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(&call->notify_lock);
40 call->notify_rx(sk, call, call->user_call_ID);
41 spin_unlock(&call->notify_lock);
43 write_lock(&rx->recvmsg_lock);
44 if (list_empty(&call->recvmsg_link)) {
45 rxrpc_get_call(call, rxrpc_call_get_notify_socket);
46 list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
48 write_unlock(&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(&call->state_lock);
91 ret = __rxrpc_set_call_completion(call, compl, abort_code, error);
92 write_unlock(&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(&call->state_lock);
111 ret = __rxrpc_call_completed(call);
112 write_unlock(&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(&call->state_lock);
135 ret = __rxrpc_abort_call(why, call, seq, abort_code, error);
136 write_unlock(&call->state_lock);
137 if (ret && test_bit(RXRPC_CALL_EXPOSED, &call->flags))
138 rxrpc_send_abort_packet(call);
143 * Pass a call terminating message to userspace.
145 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
150 switch (call->completion) {
151 case RXRPC_CALL_SUCCEEDED:
153 if (rxrpc_is_service_call(call))
154 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
156 case RXRPC_CALL_REMOTELY_ABORTED:
157 tmp = call->abort_code;
158 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
160 case RXRPC_CALL_LOCALLY_ABORTED:
161 tmp = call->abort_code;
162 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
164 case RXRPC_CALL_NETWORK_ERROR:
166 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
168 case RXRPC_CALL_LOCAL_ERROR:
170 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
173 pr_err("Invalid terminal call state %u\n", call->state);
178 trace_rxrpc_recvdata(call, rxrpc_recvmsg_terminal,
179 lower_32_bits(atomic64_read(&call->ackr_window)) - 1,
180 call->rx_pkt_offset, call->rx_pkt_len, ret);
185 * End the packet reception phase.
187 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
189 rxrpc_seq_t whigh = READ_ONCE(call->rx_highest_seq);
191 _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
193 trace_rxrpc_receive(call, rxrpc_receive_end, 0, whigh);
195 if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY)
196 rxrpc_propose_delay_ACK(call, serial, rxrpc_propose_ack_terminal_ack);
198 write_lock(&call->state_lock);
200 switch (call->state) {
201 case RXRPC_CALL_CLIENT_RECV_REPLY:
202 __rxrpc_call_completed(call);
203 write_unlock(&call->state_lock);
204 rxrpc_poke_call(call, rxrpc_call_poke_complete);
207 case RXRPC_CALL_SERVER_RECV_REQUEST:
208 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
209 call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
210 write_unlock(&call->state_lock);
211 rxrpc_propose_delay_ACK(call, serial,
212 rxrpc_propose_ack_processing_op);
215 write_unlock(&call->state_lock);
221 * Discard a packet we've used up and advance the Rx window by one.
223 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
225 struct rxrpc_skb_priv *sp;
227 rxrpc_serial_t serial;
228 rxrpc_seq_t old_consumed = call->rx_consumed, tseq;
232 _enter("%d", call->debug_id);
234 skb = skb_dequeue(&call->recvmsg_queue);
235 rxrpc_see_skb(skb, rxrpc_skb_see_rotate);
239 serial = sp->hdr.serial;
240 last = sp->hdr.flags & RXRPC_LAST_PACKET;
242 /* Barrier against rxrpc_input_data(). */
243 if (after(tseq, call->rx_consumed))
244 smp_store_release(&call->rx_consumed, tseq);
246 rxrpc_free_skb(skb, rxrpc_skb_put_rotate);
248 trace_rxrpc_receive(call, last ? rxrpc_receive_rotate_last : rxrpc_receive_rotate,
249 serial, call->rx_consumed);
251 rxrpc_end_rx_phase(call, serial);
255 /* Check to see if there's an ACK that needs sending. */
256 acked = atomic_add_return(call->rx_consumed - old_consumed,
257 &call->ackr_nr_consumed);
259 !test_and_set_bit(RXRPC_CALL_RX_IS_IDLE, &call->flags))
260 rxrpc_poke_call(call, rxrpc_call_poke_idle);
264 * Decrypt and verify a DATA packet.
266 static int rxrpc_verify_data(struct rxrpc_call *call, struct sk_buff *skb)
268 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
270 if (sp->flags & RXRPC_RX_VERIFIED)
272 return call->security->verify_packet(call, skb);
276 * Deliver messages to a call. This keeps processing packets until the buffer
277 * is filled and we find either more DATA (returns 0) or the end of the DATA
278 * (returns 1). If more packets are required, it returns -EAGAIN.
