1 /* RxRPC recvmsg() implementation
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
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.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <linux/net.h>
15 #include <linux/skbuff.h>
16 #include <linux/export.h>
17 #include <linux/sched/signal.h>
20 #include <net/af_rxrpc.h>
21 #include "ar-internal.h"
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.
27 void rxrpc_notify_socket(struct rxrpc_call *call)
29 struct rxrpc_sock *rx;
32 _enter("%d", call->debug_id);
34 if (!list_empty(&call->recvmsg_link))
39 rx = rcu_dereference(call->socket);
41 if (rx && sk->sk_state < RXRPC_CLOSE) {
42 if (call->notify_rx) {
43 call->notify_rx(sk, call, call->user_call_ID);
45 write_lock_bh(&rx->recvmsg_lock);
46 if (list_empty(&call->recvmsg_link)) {
47 rxrpc_get_call(call, rxrpc_call_got);
48 list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
50 write_unlock_bh(&rx->recvmsg_lock);
52 if (!sock_flag(sk, SOCK_DEAD)) {
53 _debug("call %ps", sk->sk_data_ready);
54 sk->sk_data_ready(sk);
64 * Pass a call terminating message to userspace.
66 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
71 switch (call->completion) {
72 case RXRPC_CALL_SUCCEEDED:
74 if (rxrpc_is_service_call(call))
75 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
77 case RXRPC_CALL_REMOTELY_ABORTED:
78 tmp = call->abort_code;
79 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
81 case RXRPC_CALL_LOCALLY_ABORTED:
82 tmp = call->abort_code;
83 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
85 case RXRPC_CALL_NETWORK_ERROR:
87 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
89 case RXRPC_CALL_LOCAL_ERROR:
91 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
94 pr_err("Invalid terminal call state %u\n", call->state);
99 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
100 call->rx_pkt_offset, call->rx_pkt_len, ret);
105 * Pass back notification of a new call. The call is added to the
106 * to-be-accepted list. This means that the next call to be accepted might not
107 * be the last call seen awaiting acceptance, but unless we leave this on the
108 * front of the queue and block all other messages until someone gives us a
109 * user_ID for it, there's not a lot we can do.
111 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
112 struct rxrpc_call *call,
113 struct msghdr *msg, int flags)
117 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
119 if (ret == 0 && !(flags & MSG_PEEK)) {
120 _debug("to be accepted");
121 write_lock_bh(&rx->recvmsg_lock);
122 list_del_init(&call->recvmsg_link);
123 write_unlock_bh(&rx->recvmsg_lock);
125 rxrpc_get_call(call, rxrpc_call_got);
126 write_lock(&rx->call_lock);
127 list_add_tail(&call->accept_link, &rx->to_be_accepted);
128 write_unlock(&rx->call_lock);
131 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
136 * End the packet reception phase.
138 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
140 _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
142 trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
143 ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
145 if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
146 rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, serial, true, false,
147 rxrpc_propose_ack_terminal_ack);
148 rxrpc_send_ack_packet(call, false);
151 write_lock_bh(&call->state_lock);
153 switch (call->state) {
154 case RXRPC_CALL_CLIENT_RECV_REPLY:
155 __rxrpc_call_completed(call);
156 write_unlock_bh(&call->state_lock);
159 case RXRPC_CALL_SERVER_RECV_REQUEST:
160 call->tx_phase = true;
161 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
162 call->ack_at = call->expire_at;
163 write_unlock_bh(&call->state_lock);
164 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial, false, true,
165 rxrpc_propose_ack_processing_op);
168 write_unlock_bh(&call->state_lock);
174 * Discard a packet we've used up and advance the Rx window by one.
176 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
178 struct rxrpc_skb_priv *sp;
180 rxrpc_serial_t serial;
181 rxrpc_seq_t hard_ack, top;
185 _enter("%d", call->debug_id);
187 hard_ack = call->rx_hard_ack;
188 top = smp_load_acquire(&call->rx_top);
189 ASSERT(before(hard_ack, top));
192 ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
193 skb = call->rxtx_buffer[ix];
194 rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
196 flags = sp->hdr.flags;
197 serial = sp->hdr.serial;
198 if (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO)
199 serial += (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO) - 1;
201 call->rxtx_buffer[ix] = NULL;
202 call->rxtx_annotations[ix] = 0;
203 /* Barrier against rxrpc_input_data(). */
204 smp_store_release(&call->rx_hard_ack, hard_ack);
206 rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
208 _debug("%u,%u,%02x", hard_ack, top, flags);
209 trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
210 if (flags & RXRPC_LAST_PACKET) {
211 rxrpc_end_rx_phase(call, serial);
213 /* Check to see if there's an ACK that needs sending. */
214 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
215 after_eq(top, call->ackr_seen + 2) ||
216 (hard_ack == top && after(hard_ack, call->ackr_consumed)))
217 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial,
219 rxrpc_propose_ack_rotate_rx);
220 if (call->ackr_reason)
221 rxrpc_send_ack_packet(call, false);
226 * Decrypt and verify a (sub)packet. The packet's length may be changed due to
227 * padding, but if this is the case, the packet length will be resident in the
228 * socket buffer. Note that we can't modify the master skb info as the skb may
229 * be the home to multiple subpackets.
