1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* RxRPC recvmsg() implementation
4 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/net.h>
11 #include <linux/skbuff.h>
12 #include <linux/export.h>
13 #include <linux/sched/signal.h>
16 #include <net/af_rxrpc.h>
17 #include "ar-internal.h"
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.
23 void rxrpc_notify_socket(struct rxrpc_call *call)
25 struct rxrpc_sock *rx;
28 _enter("%d", call->debug_id);
30 if (!list_empty(&call->recvmsg_link))
35 rx = rcu_dereference(call->socket);
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);
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);
48 write_unlock_bh(&rx->recvmsg_lock);
50 if (!sock_flag(sk, SOCK_DEAD)) {
51 _debug("call %ps", sk->sk_data_ready);
52 sk->sk_data_ready(sk);
62 * Transition a call to the complete state.
64 bool __rxrpc_set_call_completion(struct rxrpc_call *call,
65 enum rxrpc_call_completion compl,
69 if (call->state < RXRPC_CALL_COMPLETE) {
70 call->abort_code = abort_code;
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);
82 bool rxrpc_set_call_completion(struct rxrpc_call *call,
83 enum rxrpc_call_completion compl,
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);
98 * Record that a call successfully completed.
100 bool __rxrpc_call_completed(struct rxrpc_call *call)
102 return __rxrpc_set_call_completion(call, RXRPC_CALL_SUCCEEDED, 0, 0);
105 bool rxrpc_call_completed(struct rxrpc_call *call)
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);
118 * Record that a call is locally aborted.
120 bool __rxrpc_abort_call(const char *why, struct rxrpc_call *call,
121 rxrpc_seq_t seq, u32 abort_code, int error)
123 trace_rxrpc_abort(call->debug_id, why, call->cid, call->call_id, seq,
125 return __rxrpc_set_call_completion(call, RXRPC_CALL_LOCALLY_ABORTED,
129 bool rxrpc_abort_call(const char *why, struct rxrpc_call *call,
130 rxrpc_seq_t seq, u32 abort_code, int error)
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);
141 * Pass a call terminating message to userspace.
143 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
148 switch (call->completion) {
149 case RXRPC_CALL_SUCCEEDED:
151 if (rxrpc_is_service_call(call))
152 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
154 case RXRPC_CALL_REMOTELY_ABORTED:
155 tmp = call->abort_code;
156 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
158 case RXRPC_CALL_LOCALLY_ABORTED:
159 tmp = call->abort_code;
160 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
162 case RXRPC_CALL_NETWORK_ERROR:
164 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
166 case RXRPC_CALL_LOCAL_ERROR:
168 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
171 pr_err("Invalid terminal call state %u\n", call->state);
176 trace_rxrpc_recvdata(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
177 call->rx_pkt_offset, call->rx_pkt_len, ret);
182 * End the packet reception phase.
184 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
186 _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
188 trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
189 ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
191 if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY)
192 rxrpc_propose_delay_ACK(call, serial, rxrpc_propose_ack_terminal_ack);
194 write_lock_bh(&call->state_lock);
196 switch (call->state) {
197 case RXRPC_CALL_CLIENT_RECV_REPLY:
198 __rxrpc_call_completed(call);
199 write_unlock_bh(&call->state_lock);
202 case RXRPC_CALL_SERVER_RECV_REQUEST:
203 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
204 call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
205 write_unlock_bh(&call->state_lock);
206 rxrpc_propose_delay_ACK(call, serial,
207 rxrpc_propose_ack_processing_op);
210 write_unlock_bh(&call->state_lock);
216 * Discard a packet we've used up and advance the Rx window by one.
218 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
220 struct rxrpc_skb_priv *sp;
222 rxrpc_serial_t serial;
223 rxrpc_seq_t hard_ack, top;
227 _enter("%d", call->debug_id);
229 hard_ack = call->rx_hard_ack;
230 top = smp_load_acquire(&call->rx_top);
231 ASSERT(before(hard_ack, top));
234 ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
235 skb = call->rxtx_buffer[ix];
236 rxrpc_see_skb(skb, rxrpc_skb_rotated);
239 serial = sp->hdr.serial;
241 if (sp->hdr.flags & RXRPC_LAST_PACKET)
244 call->rxtx_buffer[ix] = NULL;
245 call->rxtx_annotations[ix] = 0;
246 /* Barrier against rxrpc_input_data(). */
247 smp_store_release(&call->rx_hard_ack, hard_ack);
249 rxrpc_free_skb(skb, rxrpc_skb_freed);
251 trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
253 rxrpc_end_rx_phase(call, serial);
255 /* Check to see if there's an ACK that needs sending. */
256 if (atomic_inc_return(&call->ackr_nr_consumed) > 2 &&
257 !test_and_set_bit(RXRPC_CALL_IDLE_ACK_PENDING, &call->flags)) {
258 rxrpc_send_ACK(call, RXRPC_ACK_IDLE, serial,
259 rxrpc_propose_ack_rotate_rx);
260 rxrpc_transmit_ack_packets(call->peer->local);
266 * Decrypt and verify a DATA packet.
