Merge branch 'next-lsm' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
[linux-2.6-microblaze.git] / net / rxrpc / recvmsg.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* RxRPC recvmsg() implementation
3  *
4  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
5  * Written by David Howells (dhowells@redhat.com)
6  */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/net.h>
11 #include <linux/skbuff.h>
12 #include <linux/export.h>
13 #include <linux/sched/signal.h>
14
15 #include <net/sock.h>
16 #include <net/af_rxrpc.h>
17 #include "ar-internal.h"
18
19 /*
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.
22  */
23 void rxrpc_notify_socket(struct rxrpc_call *call)
24 {
25         struct rxrpc_sock *rx;
26         struct sock *sk;
27
28         _enter("%d", call->debug_id);
29
30         if (!list_empty(&call->recvmsg_link))
31                 return;
32
33         rcu_read_lock();
34
35         rx = rcu_dereference(call->socket);
36         sk = &rx->sk;
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);
42                 } else {
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);
47                         }
48                         write_unlock_bh(&rx->recvmsg_lock);
49
50                         if (!sock_flag(sk, SOCK_DEAD)) {
51                                 _debug("call %ps", sk->sk_data_ready);
52                                 sk->sk_data_ready(sk);
53                         }
54                 }
55         }
56
57         rcu_read_unlock();
58         _leave("");
59 }
60
61 /*
62  * Pass a call terminating message to userspace.
63  */
64 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
65 {
66         u32 tmp = 0;
67         int ret;
68
69         switch (call->completion) {
70         case RXRPC_CALL_SUCCEEDED:
71                 ret = 0;
72                 if (rxrpc_is_service_call(call))
73                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
74                 break;
75         case RXRPC_CALL_REMOTELY_ABORTED:
76                 tmp = call->abort_code;
77                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
78                 break;
79         case RXRPC_CALL_LOCALLY_ABORTED:
80                 tmp = call->abort_code;
81                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
82                 break;
83         case RXRPC_CALL_NETWORK_ERROR:
84                 tmp = -call->error;
85                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
86                 break;
87         case RXRPC_CALL_LOCAL_ERROR:
88                 tmp = -call->error;
89                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
90                 break;
91         default:
92                 pr_err("Invalid terminal call state %u\n", call->state);
93                 BUG();
94                 break;
95         }
96
97         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
98                             call->rx_pkt_offset, call->rx_pkt_len, ret);
99         return ret;
100 }
101
102 /*
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.
108  */
109 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
110                                   struct rxrpc_call *call,
111                                   struct msghdr *msg, int flags)
112 {
113         int tmp = 0, ret;
114
115         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
116
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);
122
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);
127         }
128
129         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
130         return ret;
131 }
132
133 /*
134  * End the packet reception phase.
135  */
136 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
137 {
138         _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
139
140         trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
141         ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
142
143         if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
144                 rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, serial, false, true,
145                                   rxrpc_propose_ack_terminal_ack);
146                 //rxrpc_send_ack_packet(call, false, NULL);
147         }
148
149         write_lock_bh(&call->state_lock);
150
151         switch (call->state) {
152         case RXRPC_CALL_CLIENT_RECV_REPLY:
153                 __rxrpc_call_completed(call);
154                 write_unlock_bh(&call->state_lock);
155                 break;
156
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, 0, serial, false, true,
163                                   rxrpc_propose_ack_processing_op);
164                 break;
165         default:
166                 write_unlock_bh(&call->state_lock);
167                 break;
168         }
169 }
170
171 /*
172  * Discard a packet we've used up and advance the Rx window by one.
