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_recvmsg(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_ACK(call, RXRPC_ACK_IDLE, serial, false, true,
193 rxrpc_propose_ack_terminal_ack);
194 //rxrpc_send_ack_packet(call, false, NULL);
197 write_lock_bh(&call->state_lock);
199 switch (call->state) {
200 case RXRPC_CALL_CLIENT_RECV_REPLY:
201 __rxrpc_call_completed(call);
202 write_unlock_bh(&call->state_lock);
205 case RXRPC_CALL_SERVER_RECV_REQUEST:
206 call->tx_phase = true;
207 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
208 call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
209 write_unlock_bh(&call->state_lock);
210 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial, false, true,
211 rxrpc_propose_ack_processing_op);
214 write_unlock_bh(&call->state_lock);
220 * Discard a packet we've used up and advance the Rx window by one.
222 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
224 struct rxrpc_skb_priv *sp;
226 rxrpc_serial_t serial;
227 rxrpc_seq_t hard_ack, top;
232 _enter("%d", call->debug_id);
234 hard_ack = call->rx_hard_ack;
235 top = smp_load_acquire(&call->rx_top);
236 ASSERT(before(hard_ack, top));
239 ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
240 skb = call->rxtx_buffer[ix];
241 rxrpc_see_skb(skb, rxrpc_skb_rotated);
244 subpacket = call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
245 serial = sp->hdr.serial + subpacket;
247 if (subpacket == sp->nr_subpackets - 1 &&
248 sp->rx_flags & RXRPC_SKB_INCL_LAST)
251 call->rxtx_buffer[ix] = NULL;
252 call->rxtx_annotations[ix] = 0;
253 /* Barrier against rxrpc_input_data(). */
254 smp_store_release(&call->rx_hard_ack, hard_ack);
256 rxrpc_free_skb(skb, rxrpc_skb_freed);
258 trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
260 rxrpc_end_rx_phase(call, serial);
262 /* Check to see if there's an ACK that needs sending. */
263 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
264 after_eq(top, call->ackr_seen + 2) ||
265 (hard_ack == top && after(hard_ack, call->ackr_consumed)))
266 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial,
268 rxrpc_propose_ack_rotate_rx);
269 if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
270 rxrpc_send_ack_packet(call, false, NULL);
275 * Decrypt and verify a (sub)packet. The packet's length may be changed due to
276 * padding, but if this is the case, the packet length will be resident in the
277 * socket buffer. Note that we can't modify the master skb info as the skb may
278 * be the home to multiple subpackets.
280 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
282 unsigned int offset, unsigned int len)
284 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
285 rxrpc_seq_t seq = sp->hdr.seq;
286 u16 cksum = sp->hdr.cksum;
287 u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
291 /* For all but the head jumbo subpacket, the security checksum is in a
292 * jumbo header immediately prior to the data.
296 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
302 return call->security->verify_packet(call, skb, offset, len,
307 * Locate the data within a packet. This is complicated by:
309 * (1) An skb may contain a jumbo packet - so we have to find the appropriate
312 * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
313 * contains an extra header which includes the true length of the data,
314 * excluding any encrypted padding.
316 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
318 unsigned int *_offset, unsigned int *_len,
321 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
322 unsigned int offset = sizeof(struct rxrpc_wire_header);
326 u8 annotation = *_annotation;
327 u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
329 /* Locate the subpacket */
330 offset += subpacket * RXRPC_JUMBO_SUBPKTLEN;
331 len = skb->len - offset;
332 if (subpacket < sp->nr_subpackets - 1)
333 len = RXRPC_JUMBO_DATALEN;
334 else if (sp->rx_flags & RXRPC_SKB_INCL_LAST)
337 if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
338 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
341 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
347 call->security->locate_data(call, skb, _offset, _len);
352 * Deliver messages to a call. This keeps processing packets until the buffer
353 * is filled and we find either more DATA (returns 0) or the end of the DATA
354 * (returns 1). If more packets are required, it returns -EAGAIN.
