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 * 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->state);
97 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
98 call->rx_pkt_offset, call->rx_pkt_len, ret);
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.
109 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
110 struct rxrpc_call *call,
111 struct msghdr *msg, int flags)
115 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
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);
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);
129 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
134 * End the packet reception phase.
136 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
138 _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
140 trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
141 ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
143 if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
144 rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, serial, false, true,
145 rxrpc_propose_ack_terminal_ack);
146 //rxrpc_send_ack_packet(call, false, NULL);
149 write_lock_bh(&call->state_lock);
151 switch (call->state) {
152 case RXRPC_CALL_CLIENT_RECV_REPLY:
153 __rxrpc_call_completed(call);
154 write_unlock_bh(&call->state_lock);
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, serial, false, true,
163 rxrpc_propose_ack_processing_op);
166 write_unlock_bh(&call->state_lock);
172 * Discard a packet we've used up and advance the Rx window by one.
174 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
176 struct rxrpc_skb_priv *sp;
178 rxrpc_serial_t serial;
179 rxrpc_seq_t hard_ack, top;
184 _enter("%d", call->debug_id);
186 hard_ack = call->rx_hard_ack;
187 top = smp_load_acquire(&call->rx_top);
188 ASSERT(before(hard_ack, top));
191 ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
192 skb = call->rxtx_buffer[ix];
193 rxrpc_see_skb(skb, rxrpc_skb_rotated);
196 subpacket = call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
197 serial = sp->hdr.serial + subpacket;
199 if (subpacket == sp->nr_subpackets - 1 &&
200 sp->rx_flags & RXRPC_SKB_INCL_LAST)
203 call->rxtx_buffer[ix] = NULL;
204 call->rxtx_annotations[ix] = 0;
205 /* Barrier against rxrpc_input_data(). */
206 smp_store_release(&call->rx_hard_ack, hard_ack);
208 rxrpc_free_skb(skb, rxrpc_skb_freed);
210 trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
212 rxrpc_end_rx_phase(call, serial);
214 /* Check to see if there's an ACK that needs sending. */
215 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
216 after_eq(top, call->ackr_seen + 2) ||
217 (hard_ack == top && after(hard_ack, call->ackr_consumed)))
218 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial,
220 rxrpc_propose_ack_rotate_rx);
221 if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
222 rxrpc_send_ack_packet(call, false, NULL);
227 * Decrypt and verify a (sub)packet. The packet's length may be changed due to
228 * padding, but if this is the case, the packet length will be resident in the
229 * socket buffer. Note that we can't modify the master skb info as the skb may
230 * be the home to multiple subpackets.
232 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
234 unsigned int offset, unsigned int len)
236 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
237 rxrpc_seq_t seq = sp->hdr.seq;
238 u16 cksum = sp->hdr.cksum;
239 u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
243 /* For all but the head jumbo subpacket, the security checksum is in a
244 * jumbo header immediately prior to the data.
248 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
254 return call->conn->security->verify_packet(call, skb, offset, len,
259 * Locate the data within a packet. This is complicated by:
261 * (1) An skb may contain a jumbo packet - so we have to find the appropriate
264 * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
265 * contains an extra header which includes the true length of the data,
266 * excluding any encrypted padding.
268 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
270 unsigned int *_offset, unsigned int *_len)
272 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
273 unsigned int offset = sizeof(struct rxrpc_wire_header);
276 u8 annotation = *_annotation;
277 u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
279 /* Locate the subpacket */
280 offset += subpacket * RXRPC_JUMBO_SUBPKTLEN;
281 len = skb->len - offset;
282 if (subpacket < sp->nr_subpackets - 1)
283 len = RXRPC_JUMBO_DATALEN;
285 if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
286 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
289 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
294 call->conn->security->locate_data(call, skb, _offset, _len);
299 * Deliver messages to a call. This keeps processing packets until the buffer
300 * is filled and we find either more DATA (returns 0) or the end of the DATA
301 * (returns 1). If more packets are required, it returns -EAGAIN.
