net/rxrpc/recvmsg.c

   1 /* RxRPC recvmsg() implementation
   2  *
   3  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
   4  * Written by David Howells (dhowells@redhat.com)
   5  *
   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.
  10  */
  11
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13
  14 #include <linux/net.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/export.h>
  17 #include <linux/sched/signal.h>
  18
  19 #include <net/sock.h>
  20 #include <net/af_rxrpc.h>
  21 #include "ar-internal.h"
  22
  23 /*
  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.
  26  */
  27 void rxrpc_notify_socket(struct rxrpc_call *call)
  28 {
  29         struct rxrpc_sock *rx;
  30         struct sock *sk;
  31
  32         _enter("%d", call->debug_id);
  33
  34         if (!list_empty(&call->recvmsg_link))
  35                 return;
  36
  37         rcu_read_lock();
  38
  39         rx = rcu_dereference(call->socket);
  40         sk = &rx->sk;
  41         if (rx && sk->sk_state < RXRPC_CLOSE) {
  42                 if (call->notify_rx) {
  43                         spin_lock_bh(&call->notify_lock);
  44                         call->notify_rx(sk, call, call->user_call_ID);
  45                         spin_unlock_bh(&call->notify_lock);
  46                 } else {
  47                         write_lock_bh(&rx->recvmsg_lock);
  48                         if (list_empty(&call->recvmsg_link)) {
  49                                 rxrpc_get_call(call, rxrpc_call_got);
  50                                 list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
  51                         }
  52                         write_unlock_bh(&rx->recvmsg_lock);
  53
  54                         if (!sock_flag(sk, SOCK_DEAD)) {
  55                                 _debug("call %ps", sk->sk_data_ready);
  56                                 sk->sk_data_ready(sk);
  57                         }
  58                 }
  59         }
  60
  61         rcu_read_unlock();
  62         _leave("");
  63 }
  64
  65 /*
  66  * Pass a call terminating message to userspace.
  67  */
  68 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
  69 {
  70         u32 tmp = 0;
  71         int ret;
  72
  73         switch (call->completion) {
  74         case RXRPC_CALL_SUCCEEDED:
  75                 ret = 0;
  76                 if (rxrpc_is_service_call(call))
  77                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
  78                 break;
  79         case RXRPC_CALL_REMOTELY_ABORTED:
  80                 tmp = call->abort_code;
  81                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
  82                 break;
  83         case RXRPC_CALL_LOCALLY_ABORTED:
  84                 tmp = call->abort_code;
  85                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
  86                 break;
  87         case RXRPC_CALL_NETWORK_ERROR:
  88                 tmp = -call->error;
  89                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
  90                 break;
  91         case RXRPC_CALL_LOCAL_ERROR:
  92                 tmp = -call->error;
  93                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
  94                 break;
  95         default:
  96                 pr_err("Invalid terminal call state %u\n", call->state);
  97                 BUG();
  98                 break;
  99         }
 100
 101         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
 102                             call->rx_pkt_offset, call->rx_pkt_len, ret);
 103         return ret;
 104 }
 105
 106 /*
 107  * Pass back notification of a new call.  The call is added to the
 108  * to-be-accepted list.  This means that the next call to be accepted might not
 109  * be the last call seen awaiting acceptance, but unless we leave this on the
 110  * front of the queue and block all other messages until someone gives us a
 111  * user_ID for it, there's not a lot we can do.
 112  */
 113 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
 114                                   struct rxrpc_call *call,
 115                                   struct msghdr *msg, int flags)
 116 {
 117         int tmp = 0, ret;
 118
 119         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
 120
 121         if (ret == 0 && !(flags & MSG_PEEK)) {
 122                 _debug("to be accepted");
 123                 write_lock_bh(&rx->recvmsg_lock);
 124                 list_del_init(&call->recvmsg_link);
 125                 write_unlock_bh(&rx->recvmsg_lock);
 126
 127                 rxrpc_get_call(call, rxrpc_call_got);
 128                 write_lock(&rx->call_lock);
 129                 list_add_tail(&call->accept_link, &rx->to_be_accepted);
 130                 write_unlock(&rx->call_lock);
 131         }
 132
 133         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
 134         return ret;
 135 }
 136
 137 /*
 138  * End the packet reception phase.
