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 spin_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 spin_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 * Pass a call terminating message to userspace.
64 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
69 switch (call->completion) {
70 case RXRPC_CALL_SUCCEEDED:
72 if (rxrpc_is_service_call(call))
73 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
75 case RXRPC_CALL_REMOTELY_ABORTED:
76 tmp = call->abort_code;
77 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
79 case RXRPC_CALL_LOCALLY_ABORTED:
80 tmp = call->abort_code;
81 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
83 case RXRPC_CALL_NETWORK_ERROR:
85 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
87 case RXRPC_CALL_LOCAL_ERROR:
89 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
92 pr_err("Invalid terminal call state %u\n", call->completion);
97 trace_rxrpc_recvdata(call, rxrpc_recvmsg_terminal,
98 call->ackr_window - 1,
99 call->rx_pkt_offset, call->rx_pkt_len, ret);
104 * Discard a packet we've used up and advance the Rx window by one.
106 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
108 struct rxrpc_skb_priv *sp;
110 rxrpc_serial_t serial;
111 rxrpc_seq_t old_consumed = call->rx_consumed, tseq;
115 _enter("%d", call->debug_id);
117 skb = skb_dequeue(&call->recvmsg_queue);
118 rxrpc_see_skb(skb, rxrpc_skb_see_rotate);
122 serial = sp->hdr.serial;
123 last = sp->hdr.flags & RXRPC_LAST_PACKET;
125 /* Barrier against rxrpc_input_data(). */
126 if (after(tseq, call->rx_consumed))
127 smp_store_release(&call->rx_consumed, tseq);
129 rxrpc_free_skb(skb, rxrpc_skb_put_rotate);
131 trace_rxrpc_receive(call, last ? rxrpc_receive_rotate_last : rxrpc_receive_rotate,
132 serial, call->rx_consumed);
135 set_bit(RXRPC_CALL_RECVMSG_READ_ALL, &call->flags);
137 /* Check to see if there's an ACK that needs sending. */
138 acked = atomic_add_return(call->rx_consumed - old_consumed,
139 &call->ackr_nr_consumed);
141 !test_and_set_bit(RXRPC_CALL_RX_IS_IDLE, &call->flags))
142 rxrpc_poke_call(call, rxrpc_call_poke_idle);
146 * Decrypt and verify a DATA packet.
148 static int rxrpc_verify_data(struct rxrpc_call *call, struct sk_buff *skb)
150 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
152 if (sp->flags & RXRPC_RX_VERIFIED)
154 return call->security->verify_packet(call, skb);
158 * Deliver messages to a call. This keeps processing packets until the buffer
159 * is filled and we find either more DATA (returns 0) or the end of the DATA
160 * (returns 1). If more packets are required, it returns -EAGAIN and if the
161 * call has failed it returns -EIO.
163 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
164 struct msghdr *msg, struct iov_iter *iter,
165 size_t len, int flags, size_t *_offset)
167 struct rxrpc_skb_priv *sp;
171 unsigned int rx_pkt_offset, rx_pkt_len;
172 int copy, ret = -EAGAIN, ret2;
174 rx_pkt_offset = call->rx_pkt_offset;
175 rx_pkt_len = call->rx_pkt_len;
177 if (rxrpc_call_has_failed(call)) {
178 seq = call->ackr_window - 1;
183 if (test_bit(RXRPC_CALL_RECVMSG_READ_ALL, &call->flags)) {
184 seq = call->ackr_window - 1;
189 /* No one else can be removing stuff from the queue, so we shouldn't
190 * need the Rx lock to walk it.
