
642 	mutex_lock(&call->user_mutex);
643 
644 	switch (READ_ONCE(call->state)) {
645 	case RXRPC_CALL_CLIENT_RECV_REPLY:
646 	case RXRPC_CALL_SERVER_RECV_REQUEST:
647 	case RXRPC_CALL_SERVER_ACK_REQUEST:
648 		ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
649 					 iov_iter_count(iter), 0,
650 					 &offset);
651 		if (ret < 0)
652 			goto out;
653 
654 		/* We can only reach here with a partially full buffer if we
655 		 * have reached the end of the data.  We must otherwise have a
656 		 * full buffer or have been given -EAGAIN.
657 		 */
658 		if (ret == 1) {
659 			if (iov_iter_count(iter) > 0)
660 				goto short_data;
661 			if (!want_more)
662 				goto read_phase_complete;
663 			ret = 0;
664 			goto out;
665 		}
666 
667 		if (!want_more)
668 			goto excess_data;
669 		goto out;
670 
671 	case RXRPC_CALL_COMPLETE:
672 		goto call_complete;
673 
674 	default:
675 		ret = -EINPROGRESS;
676 		goto out;
677 	}
678 
679 read_phase_complete:
680 	ret = 1;
681 out:
682 	switch (call->ackr_reason) {
683 	case RXRPC_ACK_IDLE:
684 		break;
685 	case RXRPC_ACK_DELAY:
686 		if (ret != -EAGAIN)
687 			break;
688 		/* Fall through */
689 	default:
690 		rxrpc_send_ack_packet(call, false, NULL);
691 	}
692 
693 	if (_service)
694 		*_service = call->service_id;
695 	mutex_unlock(&call->user_mutex);
696 	_leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
697 	return ret;
698 
699 short_data:
700 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
701 	ret = -EBADMSG;
702 	goto out;
703 excess_data:
704 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
705 	ret = -EMSGSIZE;
706 	goto out;
707 call_complete:
708 	*_abort = call->abort_code;
709 	ret = call->error;
710 	if (call->completion == RXRPC_CALL_SUCCEEDED) {
711 		ret = 1;
712 		if (iov_iter_count(iter) > 0)
713 			ret = -ECONNRESET;
714 	}
715 	goto out;
716 }
717 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
718 
719 /**
720  * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
721  * @sock: The socket that the call exists on
722  * @call: The call to query
723  * @_ts: Where to put the timestamp
724  *
725  * Retrieve the timestamp from the first DATA packet of the reply if it is
726  * in the ring.  Returns true if successful, false if not.
727  */
728 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
729 				 ktime_t *_ts)
730 {
731 	struct sk_buff *skb;
732 	rxrpc_seq_t hard_ack, top, seq;
733 	bool success = false;
734 
735 	mutex_lock(&call->user_mutex);
736 
737 	if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
738 		goto out;
739 
740 	hard_ack = call->rx_hard_ack;
741 	if (hard_ack != 0)
742 		goto out;
743 
744 	seq = hard_ack + 1;
745 	top = smp_load_acquire(&call->rx_top);
746 	if (after(seq, top))
747 		goto out;
748 
749 	skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
750 	if (!skb)
751 		goto out;
752 
753 	*_ts = skb_get_ktime(skb);
754 	success = true;
755 
756 out:
757 	mutex_unlock(&call->user_mutex);
758 	return success;
759 }
760 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);
761