173  */
174 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
175 {
176         struct rxrpc_skb_priv *sp;
177         struct sk_buff *skb;
178         rxrpc_serial_t serial;
179         rxrpc_seq_t hard_ack, top;
180         u8 flags;
181         int ix;
182
183         _enter("%d", call->debug_id);
184
185         hard_ack = call->rx_hard_ack;
186         top = smp_load_acquire(&call->rx_top);
187         ASSERT(before(hard_ack, top));
188
189         hard_ack++;
190         ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
191         skb = call->rxtx_buffer[ix];
192         rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
193         sp = rxrpc_skb(skb);
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;
198
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);
203
204         rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
205
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);
210         } else {
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,
216                                           true, true,
217                                           rxrpc_propose_ack_rotate_rx);
218                 if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
219                         rxrpc_send_ack_packet(call, false, NULL);
220         }
221 }
222
223 /*
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.
228  */
229 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
230                                u8 annotation,
231                                unsigned int offset, unsigned int len)
232 {
233         struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
234         rxrpc_seq_t seq = sp->hdr.seq;
235         u16 cksum = sp->hdr.cksum;
236
237         _enter("");
238
239         /* For all but the head jumbo subpacket, the security checksum is in a
240          * jumbo header immediately prior to the data.
241          */
242         if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
243                 __be16 tmp;
244                 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
245                         BUG();
246                 cksum = ntohs(tmp);
247                 seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
248         }
249
250         return call->conn->security->verify_packet(call, skb, offset, len,
251                                                    seq, cksum);
252 }
253
254 /*
255  * Locate the data within a packet.  This is complicated by:
256  *
257  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
258  *     subpacket.
259  *
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.
263  */
264 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
265                              u8 *_annotation,
266                              unsigned int *_offset, unsigned int *_len)
267 {
268         unsigned int offset = sizeof(struct rxrpc_wire_header);
269         unsigned int len;
270         int ret;
271         u8 annotation = *_annotation;
272
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;
280         }
281
282         if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
283                 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
284                 if (ret < 0)
285                         return ret;
286                 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
287         }
288
289         *_offset = offset;
290         *_len = len;
291         call->conn->security->locate_data(call, skb, _offset, _len);
292         return 0;
293 }
294
295 /*
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.
299  */
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)
303 {
304         struct rxrpc_skb_priv *sp;
305         struct sk_buff *skb;
306         rxrpc_seq_t hard_ack, top, seq;
307         size_t remain;
308         bool last;
309         unsigned int rx_pkt_offset, rx_pkt_len;
310         int ix, copy, ret = -EAGAIN, ret2;
311
312         if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
313             call->ackr_reason)
314                 rxrpc_send_ack_packet(call, false, NULL);
315
316         rx_pkt_offset = call->rx_pkt_offset;
317         rx_pkt_len = call->rx_pkt_len;
318
319         if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
320                 seq = call->rx_hard_ack;
321                 ret = 1;
322                 goto done;
323         }
324
325         /* Barriers against rxrpc_input_data(). */
326         hard_ack = call->rx_hard_ack;
327         seq = hard_ack + 1;
328         while (top = smp_load_acquire(&call->rx_top),
329                before_eq(seq, top)
330                ) {
331                 ix = seq & RXRPC_RXTX_BUFF_MASK;
332                 skb = call->rxtx_buffer[ix];
333                 if (!skb) {
334                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
335                                             rx_pkt_offset, rx_pkt_len, 0);
336                         break;
337                 }
338                 smp_rmb();
339                 rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
340                 sp = rxrpc_skb(skb);
341
342                 if (!(flags & MSG_PEEK))
343                         trace_rxrpc_receive(call, rxrpc_receive_front,
344                                             sp->hdr.serial, seq);
345
346                 if (msg)
347                         sock_recv_timestamp(msg, sock->sk, skb);
348
349                 if (rx_pkt_offset == 0) {
350                         ret2 = rxrpc_locate_data(call, skb,
351                                                  &call->rxtx_annotations[ix],
352                                                  &rx_pkt_offset, &rx_pkt_len);