192 skb = skb_peek(&call->recvmsg_queue);
194 rxrpc_see_skb(skb, rxrpc_skb_see_recvmsg);
198 if (!(flags & MSG_PEEK))
199 trace_rxrpc_receive(call, rxrpc_receive_front,
200 sp->hdr.serial, seq);
203 sock_recv_timestamp(msg, sock->sk, skb);
205 if (rx_pkt_offset == 0) {
206 ret2 = rxrpc_verify_data(call, skb);
207 trace_rxrpc_recvdata(call, rxrpc_recvmsg_next, seq,
208 sp->offset, sp->len, ret2);
210 kdebug("verify = %d", ret2);
214 rx_pkt_offset = sp->offset;
215 rx_pkt_len = sp->len;
217 trace_rxrpc_recvdata(call, rxrpc_recvmsg_cont, seq,
218 rx_pkt_offset, rx_pkt_len, 0);
221 /* We have to handle short, empty and used-up DATA packets. */
222 remain = len - *_offset;
227 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
234 /* handle piecemeal consumption of data packets */
235 rx_pkt_offset += copy;
240 if (rx_pkt_len > 0) {
241 trace_rxrpc_recvdata(call, rxrpc_recvmsg_full, seq,
242 rx_pkt_offset, rx_pkt_len, 0);
243 ASSERTCMP(*_offset, ==, len);
248 /* The whole packet has been transferred. */
249 if (sp->hdr.flags & RXRPC_LAST_PACKET)
254 skb = skb_peek_next(skb, &call->recvmsg_queue);
256 if (!(flags & MSG_PEEK))
257 rxrpc_rotate_rx_window(call);
261 if (!(flags & MSG_PEEK)) {
262 call->rx_pkt_offset = rx_pkt_offset;
263 call->rx_pkt_len = rx_pkt_len;
266 trace_rxrpc_recvdata(call, rxrpc_recvmsg_data_return, seq,
267 rx_pkt_offset, rx_pkt_len, ret);
269 set_bit(RXRPC_CALL_RX_IS_IDLE, &call->flags);
274 * Receive a message from an RxRPC socket
275 * - we need to be careful about two or more threads calling recvmsg
278 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
281 struct rxrpc_call *call;
282 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
284 unsigned int call_debug_id = 0;
291 trace_rxrpc_recvmsg(0, rxrpc_recvmsg_enter, 0);
293 if (flags & (MSG_OOB | MSG_TRUNC))
296 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
301 /* Return immediately if a client socket has no outstanding calls */
302 if (RB_EMPTY_ROOT(&rx->calls) &&
303 list_empty(&rx->recvmsg_q) &&
304 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
305 release_sock(&rx->sk);
309 if (list_empty(&rx->recvmsg_q)) {
316 release_sock(&rx->sk);
318 /* Wait for something to happen */
319 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
321 ret = sock_error(&rx->sk);
325 if (list_empty(&rx->recvmsg_q)) {
326 if (signal_pending(current))
327 goto wait_interrupted;
328 trace_rxrpc_recvmsg(0, rxrpc_recvmsg_wait, 0);
329 timeo = schedule_timeout(timeo);
331 finish_wait(sk_sleep(&rx->sk), &wait);
335 /* Find the next call and dequeue it if we're not just peeking. If we
336 * do dequeue it, that comes with a ref that we will need to release.
338 spin_lock(&rx->recvmsg_lock);
339 l = rx->recvmsg_q.next;
340 call = list_entry(l, struct rxrpc_call, recvmsg_link);
341 if (!(flags & MSG_PEEK))
342 list_del_init(&call->recvmsg_link);
344 rxrpc_get_call(call, rxrpc_call_get_recvmsg);
345 spin_unlock(&rx->recvmsg_lock);
347 call_debug_id = call->debug_id;
348 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_dequeue, 0);
350 /* We're going to drop the socket lock, so we need to lock the call
351 * against interference by sendmsg.