 139  */
 140 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
 141 {
 142         _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
 143
 144         trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
 145         ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
 146
 147         if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
 148                 rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, serial, false, true,
 149                                   rxrpc_propose_ack_terminal_ack);
 150                 //rxrpc_send_ack_packet(call, false, NULL);
 151         }
 152
 153         write_lock_bh(&call->state_lock);
 154
 155         switch (call->state) {
 156         case RXRPC_CALL_CLIENT_RECV_REPLY:
 157                 __rxrpc_call_completed(call);
 158                 write_unlock_bh(&call->state_lock);
 159                 break;
 160
 161         case RXRPC_CALL_SERVER_RECV_REQUEST:
 162                 call->tx_phase = true;
 163                 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
 164                 call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
 165                 write_unlock_bh(&call->state_lock);
 166                 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial, false, true,
 167                                   rxrpc_propose_ack_processing_op);
 168                 break;
 169         default:
 170                 write_unlock_bh(&call->state_lock);
 171                 break;
 172         }
 173 }
 174
 175 /*
 176  * Discard a packet we've used up and advance the Rx window by one.
 177  */
 178 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
 179 {
 180         struct rxrpc_skb_priv *sp;
 181         struct sk_buff *skb;
 182         rxrpc_serial_t serial;
 183         rxrpc_seq_t hard_ack, top;
 184         u8 flags;
 185         int ix;
 186
 187         _enter("%d", call->debug_id);
 188
 189         hard_ack = call->rx_hard_ack;
 190         top = smp_load_acquire(&call->rx_top);
 191         ASSERT(before(hard_ack, top));
 192
 193         hard_ack++;
 194         ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
 195         skb = call->rxtx_buffer[ix];
 196         rxrpc_see_skb(skb, rxrpc_skb_rx_rotated);
 197         sp = rxrpc_skb(skb);
 198         flags = sp->hdr.flags;
 199         serial = sp->hdr.serial;
 200         if (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO)
 201                 serial += (call->rxtx_annotations[ix] & RXRPC_RX_ANNO_JUMBO) - 1;
 202
 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);
 207
 208         rxrpc_free_skb(skb, rxrpc_skb_rx_freed);
 209
 210         _debug("%u,%u,%02x", hard_ack, top, flags);
 211         trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
 212         if (flags & RXRPC_LAST_PACKET) {
 213                 rxrpc_end_rx_phase(call, serial);
 214         } else {
 215                 /* Check to see if there's an ACK that needs sending. */
 216                 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
 217                     after_eq(top, call->ackr_seen + 2) ||
 218                     (hard_ack == top && after(hard_ack, call->ackr_consumed)))
 219                         rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, 0, serial,
 220                                           true, true,
 221                                           rxrpc_propose_ack_rotate_rx);
 222                 if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
 223                         rxrpc_send_ack_packet(call, false, NULL);
 224         }
 225 }
 226
 227 /*
 228  * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
 229  * padding, but if this is the case, the packet length will be resident in the
 230  * socket buffer.  Note that we can't modify the master skb info as the skb may
 231  * be the home to multiple subpackets.
 232  */
 233 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
 234                                u8 annotation,
 235                                unsigned int offset, unsigned int len)
 236 {
 237         struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
 238         rxrpc_seq_t seq = sp->hdr.seq;
 239         u16 cksum = sp->hdr.cksum;
 240
 241         _enter("");
 242
 243         /* For all but the head jumbo subpacket, the security checksum is in a
 244          * jumbo header immediately prior to the data.
 245          */
 246         if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
 247                 __be16 tmp;
 248                 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
 249                         BUG();
 250                 cksum = ntohs(tmp);
 251                 seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
 252         }
 253
 254         return call->conn->security->verify_packet(call, skb, offset, len,
 255                                                    seq, cksum);
 256 }
 257
 258 /*
 259  * Locate the data within a packet.  This is complicated by:
 260  *
 261  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
 262  *     subpacket.
