xref: /openbmc/linux/net/rxrpc/recvmsg.c (revision 2874c5fd)
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  * Pass a call terminating message to userspace.
63  */
64 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
65 {
66 	u32 tmp = 0;
67 	int ret;
68 
69 	switch (call->completion) {
70 	case RXRPC_CALL_SUCCEEDED:
71 		ret = 0;
72 		if (rxrpc_is_service_call(call))
73 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
74 		break;
75 	case RXRPC_CALL_REMOTELY_ABORTED:
76 		tmp = call->abort_code;
77 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
78 		break;
79 	case RXRPC_CALL_LOCALLY_ABORTED:
80 		tmp = call->abort_code;
81 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
82 		break;
83 	case RXRPC_CALL_NETWORK_ERROR:
84 		tmp = -call->error;
85 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
86 		break;
87 	case RXRPC_CALL_LOCAL_ERROR:
88 		tmp = -call->error;
89 		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
90 		break;
91 	default:
92 		pr_err("Invalid terminal call state %u\n", call->state);
93 		BUG();
94 		break;
95 	}
96 
97 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
98 			    call->rx_pkt_offset, call->rx_pkt_len, ret);
99 	return ret;
100 }
101 
102 /*
103  * Pass back notification of a new call.  The call is added to the
104  * to-be-accepted list.  This means that the next call to be accepted might not
105  * be the last call seen awaiting acceptance, but unless we leave this on the
106  * front of the queue and block all other messages until someone gives us a
107  * user_ID for it, there's not a lot we can do.
108  */
109 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
110 				  struct rxrpc_call *call,
111 				  struct msghdr *msg, int flags)
112 {
113 	int tmp = 0, ret;
114 
115 	ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
116 
117 	if (ret == 0 && !(flags & MSG_PEEK)) {
118 		_debug("to be accepted");
119 		write_lock_bh(&rx->recvmsg_lock);
120 		list_del_init(&call->recvmsg_link);
121 		write_unlock_bh(&rx->recvmsg_lock);
122 
123 		rxrpc_get_call(call, rxrpc_call_got);
124 		write_lock(&rx->call_lock);
125 		list_add_tail(&call->accept_link, &rx->to_be_accepted);
126 		write_unlock(&rx->call_lock);
127 	}
128 
129 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
130 	return ret;
131 }
132 
133 /*
134  * End the packet reception phase.
135  */
136 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
137 {
138 	_enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
139 
140 	trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
141 	ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
142 
143 	if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
144 		rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, serial, false, true,
145 				  rxrpc_propose_ack_terminal_ack);
146 		//rxrpc_send_ack_packet(call, false, NULL);
147 	}
148 
149 	write_lock_bh(&call->state_lock);
150 
151 	switch (call->state) {
152 	case RXRPC_CALL_CLIENT_RECV_REPLY:
153 		__rxrpc_call_completed(call);
154 		write_unlock_bh(&call->state_lock);
155 		break;
156 
157 	case RXRPC_CALL_SERVER_RECV_REQUEST:
158 		call->tx_phase = true;
159 		call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
160 		call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
161 		write_unlock_bh(&call->state_lock);
162 		rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial, false, true,
163 				  rxrpc_propose_ack_processing_op);
164 		break;
165 	default:
166 		write_unlock_bh(&call->state_lock);
167 		break;
168 	}
169 }
170 
171 /*
172  * Discard a packet we've used up and advance the Rx window by one.