353                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
354                                             rx_pkt_offset, rx_pkt_len, ret2);
355                         if (ret2 < 0) {
356                                 ret = ret2;
357                                 goto out;
358                         }
359                 } else {
360                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
361                                             rx_pkt_offset, rx_pkt_len, 0);
362                 }
363
364                 /* We have to handle short, empty and used-up DATA packets. */
365                 remain = len - *_offset;
366                 copy = rx_pkt_len;
367                 if (copy > remain)
368                         copy = remain;
369                 if (copy > 0) {
370                         ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
371                                                       copy);
372                         if (ret2 < 0) {
373                                 ret = ret2;
374                                 goto out;
375                         }
376
377                         /* handle piecemeal consumption of data packets */
378                         rx_pkt_offset += copy;
379                         rx_pkt_len -= copy;
380                         *_offset += copy;
381                 }
382
383                 if (rx_pkt_len > 0) {
384                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
385                                             rx_pkt_offset, rx_pkt_len, 0);
386                         ASSERTCMP(*_offset, ==, len);
387                         ret = 0;
388                         break;
389                 }
390
391                 /* The whole packet has been transferred. */
392                 last = sp->hdr.flags & RXRPC_LAST_PACKET;
393                 if (!(flags & MSG_PEEK))
394                         rxrpc_rotate_rx_window(call);
395                 rx_pkt_offset = 0;
396                 rx_pkt_len = 0;
397
398                 if (last) {
399                         ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
400                         ret = 1;
401                         goto out;
402                 }
403
404                 seq++;
405         }
406
407 out:
408         if (!(flags & MSG_PEEK)) {
409                 call->rx_pkt_offset = rx_pkt_offset;
410                 call->rx_pkt_len = rx_pkt_len;
411         }
412 done:
413         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
414                             rx_pkt_offset, rx_pkt_len, ret);
415         if (ret == -EAGAIN)
416                 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
417         return ret;
418 }
419
420 /*
421  * Receive a message from an RxRPC socket
422  * - we need to be careful about two or more threads calling recvmsg
423  *   simultaneously
424  */
425 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
426                   int flags)
427 {
428         struct rxrpc_call *call;
429         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
430         struct list_head *l;
431         size_t copied = 0;
432         long timeo;
433         int ret;
434
435         DEFINE_WAIT(wait);
436
437         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
438
439         if (flags & (MSG_OOB | MSG_TRUNC))
440                 return -EOPNOTSUPP;
441
442         timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
443
444 try_again:
445         lock_sock(&rx->sk);
446
447         /* Return immediately if a client socket has no outstanding calls */
448         if (RB_EMPTY_ROOT(&rx->calls) &&
449             list_empty(&rx->recvmsg_q) &&
450             rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
451                 release_sock(&rx->sk);
452                 return -ENODATA;
453         }
454
455         if (list_empty(&rx->recvmsg_q)) {
456                 ret = -EWOULDBLOCK;
457                 if (timeo == 0) {
458                         call = NULL;
459                         goto error_no_call;
460                 }
461
462                 release_sock(&rx->sk);
463
464                 /* Wait for something to happen */
465                 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
466                                           TASK_INTERRUPTIBLE);
467                 ret = sock_error(&rx->sk);
468                 if (ret)
469                         goto wait_error;
470
471                 if (list_empty(&rx->recvmsg_q)) {
472                         if (signal_pending(current))
473                                 goto wait_interrupted;
474                         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
475                                             0, 0, 0, 0);
476                         timeo = schedule_timeout(timeo);
477                 }
478                 finish_wait(sk_sleep(&rx->sk), &wait);
479                 goto try_again;
480         }
481
482         /* Find the next call and dequeue it if we're not just peeking.  If we
483          * do dequeue it, that comes with a ref that we will need to release.
484          */
485         write_lock_bh(&rx->recvmsg_lock);
486         l = rx->recvmsg_q.next;
487         call = list_entry(l, struct rxrpc_call, recvmsg_link);
488         if (!(flags & MSG_PEEK))
489                 list_del_init(&call->recvmsg_link);
490         else
491                 rxrpc_get_call(call, rxrpc_call_got);
492         write_unlock_bh(&rx->recvmsg_lock);
493
494         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
495
496         /* We're going to drop the socket lock, so we need to lock the call
497          * against interference by sendmsg.