 263  *
 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.
 267  */
 268 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
 269                              u8 *_annotation,
 270                              unsigned int *_offset, unsigned int *_len)
 271 {
 272         unsigned int offset = sizeof(struct rxrpc_wire_header);
 273         unsigned int len;
 274         int ret;
 275         u8 annotation = *_annotation;
 276
 277         /* Locate the subpacket */
 278         len = skb->len - offset;
 279         if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
 280                 offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
 281                            RXRPC_JUMBO_SUBPKTLEN);
 282                 len = (annotation & RXRPC_RX_ANNO_JLAST) ?
 283                         skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
 284         }
 285
 286         if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
 287                 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
 288                 if (ret < 0)
 289                         return ret;
 290                 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
 291         }
 292
 293         *_offset = offset;
 294         *_len = len;
 295         call->conn->security->locate_data(call, skb, _offset, _len);
 296         return 0;
 297 }
 298
 299 /*
 300  * Deliver messages to a call.  This keeps processing packets until the buffer
 301  * is filled and we find either more DATA (returns 0) or the end of the DATA
 302  * (returns 1).  If more packets are required, it returns -EAGAIN.
 303  */
 304 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
 305                               struct msghdr *msg, struct iov_iter *iter,
 306                               size_t len, int flags, size_t *_offset)
 307 {
 308         struct rxrpc_skb_priv *sp;
 309         struct sk_buff *skb;
 310         rxrpc_seq_t hard_ack, top, seq;
 311         size_t remain;
 312         bool last;
 313         unsigned int rx_pkt_offset, rx_pkt_len;
 314         int ix, copy, ret = -EAGAIN, ret2;
 315
 316         if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
 317             call->ackr_reason)
 318                 rxrpc_send_ack_packet(call, false, NULL);
 319
 320         rx_pkt_offset = call->rx_pkt_offset;
 321         rx_pkt_len = call->rx_pkt_len;
 322
 323         if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
 324                 seq = call->rx_hard_ack;
 325                 ret = 1;
 326                 goto done;
 327         }
 328
 329         /* Barriers against rxrpc_input_data(). */
 330         hard_ack = call->rx_hard_ack;
 331         seq = hard_ack + 1;
 332         while (top = smp_load_acquire(&call->rx_top),
 333                before_eq(seq, top)
 334                ) {
 335                 ix = seq & RXRPC_RXTX_BUFF_MASK;
 336                 skb = call->rxtx_buffer[ix];
 337                 if (!skb) {
 338                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
 339                                             rx_pkt_offset, rx_pkt_len, 0);
 340                         break;
 341                 }
 342                 smp_rmb();
 343                 rxrpc_see_skb(skb, rxrpc_skb_rx_seen);
 344                 sp = rxrpc_skb(skb);
 345
 346                 if (!(flags & MSG_PEEK))
 347                         trace_rxrpc_receive(call, rxrpc_receive_front,
 348                                             sp->hdr.serial, seq);
 349
 350                 if (msg)
 351                         sock_recv_timestamp(msg, sock->sk, skb);
 352
 353                 if (rx_pkt_offset == 0) {
 354                         ret2 = rxrpc_locate_data(call, skb,
 355                                                  &call->rxtx_annotations[ix],
 356                                                  &rx_pkt_offset, &rx_pkt_len);
 357                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
 358                                             rx_pkt_offset, rx_pkt_len, ret2);
 359                         if (ret2 < 0) {
 360                                 ret = ret2;
 361                                 goto out;
 362                         }
 363                 } else {
 364                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
 365                                             rx_pkt_offset, rx_pkt_len, 0);
 366                 }
 367
 368                 /* We have to handle short, empty and used-up DATA packets. */
 369                 remain = len - *_offset;
 370                 copy = rx_pkt_len;
 371                 if (copy > remain)
 372                         copy = remain;
 373                 if (copy > 0) {
 374                         ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
 375                                                       copy);