303 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
304 struct msghdr *msg, struct iov_iter *iter,
305 size_t len, int flags, size_t *_offset)
307 struct rxrpc_skb_priv *sp;
309 rxrpc_serial_t serial;
310 rxrpc_seq_t hard_ack, top, seq;
313 unsigned int rx_pkt_offset, rx_pkt_len;
314 int ix, copy, ret = -EAGAIN, ret2;
316 if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
318 rxrpc_send_ack_packet(call, false, NULL);
320 rx_pkt_offset = call->rx_pkt_offset;
321 rx_pkt_len = call->rx_pkt_len;
323 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
324 seq = call->rx_hard_ack;
329 /* Barriers against rxrpc_input_data(). */
330 hard_ack = call->rx_hard_ack;
332 while (top = smp_load_acquire(&call->rx_top),
335 ix = seq & RXRPC_RXTX_BUFF_MASK;
336 skb = call->rxtx_buffer[ix];
338 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
339 rx_pkt_offset, rx_pkt_len, 0);
343 rxrpc_see_skb(skb, rxrpc_skb_seen);
346 if (!(flags & MSG_PEEK)) {
347 serial = sp->hdr.serial;
348 serial += call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
349 trace_rxrpc_receive(call, rxrpc_receive_front,
354 sock_recv_timestamp(msg, sock->sk, skb);
356 if (rx_pkt_offset == 0) {
357 ret2 = rxrpc_locate_data(call, skb,
358 &call->rxtx_annotations[ix],
359 &rx_pkt_offset, &rx_pkt_len);
360 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
361 rx_pkt_offset, rx_pkt_len, ret2);
367 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
368 rx_pkt_offset, rx_pkt_len, 0);
371 /* We have to handle short, empty and used-up DATA packets. */
372 remain = len - *_offset;
377 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
384 /* handle piecemeal consumption of data packets */
385 rx_pkt_offset += copy;
390 if (rx_pkt_len > 0) {
391 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
392 rx_pkt_offset, rx_pkt_len, 0);
393 ASSERTCMP(*_offset, ==, len);
398 /* The whole packet has been transferred. */
399 last = sp->hdr.flags & RXRPC_LAST_PACKET;
400 if (!(flags & MSG_PEEK))
401 rxrpc_rotate_rx_window(call);
406 ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
415 if (!(flags & MSG_PEEK)) {
416 call->rx_pkt_offset = rx_pkt_offset;
417 call->rx_pkt_len = rx_pkt_len;
420 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
421 rx_pkt_offset, rx_pkt_len, ret);
423 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
428 * Receive a message from an RxRPC socket
429 * - we need to be careful about two or more threads calling recvmsg
432 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
435 struct rxrpc_call *call;
436 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
444 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
446 if (flags & (MSG_OOB | MSG_TRUNC))
449 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
454 /* Return immediately if a client socket has no outstanding calls */
455 if (RB_EMPTY_ROOT(&rx->calls) &&
456 list_empty(&rx->recvmsg_q) &&
457 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
458 release_sock(&rx->sk);
462 if (list_empty(&rx->recvmsg_q)) {
469 release_sock(&rx->sk);
471 /* Wait for something to happen */
472 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
474 ret = sock_error(&rx->sk);
478 if (list_empty(&rx->recvmsg_q)) {
479 if (signal_pending(current))
480 goto wait_interrupted;
481 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
483 timeo = schedule_timeout(timeo);
485 finish_wait(sk_sleep(&rx->sk), &wait);
489 /* Find the next call and dequeue it if we're not just peeking. If we
490 * do dequeue it, that comes with a ref that we will need to release.
492 write_lock_bh(&rx->recvmsg_lock);
493 l = rx->recvmsg_q.next;
494 call = list_entry(l, struct rxrpc_call, recvmsg_link);
495 if (!(flags & MSG_PEEK))
496 list_del_init(&call->recvmsg_link);
498 rxrpc_get_call(call, rxrpc_call_got);
499 write_unlock_bh(&rx->recvmsg_lock);
501 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
503 /* We're going to drop the socket lock, so we need to lock the call
504 * against interference by sendmsg.