353 if (!mutex_trylock(&call->user_mutex)) {
355 if (flags & MSG_DONTWAIT)
356 goto error_requeue_call;
358 if (mutex_lock_interruptible(&call->user_mutex) < 0)
359 goto error_requeue_call;
362 release_sock(&rx->sk);
364 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
367 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
368 if (flags & MSG_CMSG_COMPAT) {
369 unsigned int id32 = call->user_call_ID;
371 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
372 sizeof(unsigned int), &id32);
374 unsigned long idl = call->user_call_ID;
376 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
377 sizeof(unsigned long), &idl);
380 goto error_unlock_call;
383 if (msg->msg_name && call->peer) {
384 size_t len = sizeof(call->dest_srx);
386 memcpy(msg->msg_name, &call->dest_srx, len);
387 msg->msg_namelen = len;
390 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
397 goto error_unlock_call;
399 if (rxrpc_call_is_complete(call) &&
400 skb_queue_empty(&call->recvmsg_queue))
402 if (rxrpc_call_has_failed(call))
405 rxrpc_notify_socket(call);
406 goto not_yet_complete;
409 rxrpc_purge_queue(&call->recvmsg_queue);
411 ret = rxrpc_recvmsg_term(call, msg);
413 goto error_unlock_call;
414 if (!(flags & MSG_PEEK))
415 rxrpc_release_call(rx, call);
416 msg->msg_flags |= MSG_EOR;
421 msg->msg_flags |= MSG_MORE;
423 msg->msg_flags &= ~MSG_MORE;
427 mutex_unlock(&call->user_mutex);
428 rxrpc_put_call(call, rxrpc_call_put_recvmsg);
429 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_return, ret);
433 if (!(flags & MSG_PEEK)) {
434 spin_lock(&rx->recvmsg_lock);
435 list_add(&call->recvmsg_link, &rx->recvmsg_q);
436 spin_unlock(&rx->recvmsg_lock);
437 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_requeue, 0);
439 rxrpc_put_call(call, rxrpc_call_put_recvmsg);
442 release_sock(&rx->sk);
444 trace_rxrpc_recvmsg(call_debug_id, rxrpc_recvmsg_return, ret);
448 ret = sock_intr_errno(timeo);
450 finish_wait(sk_sleep(&rx->sk), &wait);
456 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
457 * @sock: The socket that the call exists on
458 * @call: The call to send data through
459 * @iter: The buffer to receive into
460 * @_len: The amount of data we want to receive (decreased on return)
461 * @want_more: True if more data is expected to be read
462 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
463 * @_service: Where to store the actual service ID (may be upgraded)
465 * Allow a kernel service to receive data and pick up information about the
466 * state of a call. Returns 0 if got what was asked for and there's more
467 * available, 1 if we got what was asked for and we're at the end of the data
468 * and -EAGAIN if we need more data.
470 * Note that we may return -EAGAIN to drain empty packets at the end of the
471 * data, even if we've already copied over the requested data.
473 * *_abort should also be initialised to 0.
475 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
476 struct iov_iter *iter, size_t *_len,
477 bool want_more, u32 *_abort, u16 *_service)
482 _enter("{%d},%zu,%d", call->debug_id, *_len, want_more);
484 mutex_lock(&call->user_mutex);
486 ret = rxrpc_recvmsg_data(sock, call, NULL, iter, *_len, 0, &offset);
493 /* We can only reach here with a partially full buffer if we have
494 * reached the end of the data. We must otherwise have a full buffer
495 * or have been given -EAGAIN.
498 if (iov_iter_count(iter) > 0)
501 goto read_phase_complete;
514 *_service = call->dest_srx.srx_service;
515 mutex_unlock(&call->user_mutex);
516 _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
520 trace_rxrpc_abort(call->debug_id, rxrpc_recvmsg_short_data,
521 call->cid, call->call_id, call->rx_consumed,
526 trace_rxrpc_abort(call->debug_id, rxrpc_recvmsg_excess_data,
527 call->cid, call->call_id, call->rx_consumed,
532 *_abort = call->abort_code;
534 if (call->completion == RXRPC_CALL_SUCCEEDED) {
536 if (iov_iter_count(iter) > 0)
541 EXPORT_SYMBOL(rxrpc_kernel_recv_data);