268 static int rxrpc_verify_data(struct rxrpc_call *call, struct sk_buff *skb)
270 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
272 if (sp->flags & RXRPC_RX_VERIFIED)
274 return call->security->verify_packet(call, skb);
278 * Deliver messages to a call. This keeps processing packets until the buffer
279 * is filled and we find either more DATA (returns 0) or the end of the DATA
280 * (returns 1). If more packets are required, it returns -EAGAIN.
282 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
283 struct msghdr *msg, struct iov_iter *iter,
284 size_t len, int flags, size_t *_offset)
286 struct rxrpc_skb_priv *sp;
288 rxrpc_serial_t serial;
289 rxrpc_seq_t hard_ack, top, seq;
291 unsigned int rx_pkt_offset, rx_pkt_len;
292 int ix, copy, ret = -EAGAIN, ret2;
294 rx_pkt_offset = call->rx_pkt_offset;
295 rx_pkt_len = call->rx_pkt_len;
297 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
298 seq = call->rx_hard_ack;
303 /* Barriers against rxrpc_input_data(). */
304 hard_ack = call->rx_hard_ack;
307 while (top = smp_load_acquire(&call->rx_top),
310 ix = seq & RXRPC_RXTX_BUFF_MASK;
311 skb = call->rxtx_buffer[ix];
313 trace_rxrpc_recvdata(call, rxrpc_recvmsg_hole, seq,
314 rx_pkt_offset, rx_pkt_len, 0);
315 rxrpc_transmit_ack_packets(call->peer->local);
319 rxrpc_see_skb(skb, rxrpc_skb_seen);
322 if (!(flags & MSG_PEEK)) {
323 serial = sp->hdr.serial;
324 trace_rxrpc_receive(call, rxrpc_receive_front,
329 sock_recv_timestamp(msg, sock->sk, skb);
331 if (rx_pkt_offset == 0) {
332 ret2 = rxrpc_verify_data(call, skb);
333 rx_pkt_offset = sp->offset;
334 rx_pkt_len = sp->len;
335 trace_rxrpc_recvdata(call, rxrpc_recvmsg_next, seq,
336 rx_pkt_offset, rx_pkt_len, ret2);
342 trace_rxrpc_recvdata(call, rxrpc_recvmsg_cont, seq,
343 rx_pkt_offset, rx_pkt_len, 0);
346 /* We have to handle short, empty and used-up DATA packets. */
347 remain = len - *_offset;
352 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
359 /* handle piecemeal consumption of data packets */
360 rx_pkt_offset += copy;
365 if (rx_pkt_len > 0) {
366 trace_rxrpc_recvdata(call, rxrpc_recvmsg_full, seq,
367 rx_pkt_offset, rx_pkt_len, 0);
368 ASSERTCMP(*_offset, ==, len);
373 /* The whole packet has been transferred. */
374 if (sp->hdr.flags & RXRPC_LAST_PACKET)
379 if (!(flags & MSG_PEEK))
380 rxrpc_rotate_rx_window(call);
386 if (!(flags & MSG_PEEK)) {
387 call->rx_pkt_offset = rx_pkt_offset;
388 call->rx_pkt_len = rx_pkt_len;
391 trace_rxrpc_recvdata(call, rxrpc_recvmsg_data_return, seq,
392 rx_pkt_offset, rx_pkt_len, ret);
394 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
399 * Receive a message from an RxRPC socket
400 * - we need to be careful about two or more threads calling recvmsg
403 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
406 struct rxrpc_call *call;
407 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
415 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0);
417 if (flags & (MSG_OOB | MSG_TRUNC))
420 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
425 /* Return immediately if a client socket has no outstanding calls */
426 if (RB_EMPTY_ROOT(&rx->calls) &&
427 list_empty(&rx->recvmsg_q) &&
428 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
429 release_sock(&rx->sk);
433 if (list_empty(&rx->recvmsg_q)) {
440 release_sock(&rx->sk);
442 /* Wait for something to happen */
443 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
445 ret = sock_error(&rx->sk);
449 if (list_empty(&rx->recvmsg_q)) {
450 if (signal_pending(current))
451 goto wait_interrupted;
452 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait, 0);
453 timeo = schedule_timeout(timeo);
455 finish_wait(sk_sleep(&rx->sk), &wait);
459 /* Find the next call and dequeue it if we're not just peeking. If we
460 * do dequeue it, that comes with a ref that we will need to release.