506 if (!mutex_trylock(&call->user_mutex)) {
508 if (flags & MSG_DONTWAIT)
509 goto error_requeue_call;
511 if (mutex_lock_interruptible(&call->user_mutex) < 0)
512 goto error_requeue_call;
515 release_sock(&rx->sk);
517 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
520 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
521 if (flags & MSG_CMSG_COMPAT) {
522 unsigned int id32 = call->user_call_ID;
524 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
525 sizeof(unsigned int), &id32);
527 unsigned long idl = call->user_call_ID;
529 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
530 sizeof(unsigned long), &idl);
533 goto error_unlock_call;
537 struct sockaddr_rxrpc *srx = msg->msg_name;
538 size_t len = sizeof(call->peer->srx);
540 memcpy(msg->msg_name, &call->peer->srx, len);
541 srx->srx_service = call->service_id;
542 msg->msg_namelen = len;
545 switch (READ_ONCE(call->state)) {
546 case RXRPC_CALL_SERVER_ACCEPTING:
547 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
549 case RXRPC_CALL_CLIENT_RECV_REPLY:
550 case RXRPC_CALL_SERVER_RECV_REQUEST:
551 case RXRPC_CALL_SERVER_ACK_REQUEST:
552 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
557 if (after(call->rx_top, call->rx_hard_ack) &&
558 call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
559 rxrpc_notify_socket(call);
567 goto error_unlock_call;
569 if (call->state == RXRPC_CALL_COMPLETE) {
570 ret = rxrpc_recvmsg_term(call, msg);
572 goto error_unlock_call;
573 if (!(flags & MSG_PEEK))
574 rxrpc_release_call(rx, call);
575 msg->msg_flags |= MSG_EOR;
580 msg->msg_flags |= MSG_MORE;
582 msg->msg_flags &= ~MSG_MORE;
586 mutex_unlock(&call->user_mutex);
587 rxrpc_put_call(call, rxrpc_call_put);
588 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
592 if (!(flags & MSG_PEEK)) {
593 write_lock_bh(&rx->recvmsg_lock);
594 list_add(&call->recvmsg_link, &rx->recvmsg_q);
595 write_unlock_bh(&rx->recvmsg_lock);
596 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
598 rxrpc_put_call(call, rxrpc_call_put);
601 release_sock(&rx->sk);
603 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
607 ret = sock_intr_errno(timeo);
609 finish_wait(sk_sleep(&rx->sk), &wait);
615 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
616 * @sock: The socket that the call exists on
617 * @call: The call to send data through
618 * @iter: The buffer to receive into
619 * @want_more: True if more data is expected to be read
620 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
621 * @_service: Where to store the actual service ID (may be upgraded)
623 * Allow a kernel service to receive data and pick up information about the
624 * state of a call. Returns 0 if got what was asked for and there's more
625 * available, 1 if we got what was asked for and we're at the end of the data
626 * and -EAGAIN if we need more data.
628 * Note that we may return -EAGAIN to drain empty packets at the end of the
629 * data, even if we've already copied over the requested data.
631 * *_abort should also be initialised to 0.
633 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
634 struct iov_iter *iter,
635 bool want_more, u32 *_abort, u16 *_service)
640 _enter("{%d,%s},%zu,%d",
641 call->debug_id, rxrpc_call_states[call->state],
642 iov_iter_count(iter), want_more);
644 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
646 mutex_lock(&call->user_mutex);
648 switch (READ_ONCE(call->state)) {
649 case RXRPC_CALL_CLIENT_RECV_REPLY:
650 case RXRPC_CALL_SERVER_RECV_REQUEST:
651 case RXRPC_CALL_SERVER_ACK_REQUEST:
652 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
653 iov_iter_count(iter), 0,
658 /* We can only reach here with a partially full buffer if we
659 * have reached the end of the data. We must otherwise have a
660 * full buffer or have been given -EAGAIN.
663 if (iov_iter_count(iter) > 0)
666 goto read_phase_complete;
675 case RXRPC_CALL_COMPLETE:
686 switch (call->ackr_reason) {
689 case RXRPC_ACK_DELAY:
694 rxrpc_send_ack_packet(call, false, NULL);
698 *_service = call->service_id;
699 mutex_unlock(&call->user_mutex);
700 _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
704 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
708 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
712 *_abort = call->abort_code;
714 if (call->completion == RXRPC_CALL_SUCCEEDED) {
716 if (iov_iter_count(iter) > 0)
721 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
724 * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
725 * @sock: The socket that the call exists on
726 * @call: The call to query
727 * @_ts: Where to put the timestamp
729 * Retrieve the timestamp from the first DATA packet of the reply if it is
730 * in the ring. Returns true if successful, false if not.
732 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
736 rxrpc_seq_t hard_ack, top, seq;
737 bool success = false;
739 mutex_lock(&call->user_mutex);
741 if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
744 hard_ack = call->rx_hard_ack;
749 top = smp_load_acquire(&call->rx_top);
753 skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
757 *_ts = skb_get_ktime(skb);
761 mutex_unlock(&call->user_mutex);
764 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);