498          */
499         if (!mutex_trylock(&call->user_mutex)) {
500                 ret = -EWOULDBLOCK;
501                 if (flags & MSG_DONTWAIT)
502                         goto error_requeue_call;
503                 ret = -ERESTARTSYS;
504                 if (mutex_lock_interruptible(&call->user_mutex) < 0)
505                         goto error_requeue_call;
506         }
507
508         release_sock(&rx->sk);
509
510         if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
511                 BUG();
512
513         if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
514                 if (flags & MSG_CMSG_COMPAT) {
515                         unsigned int id32 = call->user_call_ID;
516
517                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
518                                        sizeof(unsigned int), &id32);
519                 } else {
520                         unsigned long idl = call->user_call_ID;
521
522                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
523                                        sizeof(unsigned long), &idl);
524                 }
525                 if (ret < 0)
526                         goto error_unlock_call;
527         }
528
529         if (msg->msg_name) {
530                 struct sockaddr_rxrpc *srx = msg->msg_name;
531                 size_t len = sizeof(call->peer->srx);
532
533                 memcpy(msg->msg_name, &call->peer->srx, len);
534                 srx->srx_service = call->service_id;
535                 msg->msg_namelen = len;
536         }
537
538         switch (READ_ONCE(call->state)) {
539         case RXRPC_CALL_SERVER_ACCEPTING:
540                 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
541                 break;
542         case RXRPC_CALL_CLIENT_RECV_REPLY:
543         case RXRPC_CALL_SERVER_RECV_REQUEST:
544         case RXRPC_CALL_SERVER_ACK_REQUEST:
545                 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
546                                          flags, &copied);
547                 if (ret == -EAGAIN)
548                         ret = 0;
549
550                 if (after(call->rx_top, call->rx_hard_ack) &&
551                     call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
552                         rxrpc_notify_socket(call);
553                 break;
554         default:
555                 ret = 0;
556                 break;
557         }
558
559         if (ret < 0)
560                 goto error_unlock_call;
561
562         if (call->state == RXRPC_CALL_COMPLETE) {
563                 ret = rxrpc_recvmsg_term(call, msg);
564                 if (ret < 0)
565                         goto error_unlock_call;
566                 if (!(flags & MSG_PEEK))
567                         rxrpc_release_call(rx, call);
568                 msg->msg_flags |= MSG_EOR;
569                 ret = 1;
570         }
571
572         if (ret == 0)
573                 msg->msg_flags |= MSG_MORE;
574         else
575                 msg->msg_flags &= ~MSG_MORE;
576         ret = copied;
577
578 error_unlock_call:
579         mutex_unlock(&call->user_mutex);
580         rxrpc_put_call(call, rxrpc_call_put);
581         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
582         return ret;
583
584 error_requeue_call:
585         if (!(flags & MSG_PEEK)) {
586                 write_lock_bh(&rx->recvmsg_lock);
587                 list_add(&call->recvmsg_link, &rx->recvmsg_q);
588                 write_unlock_bh(&rx->recvmsg_lock);
589                 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
590         } else {
591                 rxrpc_put_call(call, rxrpc_call_put);
592         }
593 error_no_call:
594         release_sock(&rx->sk);
595 error_trace:
596         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
597         return ret;
598
599 wait_interrupted:
600         ret = sock_intr_errno(timeo);
601 wait_error:
602         finish_wait(sk_sleep(&rx->sk), &wait);
603         call = NULL;
604         goto error_trace;
605 }
606
607 /**
608  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
609  * @sock: The socket that the call exists on
610  * @call: The call to send data through
611  * @iter: The buffer to receive into
612  * @want_more: True if more data is expected to be read
613  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
614  * @_service: Where to store the actual service ID (may be upgraded)
615  *
616  * Allow a kernel service to receive data and pick up information about the
617  * state of a call.  Returns 0 if got what was asked for and there's more
618  * available, 1 if we got what was asked for and we're at the end of the data
619  * and -EAGAIN if we need more data.