231 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
233 unsigned int offset, unsigned int len)
235 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
236 rxrpc_seq_t seq = sp->hdr.seq;
237 u16 cksum = sp->hdr.cksum;
241 /* For all but the head jumbo subpacket, the security checksum is in a
242 * jumbo header immediately prior to the data.
244 if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
246 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
249 seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
252 return call->conn->security->verify_packet(call, skb, offset, len,
257 * Locate the data within a packet. This is complicated by:
259 * (1) An skb may contain a jumbo packet - so we have to find the appropriate
262 * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
263 * contains an extra header which includes the true length of the data,
264 * excluding any encrypted padding.
266 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
268 unsigned int *_offset, unsigned int *_len)
270 unsigned int offset = sizeof(struct rxrpc_wire_header);
271 unsigned int len = *_len;
273 u8 annotation = *_annotation;
275 /* Locate the subpacket */
276 len = skb->len - offset;
277 if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
278 offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
279 RXRPC_JUMBO_SUBPKTLEN);
280 len = (annotation & RXRPC_RX_ANNO_JLAST) ?
281 skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
284 if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
285 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
288 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
293 call->conn->security->locate_data(call, skb, _offset, _len);
298 * Deliver messages to a call. This keeps processing packets until the buffer
299 * is filled and we find either more DATA (returns 0) or the end of the DATA
300 * (returns 1). If more packets are required, it returns -EAGAIN.
302 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
303 struct msghdr *msg, struct iov_iter *iter,
304 size_t len, int flags, size_t *_offset)
306 struct rxrpc_skb_priv *sp;
308 rxrpc_seq_t hard_ack, top, seq;
311 unsigned int rx_pkt_offset, rx_pkt_len;
312 int ix, copy, ret = -EAGAIN, ret2;
314 rx_pkt_offset = call->rx_pkt_offset;
315 rx_pkt_len = call->rx_pkt_len;
317 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
318 seq = call->rx_hard_ack;
323 /* Barriers against rxrpc_input_data(). */
324 hard_ack = call->rx_hard_ack;
326 while (top = smp_load_acquire(&call->rx_top),
329 ix = seq & RXRPC_RXTX_BUFF_MASK;
330 skb = call->rxtx_buffer[ix];
332 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
333 rx_pkt_offset, rx_pkt_len, 0);
337 rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
340 if (!(flags & MSG_PEEK))
341 trace_rxrpc_receive(call, rxrpc_receive_front,
342 sp->hdr.serial, seq);
345 sock_recv_timestamp(msg, sock->sk, skb);
347 if (rx_pkt_offset == 0) {
348 ret2 = rxrpc_locate_data(call, skb,
349 &call->rxtx_annotations[ix],
350 &rx_pkt_offset, &rx_pkt_len);
351 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
352 rx_pkt_offset, rx_pkt_len, ret2);
358 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
359 rx_pkt_offset, rx_pkt_len, 0);
362 /* We have to handle short, empty and used-up DATA packets. */
363 remain = len - *_offset;
368 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
375 /* handle piecemeal consumption of data packets */
376 rx_pkt_offset += copy;
381 if (rx_pkt_len > 0) {
382 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
383 rx_pkt_offset, rx_pkt_len, 0);
384 ASSERTCMP(*_offset, ==, len);
389 /* The whole packet has been transferred. */
390 last = sp->hdr.flags & RXRPC_LAST_PACKET;
391 if (!(flags & MSG_PEEK))
392 rxrpc_rotate_rx_window(call);
397 ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
406 if (!(flags & MSG_PEEK)) {
407 call->rx_pkt_offset = rx_pkt_offset;
408 call->rx_pkt_len = rx_pkt_len;
411 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
412 rx_pkt_offset, rx_pkt_len, ret);
417 * Receive a message from an RxRPC socket
418 * - we need to be careful about two or more threads calling recvmsg
421 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
424 struct rxrpc_call *call;
425 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
433 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
435 if (flags & (MSG_OOB | MSG_TRUNC))
438 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
443 /* Return immediately if a client socket has no outstanding calls */
444 if (RB_EMPTY_ROOT(&rx->calls) &&
445 list_empty(&rx->recvmsg_q) &&
446 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
447 release_sock(&rx->sk);
451 if (list_empty(&rx->recvmsg_q)) {
458 release_sock(&rx->sk);
460 /* Wait for something to happen */
461 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
463 ret = sock_error(&rx->sk);
467 if (list_empty(&rx->recvmsg_q)) {
468 if (signal_pending(current))
469 goto wait_interrupted;
470 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
472 timeo = schedule_timeout(timeo);
474 finish_wait(sk_sleep(&rx->sk), &wait);
478 /* Find the next call and dequeue it if we're not just peeking. If we
479 * do dequeue it, that comes with a ref that we will need to release.