173  */
174 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
175 {
176 	struct rxrpc_skb_priv *sp;
177 	struct sk_buff *skb;
178 	rxrpc_serial_t serial;
179 	rxrpc_seq_t hard_ack, top;
180 	u8 flags;
181 	int ix;
182 
183 	_enter("%d", call->debug_id);
184 
185 	hard_ack = call->rx_hard_ack;
186 	top = smp_load_acquire(&call->rx_top);
187 	ASSERT(before(hard_ack, top));
188 
189 	hard_ack++;
190 	ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
191 	skb = call->rxtx_buffer[ix];
192 	rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
193 	sp = rxrpc_skb(skb);
194 	flags = sp->hdr.flags;
195 	serial = sp->hdr.serial;
196 	if (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO)
197 		serial += (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO) - 1;
198 
199 	call->rxtx_buffer[ix] = NULL;
200 	call->rxtx_annotations[ix] = 0;
201 	/* Barrier against rxrpc_input_data(). */
202 	smp_store_release(&call->rx_hard_ack, hard_ack);
203 
204 	rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
205 
206 	_debug("%u,%u,%02x", hard_ack, top, flags);
207 	trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
208 	if (flags & RXRPC_LAST_PACKET) {
209 		rxrpc_end_rx_phase(call, serial);
210 	} else {
211 		/* Check to see if there's an ACK that needs sending. */
212 		if (after_eq(hard_ack, call->ackr_consumed + 2) ||
213 		    after_eq(top, call->ackr_seen + 2) ||
214 		    (hard_ack == top && after(hard_ack, call->ackr_consumed)))
215 			rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial,
216 					  true, true,
217 					  rxrpc_propose_ack_rotate_rx);
218 		if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
219 			rxrpc_send_ack_packet(call, false, NULL);
220 	}
221 }
222 
223 /*
224  * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
225  * padding, but if this is the case, the packet length will be resident in the
226  * socket buffer.  Note that we can't modify the master skb info as the skb may
227  * be the home to multiple subpackets.
228  */
229 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
230 			       u8 annotation,
231 			       unsigned int offset, unsigned int len)
232 {
233 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
234 	rxrpc_seq_t seq = sp->hdr.seq;
235 	u16 cksum = sp->hdr.cksum;
236 
237 	_enter("");
238 
239 	/* For all but the head jumbo subpacket, the security checksum is in a
240 	 * jumbo header immediately prior to the data.
241 	 */
242 	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
243 		__be16 tmp;
244 		if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
245 			BUG();
246 		cksum = ntohs(tmp);
247 		seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
248 	}
249 
250 	return call->conn->security->verify_packet(call, skb, offset, len,
251 						   seq, cksum);
252 }
253 
254 /*
255  * Locate the data within a packet.  This is complicated by:
256  *
257  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
258  *     subpacket.
259  *
260  * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
261  *     contains an extra header which includes the true length of the data,
262  *     excluding any encrypted padding.
263  */
264 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
265 			     u8 *_annotation,
266 			     unsigned int *_offset, unsigned int *_len)
267 {
268 	unsigned int offset = sizeof(struct rxrpc_wire_header);
269 	unsigned int len;
270 	int ret;
271 	u8 annotation = *_annotation;
272 
273 	/* Locate the subpacket */
274 	len = skb->len - offset;
275 	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
276 		offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
277 			   RXRPC_JUMBO_SUBPKTLEN);
278 		len = (annotation & RXRPC_RX_ANNO_JLAST) ?
279 			skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
280 	}
281 
282 	if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
283 		ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
284 		if (ret < 0)
285 			return ret;
286 		*_annotation |= RXRPC_RX_ANNO_VERIFIED;
287 	}
288 
289 	*_offset = offset;
290 	*_len = len;
291 	call->conn->security->locate_data(call, skb, _offset, _len);
292 	return 0;
293 }
294 
295 /*
296  * Deliver messages to a call.  This keeps processing packets until the buffer
297  * is filled and we find either more DATA (returns 0) or the end of the DATA
298  * (returns 1).  If more packets are required, it returns -EAGAIN.
299  */
300 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
301 			      struct msghdr *msg, struct iov_iter *iter,
302 			      size_t len, int flags, size_t *_offset)
303 {
304 	struct rxrpc_skb_priv *sp;
305 	struct sk_buff *skb;
306 	rxrpc_seq_t hard_ack, top, seq;
307 	size_t remain;
308 	bool last;
309 	unsigned int rx_pkt_offset, rx_pkt_len;
310 	int ix, copy, ret = -EAGAIN, ret2;
311 