 376                         if (ret2 < 0) {
 377                                 ret = ret2;
 378                                 goto out;
 379                         }
 380
 381                         /* handle piecemeal consumption of data packets */
 382                         rx_pkt_offset += copy;
 383                         rx_pkt_len -= copy;
 384                         *_offset += copy;
 385                 }
 386
 387                 if (rx_pkt_len > 0) {
 388                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
 389                                             rx_pkt_offset, rx_pkt_len, 0);
 390                         ASSERTCMP(*_offset, ==, len);
 391                         ret = 0;
 392                         break;
 393                 }
 394
 395                 /* The whole packet has been transferred. */
 396                 last = sp->hdr.flags & RXRPC_LAST_PACKET;
 397                 if (!(flags & MSG_PEEK))
 398                         rxrpc_rotate_rx_window(call);
 399                 rx_pkt_offset = 0;
 400                 rx_pkt_len = 0;
 401
 402                 if (last) {
 403                         ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
 404                         ret = 1;
 405                         goto out;
 406                 }
 407
 408                 seq++;
 409         }
 410
 411 out:
 412         if (!(flags & MSG_PEEK)) {
 413                 call->rx_pkt_offset = rx_pkt_offset;
 414                 call->rx_pkt_len = rx_pkt_len;
 415         }
 416 done:
 417         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
 418                             rx_pkt_offset, rx_pkt_len, ret);
 419         if (ret == -EAGAIN)
 420                 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
 421         return ret;
 422 }
 423
 424 /*
 425  * Receive a message from an RxRPC socket
 426  * - we need to be careful about two or more threads calling recvmsg
 427  *   simultaneously
 428  */
 429 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
 430                   int flags)
 431 {
 432         struct rxrpc_call *call;
 433         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 434         struct list_head *l;
 435         size_t copied = 0;
 436         long timeo;
 437         int ret;
 438
 439         DEFINE_WAIT(wait);
 440
 441         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
 442
 443         if (flags & (MSG_OOB | MSG_TRUNC))
 444                 return -EOPNOTSUPP;
 445
 446         timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
 447
 448 try_again:
 449         lock_sock(&rx->sk);
 450
 451         /* Return immediately if a client socket has no outstanding calls */
 452         if (RB_EMPTY_ROOT(&rx->calls) &&
 453             list_empty(&rx->recvmsg_q) &&
 454             rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
 455                 release_sock(&rx->sk);
 456                 return -ENODATA;
 457         }
 458
 459         if (list_empty(&rx->recvmsg_q)) {
 460                 ret = -EWOULDBLOCK;
 461                 if (timeo == 0) {
 462                         call = NULL;
 463                         goto error_no_call;
 464                 }
 465
 466                 release_sock(&rx->sk);
 467
 468                 /* Wait for something to happen */
 469                 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
 470                                           TASK_INTERRUPTIBLE);
 471                 ret = sock_error(&rx->sk);
 472                 if (ret)
 473                         goto wait_error;
 474
 475                 if (list_empty(&rx->recvmsg_q)) {
 476                         if (signal_pending(current))
 477                                 goto wait_interrupted;
 478                         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
 479                                             0, 0, 0, 0);
 480                         timeo = schedule_timeout(timeo);
 481                 }
 482                 finish_wait(sk_sleep(&rx->sk), &wait);
 483                 goto try_again;
 484         }
 485
 486         /* Find the next call and dequeue it if we're not just peeking.  If we
 487          * do dequeue it, that comes with a ref that we will need to release.
 488          */
 489         write_lock_bh(&rx->recvmsg_lock);
 490         l = rx->recvmsg_q.next;
 491         call = list_entry(l, struct rxrpc_call, recvmsg_link);
 492         if (!(flags & MSG_PEEK))
 493                 list_del_init(&call->recvmsg_link);
 494         else
 495                 rxrpc_get_call(call, rxrpc_call_got);
 496         write_unlock_bh(&rx->recvmsg_lock);
 497
 498         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
 499
 500         /* We're going to drop the socket lock, so we need to lock the call
 501          * against interference by sendmsg.