356 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
357 struct msghdr *msg, struct iov_iter *iter,
358 size_t len, int flags, size_t *_offset)
360 struct rxrpc_skb_priv *sp;
362 rxrpc_serial_t serial;
363 rxrpc_seq_t hard_ack, top, seq;
366 unsigned int rx_pkt_offset, rx_pkt_len;
367 int ix, copy, ret = -EAGAIN, ret2;
369 if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
371 rxrpc_send_ack_packet(call, false, NULL);
373 rx_pkt_offset = call->rx_pkt_offset;
374 rx_pkt_len = call->rx_pkt_len;
375 rx_pkt_last = call->rx_pkt_last;
377 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
378 seq = call->rx_hard_ack;
383 /* Barriers against rxrpc_input_data(). */
384 hard_ack = call->rx_hard_ack;
387 while (top = smp_load_acquire(&call->rx_top),
390 ix = seq & RXRPC_RXTX_BUFF_MASK;
391 skb = call->rxtx_buffer[ix];
393 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
394 rx_pkt_offset, rx_pkt_len, 0);
398 rxrpc_see_skb(skb, rxrpc_skb_seen);
401 if (!(flags & MSG_PEEK)) {
402 serial = sp->hdr.serial;
403 serial += call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
404 trace_rxrpc_receive(call, rxrpc_receive_front,
409 sock_recv_timestamp(msg, sock->sk, skb);
411 if (rx_pkt_offset == 0) {
412 ret2 = rxrpc_locate_data(call, skb,
413 &call->rxtx_annotations[ix],
414 &rx_pkt_offset, &rx_pkt_len,
416 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
417 rx_pkt_offset, rx_pkt_len, ret2);
423 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
424 rx_pkt_offset, rx_pkt_len, 0);
427 /* We have to handle short, empty and used-up DATA packets. */
428 remain = len - *_offset;
433 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
440 /* handle piecemeal consumption of data packets */
441 rx_pkt_offset += copy;
446 if (rx_pkt_len > 0) {
447 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
448 rx_pkt_offset, rx_pkt_len, 0);
449 ASSERTCMP(*_offset, ==, len);
454 /* The whole packet has been transferred. */
455 if (!(flags & MSG_PEEK))
456 rxrpc_rotate_rx_window(call);
461 ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
470 if (!(flags & MSG_PEEK)) {
471 call->rx_pkt_offset = rx_pkt_offset;
472 call->rx_pkt_len = rx_pkt_len;
473 call->rx_pkt_last = rx_pkt_last;
476 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
477 rx_pkt_offset, rx_pkt_len, ret);
479 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
484 * Receive a message from an RxRPC socket
485 * - we need to be careful about two or more threads calling recvmsg
488 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
491 struct rxrpc_call *call;
492 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
500 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
502 if (flags & (MSG_OOB | MSG_TRUNC))
505 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
510 /* Return immediately if a client socket has no outstanding calls */
511 if (RB_EMPTY_ROOT(&rx->calls) &&
512 list_empty(&rx->recvmsg_q) &&
513 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
514 release_sock(&rx->sk);
518 if (list_empty(&rx->recvmsg_q)) {
525 release_sock(&rx->sk);
527 /* Wait for something to happen */
528 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
530 ret = sock_error(&rx->sk);
534 if (list_empty(&rx->recvmsg_q)) {
535 if (signal_pending(current))
536 goto wait_interrupted;
537 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
539 timeo = schedule_timeout(timeo);
541 finish_wait(sk_sleep(&rx->sk), &wait);
545 /* Find the next call and dequeue it if we're not just peeking. If we
546 * do dequeue it, that comes with a ref that we will need to release.
548 write_lock_bh(&rx->recvmsg_lock);
549 l = rx->recvmsg_q.next;
550 call = list_entry(l, struct rxrpc_call, recvmsg_link);
551 if (!(flags & MSG_PEEK))
552 list_del_init(&call->recvmsg_link);
554 rxrpc_get_call(call, rxrpc_call_got);
555 write_unlock_bh(&rx->recvmsg_lock);
557 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
559 /* We're going to drop the socket lock, so we need to lock the call
560 * against interference by sendmsg.
