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>
18 #include <net/af_rxrpc.h>
19 #include "ar-internal.h"
22 * Post a call for attention by the socket or kernel service. Further
23 * notifications are suppressed by putting recvmsg_link on a dummy queue.
25 void rxrpc_notify_socket(struct rxrpc_call *call)
27 struct rxrpc_sock *rx;
30 _enter("%d", call->debug_id);
32 if (!list_empty(&call->recvmsg_link))
37 rx = rcu_dereference(call->socket);
39 if (rx && sk->sk_state < RXRPC_CLOSE) {
40 if (call->notify_rx) {
41 call->notify_rx(sk, call, call->user_call_ID);
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, 0, serial, true, false,
145 rxrpc_propose_ack_terminal_ack);
146 rxrpc_send_ack_packet(call, false);
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->ack_at = call->expire_at;
161 write_unlock_bh(&call->state_lock);
162 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, 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;
183 _enter("%d", call->debug_id);
185 hard_ack = call->rx_hard_ack;
186 top = smp_load_acquire(&call->rx_top);
187 ASSERT(before(hard_ack, top));
190 ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
191 skb = call->rxtx_buffer[ix];
192 rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
194 flags = sp->hdr.flags;
195 serial = sp->hdr.serial;
196 if (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO)
197 serial += (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO) - 1;
199 call->rxtx_buffer[ix] = NULL;
200 call->rxtx_annotations[ix] = 0;
201 /* Barrier against rxrpc_input_data(). */
202 smp_store_release(&call->rx_hard_ack, hard_ack);
204 rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
206 _debug("%u,%u,%02x", hard_ack, top, flags);
207 trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
208 if (flags & RXRPC_LAST_PACKET) {
209 rxrpc_end_rx_phase(call, serial);
211 /* Check to see if there's an ACK that needs sending. */
212 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
213 after_eq(top, call->ackr_seen + 2) ||
214 (hard_ack == top && after(hard_ack, call->ackr_consumed)))
215 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial,
217 rxrpc_propose_ack_rotate_rx);
218 if (call->ackr_reason)
219 rxrpc_send_ack_packet(call, false);
224 * Decrypt and verify a (sub)packet. The packet's length may be changed due to
225 * padding, but if this is the case, the packet length will be resident in the
226 * socket buffer. Note that we can't modify the master skb info as the skb may
227 * be the home to multiple subpackets.
229 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
231 unsigned int offset, unsigned int len)
233 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
234 rxrpc_seq_t seq = sp->hdr.seq;
235 u16 cksum = sp->hdr.cksum;
239 /* For all but the head jumbo subpacket, the security checksum is in a
240 * jumbo header immediately prior to the data.
242 if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
244 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
247 seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
250 return call->conn->security->verify_packet(call, skb, offset, len,
255 * Locate the data within a packet. This is complicated by:
257 * (1) An skb may contain a jumbo packet - so we have to find the appropriate
260 * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
261 * contains an extra header which includes the true length of the data,
262 * excluding any encrypted padding.
264 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
266 unsigned int *_offset, unsigned int *_len)
268 unsigned int offset = sizeof(struct rxrpc_wire_header);
269 unsigned int len = *_len;
271 u8 annotation = *_annotation;
273 /* Locate the subpacket */
274 len = skb->len - offset;
275 if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
276 offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
277 RXRPC_JUMBO_SUBPKTLEN);
278 len = (annotation & RXRPC_RX_ANNO_JLAST) ?
279 skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
282 if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
283 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
286 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
291 call->conn->security->locate_data(call, skb, _offset, _len);
296 * Deliver messages to a call. This keeps processing packets until the buffer
297 * is filled and we find either more DATA (returns 0) or the end of the DATA
298 * (returns 1). If more packets are required, it returns -EAGAIN.
300 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
301 struct msghdr *msg, struct iov_iter *iter,
302 size_t len, int flags, size_t *_offset)
304 struct rxrpc_skb_priv *sp;
306 rxrpc_seq_t hard_ack, top, seq;
309 unsigned int rx_pkt_offset, rx_pkt_len;
310 int ix, copy, ret = -EAGAIN, ret2;
312 rx_pkt_offset = call->rx_pkt_offset;
313 rx_pkt_len = call->rx_pkt_len;
315 if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
316 seq = call->rx_hard_ack;
321 /* Barriers against rxrpc_input_data(). */
322 hard_ack = call->rx_hard_ack;
324 while (top = smp_load_acquire(&call->rx_top),
327 ix = seq & RXRPC_RXTX_BUFF_MASK;
328 skb = call->rxtx_buffer[ix];
330 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
331 rx_pkt_offset, rx_pkt_len, 0);
335 rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
338 if (!(flags & MSG_PEEK))
339 trace_rxrpc_receive(call, rxrpc_receive_front,
340 sp->hdr.serial, seq);
343 sock_recv_timestamp(msg, sock->sk, skb);
345 if (rx_pkt_offset == 0) {
346 ret2 = rxrpc_locate_data(call, skb,
347 &call->rxtx_annotations[ix],
348 &rx_pkt_offset, &rx_pkt_len);
349 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
350 rx_pkt_offset, rx_pkt_len, ret2);
356 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
357 rx_pkt_offset, rx_pkt_len, 0);
360 /* We have to handle short, empty and used-up DATA packets. */
361 remain = len - *_offset;
366 ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
373 /* handle piecemeal consumption of data packets */
374 rx_pkt_offset += copy;
379 if (rx_pkt_len > 0) {
380 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
381 rx_pkt_offset, rx_pkt_len, 0);
382 ASSERTCMP(*_offset, ==, len);
387 /* The whole packet has been transferred. */
388 last = sp->hdr.flags & RXRPC_LAST_PACKET;
389 if (!(flags & MSG_PEEK))
390 rxrpc_rotate_rx_window(call);
395 ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
404 if (!(flags & MSG_PEEK)) {
405 call->rx_pkt_offset = rx_pkt_offset;
406 call->rx_pkt_len = rx_pkt_len;
409 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
410 rx_pkt_offset, rx_pkt_len, ret);
415 * Receive a message from an RxRPC socket
416 * - we need to be careful about two or more threads calling recvmsg
419 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
422 struct rxrpc_call *call;
423 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
431 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
433 if (flags & (MSG_OOB | MSG_TRUNC))
436 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
441 /* Return immediately if a client socket has no outstanding calls */
442 if (RB_EMPTY_ROOT(&rx->calls) &&
443 list_empty(&rx->recvmsg_q) &&
444 rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
445 release_sock(&rx->sk);
449 if (list_empty(&rx->recvmsg_q)) {
456 release_sock(&rx->sk);
458 /* Wait for something to happen */
459 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
461 ret = sock_error(&rx->sk);
465 if (list_empty(&rx->recvmsg_q)) {
466 if (signal_pending(current))
467 goto wait_interrupted;
468 trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
470 timeo = schedule_timeout(timeo);
472 finish_wait(sk_sleep(&rx->sk), &wait);
476 /* Find the next call and dequeue it if we're not just peeking. If we
477 * do dequeue it, that comes with a ref that we will need to release.