312 	if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
313 	    call->ackr_reason)
314 		rxrpc_send_ack_packet(call, false, NULL);
315 
316 	rx_pkt_offset = call->rx_pkt_offset;
317 	rx_pkt_len = call->rx_pkt_len;
318 
319 	if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
320 		seq = call->rx_hard_ack;
321 		ret = 1;
322 		goto done;
323 	}
324 
325 	/* Barriers against rxrpc_input_data(). */
326 	hard_ack = call->rx_hard_ack;
327 	seq = hard_ack + 1;
328 	while (top = smp_load_acquire(&call->rx_top),
329 	       before_eq(seq, top)
330 	       ) {
331 		ix = seq & RXRPC_RXTX_BUFF_MASK;
332 		skb = call->rxtx_buffer[ix];
333 		if (!skb) {
334 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
335 					    rx_pkt_offset, rx_pkt_len, 0);
336 			break;
337 		}
338 		smp_rmb();
339 		rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
340 		sp = rxrpc_skb(skb);
341 
342 		if (!(flags & MSG_PEEK))
343 			trace_rxrpc_receive(call, rxrpc_receive_front,
344 					    sp->hdr.serial, seq);
345 
346 		if (msg)
347 			sock_recv_timestamp(msg, sock->sk, skb);
348 
349 		if (rx_pkt_offset == 0) {
350 			ret2 = rxrpc_locate_data(call, skb,
351 						 &call->rxtx_annotations[ix],
352 						 &rx_pkt_offset, &rx_pkt_len);
353 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
354 					    rx_pkt_offset, rx_pkt_len, ret2);
355 			if (ret2 < 0) {
356 				ret = ret2;
357 				goto out;
358 			}
359 		} else {
360 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
361 					    rx_pkt_offset, rx_pkt_len, 0);
362 		}
363 
364 		/* We have to handle short, empty and used-up DATA packets. */
365 		remain = len - *_offset;
366 		copy = rx_pkt_len;
367 		if (copy > remain)
368 			copy = remain;
369 		if (copy > 0) {
370 			ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
371 						      copy);
372 			if (ret2 < 0) {
373 				ret = ret2;
374 				goto out;
375 			}
376 
377 			/* handle piecemeal consumption of data packets */
378 			rx_pkt_offset += copy;
379 			rx_pkt_len -= copy;
380 			*_offset += copy;
381 		}
382 
383 		if (rx_pkt_len > 0) {
384 			trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
385 					    rx_pkt_offset, rx_pkt_len, 0);
386 			ASSERTCMP(*_offset, ==, len);
387 			ret = 0;
388 			break;
389 		}
390 
391 		/* The whole packet has been transferred. */
392 		last = sp->hdr.flags & RXRPC_LAST_PACKET;
393 		if (!(flags & MSG_PEEK))
394 			rxrpc_rotate_rx_window(call);
395 		rx_pkt_offset = 0;
396 		rx_pkt_len = 0;
397 
398 		if (last) {
399 			ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
400 			ret = 1;
401 			goto out;
402 		}
403 
404 		seq++;
405 	}
406 
407 out:
408 	if (!(flags & MSG_PEEK)) {
409 		call->rx_pkt_offset = rx_pkt_offset;
410 		call->rx_pkt_len = rx_pkt_len;
411 	}
412 done:
413 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
414 			    rx_pkt_offset, rx_pkt_len, ret);
415 	if (ret == -EAGAIN)
416 		set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
417 	return ret;
418 }
419 
420 /*
421  * Receive a message from an RxRPC socket
422  * - we need to be careful about two or more threads calling recvmsg
423  *   simultaneously
424  */
425 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
426 		  int flags)
427 {
428 	struct rxrpc_call *call;
429 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
430 	struct list_head *l;
431 	size_t copied = 0;
432 	long timeo;
433 	int ret;
434 
435 	DEFINE_WAIT(wait);
436 
437 	trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
438 
439 	if (flags & (MSG_OOB | MSG_TRUNC))
440 		return -EOPNOTSUPP;
441 
442 	timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
443 
444 try_again:
445 	lock_sock(&rx->sk);
446 
447 	/* Return immediately if a client socket has no outstanding calls */
448 	if (RB_EMPTY_ROOT(&rx->calls) &&
449 	    list_empty(&rx->recvmsg_q) &&
450 	    rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
451 		release_sock(&rx->sk);
452 		return -ENODATA;
453 	}
454 
455 	if (list_empty(&rx->recvmsg_q)) {
456 		ret = -EWOULDBLOCK;
457 		if (timeo == 0) {
458 			call = NULL;
459 			goto error_no_call;
460 		}
461 
462 		release_sock(&rx->sk);
463 
464 		/* Wait for something to happen */
465 		prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
466 					  TASK_INTERRUPTIBLE);
467 		ret = sock_error(&rx->sk);
468 		if (ret)
469 			goto wait_error;
470 
471 		if (list_empty(&rx->recvmsg_q)) {
472 			if (signal_pending(current))
473 				goto wait_interrupted;
474 			trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
475 					    0, 0, 0, 0);
476 			timeo = schedule_timeout(timeo);
477 		}
478 		finish_wait(sk_sleep(&rx->sk), &wait);
479 		goto try_again;
480 	}
481 
482 	/* Find the next call and dequeue it if we're not just peeking.  If we
483 	 * do dequeue it, that comes with a ref that we will need to release.