620  *
621  * Note that we may return -EAGAIN to drain empty packets at the end of the
622  * data, even if we've already copied over the requested data.
623  *
624  * *_abort should also be initialised to 0.
625  */
626 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
627                            struct iov_iter *iter,
628                            bool want_more, u32 *_abort, u16 *_service)
629 {
630         size_t offset = 0;
631         int ret;
632
633         _enter("{%d,%s},%zu,%d",
634                call->debug_id, rxrpc_call_states[call->state],
635                iov_iter_count(iter), want_more);
636
637         ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
638
639         mutex_lock(&call->user_mutex);
640
641         switch (READ_ONCE(call->state)) {
642         case RXRPC_CALL_CLIENT_RECV_REPLY:
643         case RXRPC_CALL_SERVER_RECV_REQUEST:
644         case RXRPC_CALL_SERVER_ACK_REQUEST:
645                 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
646                                          iov_iter_count(iter), 0,
647                                          &offset);
648                 if (ret < 0)
649                         goto out;
650
651                 /* We can only reach here with a partially full buffer if we
652                  * have reached the end of the data.  We must otherwise have a
653                  * full buffer or have been given -EAGAIN.
654                  */
655                 if (ret == 1) {
656                         if (iov_iter_count(iter) > 0)
657                                 goto short_data;
658                         if (!want_more)
659                                 goto read_phase_complete;
660                         ret = 0;
661                         goto out;
662                 }
663
664                 if (!want_more)
665                         goto excess_data;
666                 goto out;
667
668         case RXRPC_CALL_COMPLETE:
669                 goto call_complete;
670
671         default:
672                 ret = -EINPROGRESS;
673                 goto out;
674         }
675
676 read_phase_complete:
677         ret = 1;
678 out:
679         switch (call->ackr_reason) {
680         case RXRPC_ACK_IDLE:
681                 break;
682         case RXRPC_ACK_DELAY:
683                 if (ret != -EAGAIN)
684                         break;
685                 /* Fall through */
686         default:
687                 rxrpc_send_ack_packet(call, false, NULL);
688         }
689
690         if (_service)
691                 *_service = call->service_id;
692         mutex_unlock(&call->user_mutex);
693         _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
694         return ret;
695
696 short_data:
697         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
698         ret = -EBADMSG;
699         goto out;
700 excess_data:
701         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
702         ret = -EMSGSIZE;
703         goto out;
704 call_complete:
705         *_abort = call->abort_code;
706         ret = call->error;
707         if (call->completion == RXRPC_CALL_SUCCEEDED) {
708                 ret = 1;
709                 if (iov_iter_count(iter) > 0)
710                         ret = -ECONNRESET;
711         }
712         goto out;
713 }
714 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
715
716 /**
717  * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
718  * @sock: The socket that the call exists on
719  * @call: The call to query
720  * @_ts: Where to put the timestamp
721  *
722  * Retrieve the timestamp from the first DATA packet of the reply if it is
723  * in the ring.  Returns true if successful, false if not.
724  */
725 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
726                                  ktime_t *_ts)
727 {
728         struct sk_buff *skb;
729         rxrpc_seq_t hard_ack, top, seq;
730         bool success = false;
731
732         mutex_lock(&call->user_mutex);
733
734         if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
735                 goto out;
736
737         hard_ack = call->rx_hard_ack;
738         if (hard_ack != 0)
739                 goto out;
740
741         seq = hard_ack + 1;
742         top = smp_load_acquire(&call->rx_top);
743         if (after(seq, top))
744                 goto out;
745
746         skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
747         if (!skb)
748                 goto out;
749
750         *_ts = skb_get_ktime(skb);
751         success = true;
752
753 out:
754         mutex_unlock(&call->user_mutex);
755         return success;
756 }
757 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);