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  * Transition a call to the complete state.
  63  */
  64 bool __rxrpc_set_call_completion(struct rxrpc_call *call,
  65                                  enum rxrpc_call_completion compl,
  66                                  u32 abort_code,
  67                                  int error)
  68 {
  69         if (call->state < RXRPC_CALL_COMPLETE) {
  70                 call->abort_code = abort_code;
  71                 call->error = error;
  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);
  77                 return true;
  78         }
  79         return false;
  80 }
  81
  82 bool rxrpc_set_call_completion(struct rxrpc_call *call,
  83                                enum rxrpc_call_completion compl,
  84                                u32 abort_code,
  85                                int error)
  86 {
  87         bool ret = false;
  88
  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);
  93         }
  94         return ret;
  95 }
  96
  97 /*
  98  * Record that a call successfully completed.
  99  */
 100 bool __rxrpc_call_completed(struct rxrpc_call *call)
 101 {
 102         return __rxrpc_set_call_completion(call, RXRPC_CALL_SUCCEEDED, 0, 0);
 103 }
 104
 105 bool rxrpc_call_completed(struct rxrpc_call *call)
 106 {
 107         bool ret = false;
 108
 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);
 113         }
 114         return ret;
 115 }
 116
 117 /*
 118  * Record that a call is locally aborted.
 119  */
 120 bool __rxrpc_abort_call(const char *why, struct rxrpc_call *call,
 121                         rxrpc_seq_t seq, u32 abort_code, int error)
 122 {
 123         trace_rxrpc_abort(call->debug_id, why, call->cid, call->call_id, seq,
 124                           abort_code, error);
 125         return __rxrpc_set_call_completion(call, RXRPC_CALL_LOCALLY_ABORTED,
 126                                            abort_code, error);
 127 }
 128
 129 bool rxrpc_abort_call(const char *why, struct rxrpc_call *call,
 130                       rxrpc_seq_t seq, u32 abort_code, int error)
 131 {
 132         bool ret;
 133
 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);
 137         return ret;
 138 }
 139
 140 /*
 141  * Pass a call terminating message to userspace.
 142  */
 143 static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
 144 {
 145         u32 tmp = 0;
 146         int ret;
 147
 148         switch (call->completion) {
 149         case RXRPC_CALL_SUCCEEDED:
 150                 ret = 0;
 151                 if (rxrpc_is_service_call(call))
 152                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
 153                 break;
 154         case RXRPC_CALL_REMOTELY_ABORTED:
 155                 tmp = call->abort_code;
 156                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
 157                 break;
 158         case RXRPC_CALL_LOCALLY_ABORTED:
 159                 tmp = call->abort_code;
 160                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
 161                 break;
 162         case RXRPC_CALL_NETWORK_ERROR:
 163                 tmp = -call->error;
 164                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
 165                 break;
 166         case RXRPC_CALL_LOCAL_ERROR:
 167                 tmp = -call->error;
 168                 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
 169                 break;
 170         default:
 171                 pr_err("Invalid terminal call state %u\n", call->state);
 172                 BUG();
 173                 break;
 174         }
 175
 176         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_terminal, call->rx_hard_ack,
 177                             call->rx_pkt_offset, call->rx_pkt_len, ret);
 178         return ret;
 179 }
 180
 181 /*
 182  * Pass back notification of a new call.  The call is added to the
 183  * to-be-accepted list.  This means that the next call to be accepted might not
 184  * be the last call seen awaiting acceptance, but unless we leave this on the
 185  * front of the queue and block all other messages until someone gives us a
 186  * user_ID for it, there's not a lot we can do.
 187  */
 188 static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
 189                                   struct rxrpc_call *call,
 190                                   struct msghdr *msg, int flags)
 191 {
 192         int tmp = 0, ret;
 193
 194         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
 195
 196         if (ret == 0 && !(flags & MSG_PEEK)) {
 197                 _debug("to be accepted");
 198                 write_lock_bh(&rx->recvmsg_lock);
 199                 list_del_init(&call->recvmsg_link);
 200                 write_unlock_bh(&rx->recvmsg_lock);
 201
 202                 rxrpc_get_call(call, rxrpc_call_got);
 203                 write_lock(&rx->call_lock);
 204                 list_add_tail(&call->accept_link, &rx->to_be_accepted);
 205                 write_unlock(&rx->call_lock);
 206         }
 207
 208         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_to_be_accepted, 1, 0, 0, ret);
 209         return ret;
 210 }
 211
 212 /*
 213  * End the packet reception phase.