484 	 */
485 	write_lock_bh(&rx->recvmsg_lock);
486 	l = rx->recvmsg_q.next;
487 	call = list_entry(l, struct rxrpc_call, recvmsg_link);
488 	if (!(flags & MSG_PEEK))
489 		list_del_init(&call->recvmsg_link);
490 	else
491 		rxrpc_get_call(call, rxrpc_call_got);
492 	write_unlock_bh(&rx->recvmsg_lock);
493 
494 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
495 
496 	/* We're going to drop the socket lock, so we need to lock the call
497 	 * against interference by sendmsg.
498 	 */
499 	if (!mutex_trylock(&call->user_mutex)) {
500 		ret = -EWOULDBLOCK;
501 		if (flags & MSG_DONTWAIT)
502 			goto error_requeue_call;
503 		ret = -ERESTARTSYS;
504 		if (mutex_lock_interruptible(&call->user_mutex) < 0)
505 			goto error_requeue_call;
506 	}
507 
508 	release_sock(&rx->sk);
509 
510 	if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
511 		BUG();
512 
513 	if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
514 		if (flags & MSG_CMSG_COMPAT) {
515 			unsigned int id32 = call->user_call_ID;
516 
517 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
518 				       sizeof(unsigned int), &id32);
519 		} else {
520 			unsigned long idl = call->user_call_ID;
521 
522 			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
523 				       sizeof(unsigned long), &idl);
524 		}
525 		if (ret < 0)
526 			goto error_unlock_call;
527 	}
528 
529 	if (msg->msg_name) {
530 		struct sockaddr_rxrpc *srx = msg->msg_name;
531 		size_t len = sizeof(call->peer->srx);
532 
533 		memcpy(msg->msg_name, &call->peer->srx, len);
534 		srx->srx_service = call->service_id;
535 		msg->msg_namelen = len;
536 	}
537 
538 	switch (READ_ONCE(call->state)) {
539 	case RXRPC_CALL_SERVER_ACCEPTING:
540 		ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
541 		break;
542 	case RXRPC_CALL_CLIENT_RECV_REPLY:
543 	case RXRPC_CALL_SERVER_RECV_REQUEST:
544 	case RXRPC_CALL_SERVER_ACK_REQUEST:
545 		ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
546 					 flags, &copied);
547 		if (ret == -EAGAIN)
548 			ret = 0;
549 
550 		if (after(call->rx_top, call->rx_hard_ack) &&
551 		    call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
552 			rxrpc_notify_socket(call);
553 		break;
554 	default:
555 		ret = 0;
556 		break;
557 	}
558 
559 	if (ret < 0)
560 		goto error_unlock_call;
561 
562 	if (call->state == RXRPC_CALL_COMPLETE) {
563 		ret = rxrpc_recvmsg_term(call, msg);
564 		if (ret < 0)
565 			goto error_unlock_call;
566 		if (!(flags & MSG_PEEK))
567 			rxrpc_release_call(rx, call);
568 		msg->msg_flags |= MSG_EOR;
569 		ret = 1;
570 	}
571 
572 	if (ret == 0)
573 		msg->msg_flags |= MSG_MORE;
574 	else
575 		msg->msg_flags &= ~MSG_MORE;
576 	ret = copied;
577 
578 error_unlock_call:
579 	mutex_unlock(&call->user_mutex);
580 	rxrpc_put_call(call, rxrpc_call_put);
581 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
582 	return ret;
583 
584 error_requeue_call:
585 	if (!(flags & MSG_PEEK)) {
586 		write_lock_bh(&rx->recvmsg_lock);
587 		list_add(&call->recvmsg_link, &rx->recvmsg_q);
588 		write_unlock_bh(&rx->recvmsg_lock);
589 		trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
590 	} else {
591 		rxrpc_put_call(call, rxrpc_call_put);
592 	}
593 error_no_call:
594 	release_sock(&rx->sk);
595 error_trace:
596 	trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
597 	return ret;
598 
599 wait_interrupted:
600 	ret = sock_intr_errno(timeo);
601 wait_error:
602 	finish_wait(sk_sleep(&rx->sk), &wait);
603 	call = NULL;
604 	goto error_trace;
605 }
606 
607 /**
608  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
609  * @sock: The socket that the call exists on
610  * @call: The call to send data through
611  * @iter: The buffer to receive into
612  * @want_more: True if more data is expected to be read
613  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
614  * @_service: Where to store the actual service ID (may be upgraded)
615  *
616  * Allow a kernel service to receive data and pick up information about the
617  * state of a call.  Returns 0 if got what was asked for and there's more
618  * available, 1 if we got what was asked for and we're at the end of the data
619  * and -EAGAIN if we need more data.