 502          */
 503         if (!mutex_trylock(&call->user_mutex)) {
 504                 ret = -EWOULDBLOCK;
 505                 if (flags & MSG_DONTWAIT)
 506                         goto error_requeue_call;
 507                 ret = -ERESTARTSYS;
 508                 if (mutex_lock_interruptible(&call->user_mutex) < 0)
 509                         goto error_requeue_call;
 510         }
 511
 512         release_sock(&rx->sk);
 513
 514         if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
 515                 BUG();
 516
 517         if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
 518                 if (flags & MSG_CMSG_COMPAT) {
 519                         unsigned int id32 = call->user_call_ID;
 520
 521                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
 522                                        sizeof(unsigned int), &id32);
 523                 } else {
 524                         unsigned long idl = call->user_call_ID;
 525
 526                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
 527                                        sizeof(unsigned long), &idl);
 528                 }
 529                 if (ret < 0)
 530                         goto error_unlock_call;
 531         }
 532
 533         if (msg->msg_name) {
 534                 struct sockaddr_rxrpc *srx = msg->msg_name;
 535                 size_t len = sizeof(call->peer->srx);
 536
 537                 memcpy(msg->msg_name, &call->peer->srx, len);
 538                 srx->srx_service = call->service_id;
 539                 msg->msg_namelen = len;
 540         }
 541
 542         switch (READ_ONCE(call->state)) {
 543         case RXRPC_CALL_SERVER_ACCEPTING:
 544                 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
 545                 break;
 546         case RXRPC_CALL_CLIENT_RECV_REPLY:
 547         case RXRPC_CALL_SERVER_RECV_REQUEST:
 548         case RXRPC_CALL_SERVER_ACK_REQUEST:
 549                 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
 550                                          flags, &copied);
 551                 if (ret == -EAGAIN)
 552                         ret = 0;
 553
 554                 if (after(call->rx_top, call->rx_hard_ack) &&
 555                     call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
 556                         rxrpc_notify_socket(call);
 557                 break;
 558         default:
 559                 ret = 0;
 560                 break;
 561         }
 562
 563         if (ret < 0)
 564                 goto error_unlock_call;
 565
 566         if (call->state == RXRPC_CALL_COMPLETE) {
 567                 ret = rxrpc_recvmsg_term(call, msg);
 568                 if (ret < 0)
 569                         goto error_unlock_call;
 570                 if (!(flags & MSG_PEEK))
 571                         rxrpc_release_call(rx, call);
 572                 msg->msg_flags |= MSG_EOR;
 573                 ret = 1;
 574         }
 575
 576         if (ret == 0)
 577                 msg->msg_flags |= MSG_MORE;
 578         else
 579                 msg->msg_flags &= ~MSG_MORE;
 580         ret = copied;
 581
 582 error_unlock_call:
 583         mutex_unlock(&call->user_mutex);
 584         rxrpc_put_call(call, rxrpc_call_put);
 585         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
 586         return ret;
 587
 588 error_requeue_call:
 589         if (!(flags & MSG_PEEK)) {
 590                 write_lock_bh(&rx->recvmsg_lock);
 591                 list_add(&call->recvmsg_link, &rx->recvmsg_q);
 592                 write_unlock_bh(&rx->recvmsg_lock);
 593                 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
 594         } else {
 595                 rxrpc_put_call(call, rxrpc_call_put);
 596         }
 597 error_no_call:
 598         release_sock(&rx->sk);
 599 error_trace:
 600         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
 601         return ret;
 602
 603 wait_interrupted:
 604         ret = sock_intr_errno(timeo);
 605 wait_error:
 606         finish_wait(sk_sleep(&rx->sk), &wait);
 607         call = NULL;
 608         goto error_trace;
 609 }
 610
 611 /**
 612  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
 613  * @sock: The socket that the call exists on
 614  * @call: The call to send data through
 615  * @iter: The buffer to receive into
 616  * @want_more: True if more data is expected to be read
 617  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
 618  * @_service: Where to store the actual service ID (may be upgraded)
 619  *
 620  * Allow a kernel service to receive data and pick up information about the
 621  * state of a call.  Returns 0 if got what was asked for and there's more
 622  * available, 1 if we got what was asked for and we're at the end of the data
 623  * and -EAGAIN if we need more data.