280 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
281 struct msghdr *msg, struct iov_iter *iter,
282 size_t len, int flags, size_t *_offset)
284 struct rxrpc_skb_priv *sp;
288 unsigned int rx_pkt_offset, rx_pkt_len;
289 int copy, ret = -EAGAIN, ret2;
291 rx_pkt_offset = call->rx_pkt_offset;
292 rx_pkt_len = call->rx_pkt_len;
294 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
295 seq = lower_32_bits(atomic64_read(&call->ackr_window)) - 1;
300 /* No one else can be removing stuff from the queue, so we shouldn't
301 * need the Rx lock to walk it.
303 skb = skb_peek(&call->recvmsg_queue);
305 rxrpc_see_skb(skb, rxrpc_skb_see_recvmsg);
309 if (!(flags & MSG_PEEK))
310 trace_rxrpc_receive(call, rxrpc_receive_front,
311 sp->hdr.serial, seq);
314 sock_recv_timestamp(msg, sock->sk, skb);
316 if (rx_pkt_offset == 0) {
317 ret2 = rxrpc_verify_data(call, skb);
318 rx_pkt_offset = sp->offset;
319 rx_pkt_len = sp->len;
320 trace_rxrpc_recvdata(call, rxrpc_recvmsg_next, seq,
321 rx_pkt_offset, rx_pkt_len, ret2);
327 trace_rxrpc_recvdata(call, rxrpc_recvmsg_cont, seq,
328 rx_pkt_offset, rx_pkt_len, 0);
331 /* We have to handle short, empty and used-up DATA packets. */
332 remain = len - *_offset;
337 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
344 /* handle piecemeal consumption of data packets */
345 rx_pkt_offset += copy;
350 if (rx_pkt_len > 0) {
351 trace_rxrpc_recvdata(call, rxrpc_recvmsg_full, seq,
352 rx_pkt_offset, rx_pkt_len, 0);
353 ASSERTCMP(*_offset, ==, len);
358 /* The whole packet has been transferred. */
359 if (sp->hdr.flags & RXRPC_LAST_PACKET)
364 skb = skb_peek_next(skb, &call->recvmsg_queue);
366 if (!(flags & MSG_PEEK))
367 rxrpc_rotate_rx_window(call);
371 if (!(flags & MSG_PEEK)) {
372 call->rx_pkt_offset = rx_pkt_offset;
373 call->rx_pkt_len = rx_pkt_len;
376 trace_rxrpc_recvdata(call, rxrpc_recvmsg_data_return, seq,
377 rx_pkt_offset, rx_pkt_len, ret);
379 set_bit(RXRPC_CALL_RX_IS_IDLE, &call->flags);
384 * Receive a message from an RxRPC socket
385 * - we need to be careful about two or more threads calling recvmsg
388 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
391 struct rxrpc_call *call;
392 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
394 unsigned int call_debug_id = 0;
401 trace_rxrpc_recvmsg(0, rxrpc_recvmsg_enter, 0);
403 if (flags & (MSG_OOB | MSG_TRUNC))
406 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
411 /* Return immediately if a client socket has no outstanding calls */
412 if (RB_EMPTY_ROOT(&rx->calls) &&
413 list_empty(&rx->recvmsg_q) &&
414 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
415 release_sock(&rx->sk);
419 if (list_empty(&rx->recvmsg_q)) {
426 release_sock(&rx->sk);
428 /* Wait for something to happen */
429 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
431 ret = sock_error(&rx->sk);
435 if (list_empty(&rx->recvmsg_q)) {
436 if (signal_pending(current))
437 goto wait_interrupted;
438 trace_rxrpc_recvmsg(0, rxrpc_recvmsg_wait, 0);
439 timeo = schedule_timeout(timeo);
441 finish_wait(sk_sleep(&rx->sk), &wait);
445 /* Find the next call and dequeue it if we're not just peeking. If we
446 * do dequeue it, that comes with a ref that we will need to release.
448 write_lock(&rx->recvmsg_lock);
449 l = rx->recvmsg_q.next;
450 call = list_entry(l, struct rxrpc_call, recvmsg_link);
451 if (!(flags & MSG_PEEK))
452 list_del_init(&call->recvmsg_link);
454 rxrpc_get_call(call, rxrpc_call_get_recvmsg);
455 write_unlock(&rx->recvmsg_lock);
457 call_debug_id = call->debug_id;
458 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_dequeue, 0);
460 /* We're going to drop the socket lock, so we need to lock the call
461 * against interference by sendmsg.