620  *
621  * Note that we may return -EAGAIN to drain empty packets at the end of the
622  * data, even if we've already copied over the requested data.
623  *
624  * *_abort should also be initialised to 0.
625  */
626 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
627 			   struct iov_iter *iter,
628 			   bool want_more, u32 *_abort, u16 *_service)
629 {
630 	size_t offset = 0;
631 	int ret;
632 
633 	_enter("{%d,%s},%zu,%d",
634 	       call->debug_id, rxrpc_call_states[call->state],
635 	       iov_iter_count(iter), want_more);
636 
637 	ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
638 
639 	mutex_lock(&call->user_mutex);
640 
641 	switch (READ_ONCE(call->state)) {
642 	case RXRPC_CALL_CLIENT_RECV_REPLY:
643 	case RXRPC_CALL_SERVER_RECV_REQUEST:
644 	case RXRPC_CALL_SERVER_ACK_REQUEST:
645 		ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
646 					 iov_iter_count(iter), 0,
647 					 &offset);
648 		if (ret < 0)
649 			goto out;
650 
651 		/* We can only reach here with a partially full buffer if we
652 		 * have reached the end of the data.  We must otherwise have a
653 		 * full buffer or have been given -EAGAIN.
654 		 */
655 		if (ret == 1) {
656 			if (iov_iter_count(iter) > 0)
657 				goto short_data;
658 			if (!want_more)
659 				goto read_phase_complete;
660 			ret = 0;
661 			goto out;
662 		}
663 
664 		if (!want_more)
665 			goto excess_data;
666 		goto out;
667 
668 	case RXRPC_CALL_COMPLETE:
669 		goto call_complete;
670 
671 	default:
672 		ret = -EINPROGRESS;
673 		goto out;
674 	}
675 
676 read_phase_complete:
677 	ret = 1;
678 out:
679 	switch (call->ackr_reason) {
680 	case RXRPC_ACK_IDLE:
681 		break;
682 	case RXRPC_ACK_DELAY:
683 		if (ret != -EAGAIN)
684 			break;
685 		/* Fall through */
686 	default:
687 		rxrpc_send_ack_packet(call, false, NULL);
688 	}
689 
690 	if (_service)
691 		*_service = call->service_id;
692 	mutex_unlock(&call->user_mutex);
693 	_leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
694 	return ret;
695 
696 short_data:
697 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
698 	ret = -EBADMSG;
699 	goto out;
700 excess_data:
701 	trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
702 	ret = -EMSGSIZE;
703 	goto out;
704 call_complete:
705 	*_abort = call->abort_code;
706 	ret = call->error;
707 	if (call->completion == RXRPC_CALL_SUCCEEDED) {
708 		ret = 1;
709 		if (iov_iter_count(iter) > 0)
710 			ret = -ECONNRESET;
711 	}
712 	goto out;
713 }
714 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
715 
716 /**
717  * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
718  * @sock: The socket that the call exists on
719  * @call: The call to query
720  * @_ts: Where to put the timestamp
721  *
722  * Retrieve the timestamp from the first DATA packet of the reply if it is
723  * in the ring.  Returns true if successful, false if not.
724  */
725 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
726 				 ktime_t *_ts)
727 {
728 	struct sk_buff *skb;
729 	rxrpc_seq_t hard_ack, top, seq;
730 	bool success = false;
731 
732 	mutex_lock(&call->user_mutex);
733 
734 	if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
735 		goto out;
736 
737 	hard_ack = call->rx_hard_ack;
738 	if (hard_ack != 0)
739 		goto out;
740 
741 	seq = hard_ack + 1;
742 	top = smp_load_acquire(&call->rx_top);
743 	if (after(seq, top))
744 		goto out;
745 
746 	skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
747 	if (!skb)
748 		goto out;
749 
750 	*_ts = skb_get_ktime(skb);
751 	success = true;
752 
753 out:
754 	mutex_unlock(&call->user_mutex);
755 	return success;
756 }
757 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);
758