479 write_lock_bh(&rx->recvmsg_lock);
480 l = rx->recvmsg_q.next;
481 call = list_entry(l, struct rxrpc_call, recvmsg_link);
482 if (!(flags & MSG_PEEK))
483 list_del_init(&call->recvmsg_link);
485 rxrpc_get_call(call, rxrpc_call_got);
486 write_unlock_bh(&rx->recvmsg_lock);
488 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
490 if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
493 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
494 if (flags & MSG_CMSG_COMPAT) {
495 unsigned int id32 = call->user_call_ID;
497 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
498 sizeof(unsigned int), &id32);
500 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
501 sizeof(unsigned long),
502 &call->user_call_ID);
509 size_t len = sizeof(call->conn->params.peer->srx);
510 memcpy(msg->msg_name, &call->conn->params.peer->srx, len);
511 msg->msg_namelen = len;
514 switch (call->state) {
515 case RXRPC_CALL_SERVER_ACCEPTING:
516 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
518 case RXRPC_CALL_CLIENT_RECV_REPLY:
519 case RXRPC_CALL_SERVER_RECV_REQUEST:
520 case RXRPC_CALL_SERVER_ACK_REQUEST:
521 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
526 if (after(call->rx_top, call->rx_hard_ack) &&
527 call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
528 rxrpc_notify_socket(call);
538 if (call->state == RXRPC_CALL_COMPLETE) {
539 ret = rxrpc_recvmsg_term(call, msg);
542 if (!(flags & MSG_PEEK))
543 rxrpc_release_call(rx, call);
544 msg->msg_flags |= MSG_EOR;
549 msg->msg_flags |= MSG_MORE;
551 msg->msg_flags &= ~MSG_MORE;
555 rxrpc_put_call(call, rxrpc_call_put);
557 release_sock(&rx->sk);
558 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
562 ret = sock_intr_errno(timeo);
564 finish_wait(sk_sleep(&rx->sk), &wait);
570 * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
571 * @sock: The socket that the call exists on
572 * @call: The call to send data through
573 * @buf: The buffer to receive into
574 * @size: The size of the buffer, including data already read
575 * @_offset: The running offset into the buffer.
576 * @want_more: True if more data is expected to be read
577 * @_abort: Where the abort code is stored if -ECONNABORTED is returned
579 * Allow a kernel service to receive data and pick up information about the
580 * state of a call. Returns 0 if got what was asked for and there's more
581 * available, 1 if we got what was asked for and we're at the end of the data
582 * and -EAGAIN if we need more data.
584 * Note that we may return -EAGAIN to drain empty packets at the end of the
585 * data, even if we've already copied over the requested data.
587 * This function adds the amount it transfers to *_offset, so this should be
588 * precleared as appropriate. Note that the amount remaining in the buffer is
589 * taken to be size - *_offset.
591 * *_abort should also be initialised to 0.
593 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
594 void *buf, size_t size, size_t *_offset,
595 bool want_more, u32 *_abort)
597 struct iov_iter iter;
601 _enter("{%d,%s},%zu/%zu,%d",
602 call->debug_id, rxrpc_call_states[call->state],
603 *_offset, size, want_more);
605 ASSERTCMP(*_offset, <=, size);
606 ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
608 iov.iov_base = buf + *_offset;
609 iov.iov_len = size - *_offset;
610 iov_iter_kvec(&iter, ITER_KVEC | READ, &iov, 1, size - *_offset);
614 switch (call->state) {
615 case RXRPC_CALL_CLIENT_RECV_REPLY:
616 case RXRPC_CALL_SERVER_RECV_REQUEST:
617 case RXRPC_CALL_SERVER_ACK_REQUEST:
618 ret = rxrpc_recvmsg_data(sock, call, NULL, &iter, size, 0,
623 /* We can only reach here with a partially full buffer if we
624 * have reached the end of the data. We must otherwise have a
625 * full buffer or have been given -EAGAIN.
631 goto read_phase_complete;
640 case RXRPC_CALL_COMPLETE:
651 release_sock(sock->sk);
652 _leave(" = %d [%zu,%d]", ret, *_offset, *_abort);
662 *_abort = call->abort_code;
664 if (call->completion == RXRPC_CALL_SUCCEEDED) {
671 EXPORT_SYMBOL(rxrpc_kernel_recv_data);