 624  *
 625  * Note that we may return -EAGAIN to drain empty packets at the end of the
 626  * data, even if we've already copied over the requested data.
 627  *
 628  * *_abort should also be initialised to 0.
 629  */
 630 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
 631                            struct iov_iter *iter,
 632                            bool want_more, u32 *_abort, u16 *_service)
 633 {
 634         size_t offset = 0;
 635         int ret;
 636
 637         _enter("{%d,%s},%zu,%d",
 638                call->debug_id, rxrpc_call_states[call->state],
 639                iov_iter_count(iter), want_more);
 640
 641         ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
 642
 643         mutex_lock(&call->user_mutex);
 644
 645         switch (READ_ONCE(call->state)) {
 646         case RXRPC_CALL_CLIENT_RECV_REPLY:
 647         case RXRPC_CALL_SERVER_RECV_REQUEST:
 648         case RXRPC_CALL_SERVER_ACK_REQUEST:
 649                 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
 650                                          iov_iter_count(iter), 0,
 651                                          &offset);
 652                 if (ret < 0)
 653                         goto out;
 654
 655                 /* We can only reach here with a partially full buffer if we
 656                  * have reached the end of the data.  We must otherwise have a
 657                  * full buffer or have been given -EAGAIN.
 658                  */
 659                 if (ret == 1) {
 660                         if (iov_iter_count(iter) > 0)
 661                                 goto short_data;
 662                         if (!want_more)
 663                                 goto read_phase_complete;
 664                         ret = 0;
 665                         goto out;
 666                 }
 667
 668                 if (!want_more)
 669                         goto excess_data;
 670                 goto out;
 671
 672         case RXRPC_CALL_COMPLETE:
 673                 goto call_complete;
 674
 675         default:
 676                 ret = -EINPROGRESS;
 677                 goto out;
 678         }
 679
 680 read_phase_complete:
 681         ret = 1;
 682 out:
 683         switch (call->ackr_reason) {
 684         case RXRPC_ACK_IDLE:
 685                 break;
 686         case RXRPC_ACK_DELAY:
 687                 if (ret != -EAGAIN)
 688                         break;
 689                 /* Fall through */
 690         default:
 691                 rxrpc_send_ack_packet(call, false, NULL);
 692         }
 693
 694         if (_service)
 695                 *_service = call->service_id;
 696         mutex_unlock(&call->user_mutex);
 697         _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
 698         return ret;
 699
 700 short_data:
 701         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
 702         ret = -EBADMSG;
 703         goto out;
 704 excess_data:
 705         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
 706         ret = -EMSGSIZE;
 707         goto out;
 708 call_complete:
 709         *_abort = call->abort_code;
 710         ret = call->error;
 711         if (call->completion == RXRPC_CALL_SUCCEEDED) {
 712                 ret = 1;
 713                 if (iov_iter_count(iter) > 0)
 714                         ret = -ECONNRESET;
 715         }
 716         goto out;
 717 }
 718 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
 719
 720 /**
 721  * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
 722  * @sock: The socket that the call exists on
 723  * @call: The call to query
 724  * @_ts: Where to put the timestamp
 725  *
 726  * Retrieve the timestamp from the first DATA packet of the reply if it is
 727  * in the ring.  Returns true if successful, false if not.
 728  */
 729 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
 730                                  ktime_t *_ts)
 731 {
 732         struct sk_buff *skb;
 733         rxrpc_seq_t hard_ack, top, seq;
 734         bool success = false;
 735
 736         mutex_lock(&call->user_mutex);
 737
 738         if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
 739                 goto out;
 740
 741         hard_ack = call->rx_hard_ack;
 742         if (hard_ack != 0)
 743                 goto out;
 744
 745         seq = hard_ack + 1;
 746         top = smp_load_acquire(&call->rx_top);
 747         if (after(seq, top))
 748                 goto out;
 749
 750         skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
 751         if (!skb)
 752                 goto out;
 753
 754         *_ts = skb_get_ktime(skb);
 755         success = true;
 756
 757 out:
 758         mutex_unlock(&call->user_mutex);
 759         return success;
 760 }
 761 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);