562 if (!mutex_trylock(&call->user_mutex)) {
564 if (flags & MSG_DONTWAIT)
565 goto error_requeue_call;
567 if (mutex_lock_interruptible(&call->user_mutex) < 0)
568 goto error_requeue_call;
571 release_sock(&rx->sk);
573 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
576 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
577 if (flags & MSG_CMSG_COMPAT) {
578 unsigned int id32 = call->user_call_ID;
580 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
581 sizeof(unsigned int), &id32);
583 unsigned long idl = call->user_call_ID;
585 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
586 sizeof(unsigned long), &idl);
589 goto error_unlock_call;
592 if (msg->msg_name && call->peer) {
593 struct sockaddr_rxrpc *srx = msg->msg_name;
594 size_t len = sizeof(call->peer->srx);
596 memcpy(msg->msg_name, &call->peer->srx, len);
597 srx->srx_service = call->service_id;
598 msg->msg_namelen = len;
601 switch (READ_ONCE(call->state)) {
602 case RXRPC_CALL_CLIENT_RECV_REPLY:
603 case RXRPC_CALL_SERVER_RECV_REQUEST:
604 case RXRPC_CALL_SERVER_ACK_REQUEST:
605 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
610 if (after(call->rx_top, call->rx_hard_ack) &&
611 call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
612 rxrpc_notify_socket(call);
620 goto error_unlock_call;
622 if (call->state == RXRPC_CALL_COMPLETE) {
623 ret = rxrpc_recvmsg_term(call, msg);
625 goto error_unlock_call;
626 if (!(flags & MSG_PEEK))
627 rxrpc_release_call(rx, call);
628 msg->msg_flags |= MSG_EOR;
633 msg->msg_flags |= MSG_MORE;
635 msg->msg_flags &= ~MSG_MORE;
639 mutex_unlock(&call->user_mutex);
640 rxrpc_put_call(call, rxrpc_call_put);
641 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
645 if (!(flags & MSG_PEEK)) {
646 write_lock_bh(&rx->recvmsg_lock);
647 list_add(&call->recvmsg_link, &rx->recvmsg_q);
648 write_unlock_bh(&rx->recvmsg_lock);
649 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
651 rxrpc_put_call(call, rxrpc_call_put);
654 release_sock(&rx->sk);
656 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
660 ret = sock_intr_errno(timeo);
662 finish_wait(sk_sleep(&rx->sk), &wait);
668 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
669 * @sock: The socket that the call exists on
670 * @call: The call to send data through
671 * @iter: The buffer to receive into
672 * @_len: The amount of data we want to receive (decreased on return)
673 * @want_more: True if more data is expected to be read
674 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
675 * @_service: Where to store the actual service ID (may be upgraded)
677 * Allow a kernel service to receive data and pick up information about the
678 * state of a call. Returns 0 if got what was asked for and there's more
679 * available, 1 if we got what was asked for and we're at the end of the data
680 * and -EAGAIN if we need more data.
682 * Note that we may return -EAGAIN to drain empty packets at the end of the
683 * data, even if we've already copied over the requested data.
685 * *_abort should also be initialised to 0.
687 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
688 struct iov_iter *iter, size_t *_len,
689 bool want_more, u32 *_abort, u16 *_service)
694 _enter("{%d,%s},%zu,%d",
695 call->debug_id, rxrpc_call_states[call->state],
698 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_SECURING);
700 mutex_lock(&call->user_mutex);
702 switch (READ_ONCE(call->state)) {
703 case RXRPC_CALL_CLIENT_RECV_REPLY:
704 case RXRPC_CALL_SERVER_RECV_REQUEST:
705 case RXRPC_CALL_SERVER_ACK_REQUEST:
706 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
712 /* We can only reach here with a partially full buffer if we
713 * have reached the end of the data. We must otherwise have a
714 * full buffer or have been given -EAGAIN.
717 if (iov_iter_count(iter) > 0)
720 goto read_phase_complete;
729 case RXRPC_CALL_COMPLETE:
740 switch (call->ackr_reason) {
743 case RXRPC_ACK_DELAY:
748 rxrpc_send_ack_packet(call, false, NULL);
752 *_service = call->service_id;
753 mutex_unlock(&call->user_mutex);
754 _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
758 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
762 trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
766 *_abort = call->abort_code;
768 if (call->completion == RXRPC_CALL_SUCCEEDED) {
770 if (iov_iter_count(iter) > 0)
775 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
778 * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
779 * @sock: The socket that the call exists on
780 * @call: The call to query
781 * @_ts: Where to put the timestamp
783 * Retrieve the timestamp from the first DATA packet of the reply if it is
784 * in the ring. Returns true if successful, false if not.
786 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
790 rxrpc_seq_t hard_ack, top, seq;
791 bool success = false;
793 mutex_lock(&call->user_mutex);
795 if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
798 hard_ack = call->rx_hard_ack;
803 top = smp_load_acquire(&call->rx_top);
807 skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
811 *_ts = skb_get_ktime(skb);
815 mutex_unlock(&call->user_mutex);
818 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);