481 write_lock_bh(&rx->recvmsg_lock);
482 l = rx->recvmsg_q.next;
483 call = list_entry(l, struct rxrpc_call, recvmsg_link);
484 if (!(flags & MSG_PEEK))
485 list_del_init(&call->recvmsg_link);
487 rxrpc_get_call(call, rxrpc_call_got);
488 write_unlock_bh(&rx->recvmsg_lock);
490 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
492 /* We're going to drop the socket lock, so we need to lock the call
493 * against interference by sendmsg.
495 if (!mutex_trylock(&call->user_mutex)) {
497 if (flags & MSG_DONTWAIT)
498 goto error_requeue_call;
500 if (mutex_lock_interruptible(&call->user_mutex) < 0)
501 goto error_requeue_call;
504 release_sock(&rx->sk);
506 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
509 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
510 if (flags & MSG_CMSG_COMPAT) {
511 unsigned int id32 = call->user_call_ID;
513 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
514 sizeof(unsigned int), &id32);
516 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
517 sizeof(unsigned long),
518 &call->user_call_ID);
521 goto error_unlock_call;
525 size_t len = sizeof(call->conn->params.peer->srx);
526 memcpy(msg->msg_name, &call->conn->params.peer->srx, len);
527 msg->msg_namelen = len;
530 switch (READ_ONCE(call->state)) {
531 case RXRPC_CALL_SERVER_ACCEPTING:
532 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
534 case RXRPC_CALL_CLIENT_RECV_REPLY:
535 case RXRPC_CALL_SERVER_RECV_REQUEST:
536 case RXRPC_CALL_SERVER_ACK_REQUEST:
537 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
542 if (after(call->rx_top, call->rx_hard_ack) &&
543 call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
544 rxrpc_notify_socket(call);
552 goto error_unlock_call;
554 if (call->state == RXRPC_CALL_COMPLETE) {
555 ret = rxrpc_recvmsg_term(call, msg);
557 goto error_unlock_call;
558 if (!(flags & MSG_PEEK))
559 rxrpc_release_call(rx, call);
560 msg->msg_flags |= MSG_EOR;
565 msg->msg_flags |= MSG_MORE;
567 msg->msg_flags &= ~MSG_MORE;
571 mutex_unlock(&call->user_mutex);
572 rxrpc_put_call(call, rxrpc_call_put);
573 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
577 if (!(flags & MSG_PEEK)) {
578 write_lock_bh(&rx->recvmsg_lock);
579 list_add(&call->recvmsg_link, &rx->recvmsg_q);
580 write_unlock_bh(&rx->recvmsg_lock);
581 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
583 rxrpc_put_call(call, rxrpc_call_put);
586 release_sock(&rx->sk);
587 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
591 ret = sock_intr_errno(timeo);
593 finish_wait(sk_sleep(&rx->sk), &wait);
599 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
600 * @sock: The socket that the call exists on
601 * @call: The call to send data through
602 * @buf: The buffer to receive into
603 * @size: The size of the buffer, including data already read
604 * @_offset: The running offset into the buffer.
605 * @want_more: True if more data is expected to be read
606 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
608 * Allow a kernel service to receive data and pick up information about the
609 * state of a call. Returns 0 if got what was asked for and there's more
610 * available, 1 if we got what was asked for and we're at the end of the data
611 * and -EAGAIN if we need more data.
613 * Note that we may return -EAGAIN to drain empty packets at the end of the
614 * data, even if we've already copied over the requested data.
616 * This function adds the amount it transfers to *_offset, so this should be
617 * precleared as appropriate. Note that the amount remaining in the buffer is
618 * taken to be size - *_offset.
620 * *_abort should also be initialised to 0.
622 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
623 void *buf, size_t size, size_t *_offset,
624 bool want_more, u32 *_abort)
626 struct iov_iter iter;
630 _enter("{%d,%s},%zu/%zu,%d",
631 call->debug_id, rxrpc_call_states[call->state],
632 *_offset, size, want_more);
634 ASSERTCMP(*_offset, <=, size);
635 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
637 iov.iov_base = buf + *_offset;
638 iov.iov_len = size - *_offset;
639 iov_iter_kvec(&iter, ITER_KVEC | READ, &iov, 1, size - *_offset);
641 mutex_lock(&call->user_mutex);
643 switch (READ_ONCE(call->state)) {
644 case RXRPC_CALL_CLIENT_RECV_REPLY:
645 case RXRPC_CALL_SERVER_RECV_REQUEST:
646 case RXRPC_CALL_SERVER_ACK_REQUEST:
647 ret = rxrpc_recvmsg_data(sock, call, NULL, &iter, size, 0,
652 /* We can only reach here with a partially full buffer if we
653 * have reached the end of the data. We must otherwise have a
654 * full buffer or have been given -EAGAIN.
660 goto read_phase_complete;
669 case RXRPC_CALL_COMPLETE:
680 mutex_unlock(&call->user_mutex);
681 _leave(" = %d [%zu,%d]", ret, *_offset, *_abort);
691 *_abort = call->abort_code;
693 if (call->completion == RXRPC_CALL_SUCCEEDED) {
700 EXPORT_SYMBOL(rxrpc_kernel_recv_data);