462 write_lock_bh(&rx->recvmsg_lock);
463 l = rx->recvmsg_q.next;
464 call = list_entry(l, struct rxrpc_call, recvmsg_link);
465 if (!(flags & MSG_PEEK))
466 list_del_init(&call->recvmsg_link);
468 rxrpc_get_call(call, rxrpc_call_got);
469 write_unlock_bh(&rx->recvmsg_lock);
471 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0);
473 /* We're going to drop the socket lock, so we need to lock the call
474 * against interference by sendmsg.
476 if (!mutex_trylock(&call->user_mutex)) {
478 if (flags & MSG_DONTWAIT)
479 goto error_requeue_call;
481 if (mutex_lock_interruptible(&call->user_mutex) < 0)
482 goto error_requeue_call;
485 release_sock(&rx->sk);
487 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
490 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
491 if (flags & MSG_CMSG_COMPAT) {
492 unsigned int id32 = call->user_call_ID;
494 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
495 sizeof(unsigned int), &id32);
497 unsigned long idl = call->user_call_ID;
499 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
500 sizeof(unsigned long), &idl);
503 goto error_unlock_call;
506 if (msg->msg_name && call->peer) {
507 struct sockaddr_rxrpc *srx = msg->msg_name;
508 size_t len = sizeof(call->peer->srx);
510 memcpy(msg->msg_name, &call->peer->srx, len);
511 srx->srx_service = call->service_id;
512 msg->msg_namelen = len;
515 switch (READ_ONCE(call->state)) {
516 case RXRPC_CALL_CLIENT_RECV_REPLY:
517 case RXRPC_CALL_SERVER_RECV_REQUEST:
518 case RXRPC_CALL_SERVER_ACK_REQUEST:
519 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
524 rxrpc_transmit_ack_packets(call->peer->local);
525 if (after(call->rx_top, call->rx_hard_ack) &&
526 call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
527 rxrpc_notify_socket(call);
535 goto error_unlock_call;
537 if (call->state == RXRPC_CALL_COMPLETE) {
538 ret = rxrpc_recvmsg_term(call, msg);
540 goto error_unlock_call;
541 if (!(flags & MSG_PEEK))
542 rxrpc_release_call(rx, call);
543 msg->msg_flags |= MSG_EOR;
548 msg->msg_flags |= MSG_MORE;
550 msg->msg_flags &= ~MSG_MORE;
554 mutex_unlock(&call->user_mutex);
555 rxrpc_put_call(call, rxrpc_call_put);
556 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, ret);
560 if (!(flags & MSG_PEEK)) {
561 write_lock_bh(&rx->recvmsg_lock);
562 list_add(&call->recvmsg_link, &rx->recvmsg_q);
563 write_unlock_bh(&rx->recvmsg_lock);
564 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0);
566 rxrpc_put_call(call, rxrpc_call_put);
569 release_sock(&rx->sk);
571 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, ret);
575 ret = sock_intr_errno(timeo);
577 finish_wait(sk_sleep(&rx->sk), &wait);
583 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
584 * @sock: The socket that the call exists on
585 * @call: The call to send data through
586 * @iter: The buffer to receive into
587 * @_len: The amount of data we want to receive (decreased on return)
588 * @want_more: True if more data is expected to be read
589 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
590 * @_service: Where to store the actual service ID (may be upgraded)
592 * Allow a kernel service to receive data and pick up information about the
593 * state of a call. Returns 0 if got what was asked for and there's more
594 * available, 1 if we got what was asked for and we're at the end of the data
595 * and -EAGAIN if we need more data.
597 * Note that we may return -EAGAIN to drain empty packets at the end of the
598 * data, even if we've already copied over the requested data.
600 * *_abort should also be initialised to 0.
602 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
603 struct iov_iter *iter, size_t *_len,
604 bool want_more, u32 *_abort, u16 *_service)
609 _enter("{%d,%s},%zu,%d",
610 call->debug_id, rxrpc_call_states[call->state],
613 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_SECURING);
615 mutex_lock(&call->user_mutex);
617 switch (READ_ONCE(call->state)) {
618 case RXRPC_CALL_CLIENT_RECV_REPLY:
619 case RXRPC_CALL_SERVER_RECV_REQUEST:
620 case RXRPC_CALL_SERVER_ACK_REQUEST:
621 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
627 /* We can only reach here with a partially full buffer if we
628 * have reached the end of the data. We must otherwise have a
629 * full buffer or have been given -EAGAIN.
632 if (iov_iter_count(iter) > 0)
635 goto read_phase_complete;
644 case RXRPC_CALL_COMPLETE:
655 rxrpc_transmit_ack_packets(call->peer->local);
657 *_service = call->service_id;
658 mutex_unlock(&call->user_mutex);
659 _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
663 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
667 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
671 *_abort = call->abort_code;
673 if (call->completion == RXRPC_CALL_SUCCEEDED) {
675 if (iov_iter_count(iter) > 0)
680 EXPORT_SYMBOL(rxrpc_kernel_recv_data);