463 if (!mutex_trylock(&call->user_mutex)) {
465 if (flags & MSG_DONTWAIT)
466 goto error_requeue_call;
468 if (mutex_lock_interruptible(&call->user_mutex) < 0)
469 goto error_requeue_call;
472 release_sock(&rx->sk);
474 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
477 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
478 if (flags & MSG_CMSG_COMPAT) {
479 unsigned int id32 = call->user_call_ID;
481 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
482 sizeof(unsigned int), &id32);
484 unsigned long idl = call->user_call_ID;
486 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
487 sizeof(unsigned long), &idl);
490 goto error_unlock_call;
493 if (msg->msg_name && call->peer) {
494 size_t len = sizeof(call->dest_srx);
496 memcpy(msg->msg_name, &call->dest_srx, len);
497 msg->msg_namelen = len;
500 switch (READ_ONCE(call->state)) {
501 case RXRPC_CALL_CLIENT_RECV_REPLY:
502 case RXRPC_CALL_SERVER_RECV_REQUEST:
503 case RXRPC_CALL_SERVER_ACK_REQUEST:
504 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
509 if (!skb_queue_empty(&call->recvmsg_queue))
510 rxrpc_notify_socket(call);
518 goto error_unlock_call;
520 if (call->state == RXRPC_CALL_COMPLETE) {
521 ret = rxrpc_recvmsg_term(call, msg);
523 goto error_unlock_call;
524 if (!(flags & MSG_PEEK))
525 rxrpc_release_call(rx, call);
526 msg->msg_flags |= MSG_EOR;
531 msg->msg_flags |= MSG_MORE;
533 msg->msg_flags &= ~MSG_MORE;
537 mutex_unlock(&call->user_mutex);
538 rxrpc_put_call(call, rxrpc_call_put_recvmsg);
539 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_return, ret);
543 if (!(flags & MSG_PEEK)) {
544 write_lock(&rx->recvmsg_lock);
545 list_add(&call->recvmsg_link, &rx->recvmsg_q);
546 write_unlock(&rx->recvmsg_lock);
547 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_requeue, 0);
549 rxrpc_put_call(call, rxrpc_call_put_recvmsg);
552 release_sock(&rx->sk);
554 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_return, ret);
558 ret = sock_intr_errno(timeo);
560 finish_wait(sk_sleep(&rx->sk), &wait);
566 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
567 * @sock: The socket that the call exists on
568 * @call: The call to send data through
569 * @iter: The buffer to receive into
570 * @_len: The amount of data we want to receive (decreased on return)
571 * @want_more: True if more data is expected to be read
572 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
573 * @_service: Where to store the actual service ID (may be upgraded)
575 * Allow a kernel service to receive data and pick up information about the
576 * state of a call. Returns 0 if got what was asked for and there's more
577 * available, 1 if we got what was asked for and we're at the end of the data
578 * and -EAGAIN if we need more data.
580 * Note that we may return -EAGAIN to drain empty packets at the end of the
581 * data, even if we've already copied over the requested data.
583 * *_abort should also be initialised to 0.
585 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
586 struct iov_iter *iter, size_t *_len,
587 bool want_more, u32 *_abort, u16 *_service)
592 _enter("{%d,%s},%zu,%d",
593 call->debug_id, rxrpc_call_states[call->state],
596 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_SECURING);
598 mutex_lock(&call->user_mutex);
600 switch (READ_ONCE(call->state)) {
601 case RXRPC_CALL_CLIENT_RECV_REPLY:
602 case RXRPC_CALL_SERVER_RECV_REQUEST:
603 case RXRPC_CALL_SERVER_ACK_REQUEST:
604 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
610 /* We can only reach here with a partially full buffer if we
611 * have reached the end of the data. We must otherwise have a
612 * full buffer or have been given -EAGAIN.
615 if (iov_iter_count(iter) > 0)
618 goto read_phase_complete;
627 case RXRPC_CALL_COMPLETE:
639 *_service = call->dest_srx.srx_service;
640 mutex_unlock(&call->user_mutex);
641 _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
645 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
649 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
653 *_abort = call->abort_code;
655 if (call->completion == RXRPC_CALL_SUCCEEDED) {
657 if (iov_iter_count(iter) > 0)
662 EXPORT_SYMBOL(rxrpc_kernel_recv_data);