 214  */
 215 static void rxrpc_end_rx_phase(struct rxrpc_call *call, rxrpc_serial_t serial)
 216 {
 217         _enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
 218
 219         trace_rxrpc_receive(call, rxrpc_receive_end, 0, call->rx_top);
 220         ASSERTCMP(call->rx_hard_ack, ==, call->rx_top);
 221
 222         if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
 223                 rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, serial, false, true,
 224                                   rxrpc_propose_ack_terminal_ack);
 225                 //rxrpc_send_ack_packet(call, false, NULL);
 226         }
 227
 228         write_lock_bh(&call->state_lock);
 229
 230         switch (call->state) {
 231         case RXRPC_CALL_CLIENT_RECV_REPLY:
 232                 __rxrpc_call_completed(call);
 233                 write_unlock_bh(&call->state_lock);
 234                 break;
 235
 236         case RXRPC_CALL_SERVER_RECV_REQUEST:
 237                 call->tx_phase = true;
 238                 call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
 239                 call->expect_req_by = jiffies + MAX_JIFFY_OFFSET;
 240                 write_unlock_bh(&call->state_lock);
 241                 rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial, false, true,
 242                                   rxrpc_propose_ack_processing_op);
 243                 break;
 244         default:
 245                 write_unlock_bh(&call->state_lock);
 246                 break;
 247         }
 248 }
 249
 250 /*
 251  * Discard a packet we've used up and advance the Rx window by one.
 252  */
 253 static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
 254 {
 255         struct rxrpc_skb_priv *sp;
 256         struct sk_buff *skb;
 257         rxrpc_serial_t serial;
 258         rxrpc_seq_t hard_ack, top;
 259         bool last = false;
 260         u8 subpacket;
 261         int ix;
 262
 263         _enter("%d", call->debug_id);
 264
 265         hard_ack = call->rx_hard_ack;
 266         top = smp_load_acquire(&call->rx_top);
 267         ASSERT(before(hard_ack, top));
 268
 269         hard_ack++;
 270         ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
 271         skb = call->rxtx_buffer[ix];
 272         rxrpc_see_skb(skb, rxrpc_skb_rotated);
 273         sp = rxrpc_skb(skb);
 274
 275         subpacket = call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
 276         serial = sp->hdr.serial + subpacket;
 277
 278         if (subpacket == sp->nr_subpackets - 1 &&
 279             sp->rx_flags & RXRPC_SKB_INCL_LAST)
 280                 last = true;
 281
 282         call->rxtx_buffer[ix] = NULL;
 283         call->rxtx_annotations[ix] = 0;
 284         /* Barrier against rxrpc_input_data(). */
 285         smp_store_release(&call->rx_hard_ack, hard_ack);
 286
 287         rxrpc_free_skb(skb, rxrpc_skb_freed);
 288
 289         trace_rxrpc_receive(call, rxrpc_receive_rotate, serial, hard_ack);
 290         if (last) {
 291                 rxrpc_end_rx_phase(call, serial);
 292         } else {
 293                 /* Check to see if there's an ACK that needs sending. */
 294                 if (after_eq(hard_ack, call->ackr_consumed + 2) ||
 295                     after_eq(top, call->ackr_seen + 2) ||
 296                     (hard_ack == top && after(hard_ack, call->ackr_consumed)))
 297                         rxrpc_propose_ACK(call, RXRPC_ACK_DELAY, serial,
 298                                           true, true,
 299                                           rxrpc_propose_ack_rotate_rx);
 300                 if (call->ackr_reason && call->ackr_reason != RXRPC_ACK_DELAY)
 301                         rxrpc_send_ack_packet(call, false, NULL);
 302         }
 303 }
 304
 305 /*
 306  * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
 307  * padding, but if this is the case, the packet length will be resident in the
 308  * socket buffer.  Note that we can't modify the master skb info as the skb may
 309  * be the home to multiple subpackets.
 310  */
 311 static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
 312                                u8 annotation,
 313                                unsigned int offset, unsigned int len)
 314 {
 315         struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
 316         rxrpc_seq_t seq = sp->hdr.seq;
 317         u16 cksum = sp->hdr.cksum;
 318         u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
 319
 320         _enter("");
 321
 322         /* For all but the head jumbo subpacket, the security checksum is in a
 323          * jumbo header immediately prior to the data.
 324          */
 325         if (subpacket > 0) {
 326                 __be16 tmp;
 327                 if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
 328                         BUG();
 329                 cksum = ntohs(tmp);
 330                 seq += subpacket;
 331         }
 332
 333         return call->security->verify_packet(call, skb, offset, len,
 334                                              seq, cksum);
 335 }
 336
 337 /*
 338  * Locate the data within a packet.  This is complicated by:
 339  *
 340  * (1) An skb may contain a jumbo packet - so we have to find the appropriate
 341  *     subpacket.
 342  *
 343  * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
 344  *     contains an extra header which includes the true length of the data,
 345  *     excluding any encrypted padding.
 346  */
 347 static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
 348                              u8 *_annotation,
 349                              unsigned int *_offset, unsigned int *_len,
 350                              bool *_last)
 351 {
 352         struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
 353         unsigned int offset = sizeof(struct rxrpc_wire_header);
 354         unsigned int len;
 355         bool last = false;
 356         int ret;
 357         u8 annotation = *_annotation;
 358         u8 subpacket = annotation & RXRPC_RX_ANNO_SUBPACKET;
 359
 360         /* Locate the subpacket */
 361         offset += subpacket * RXRPC_JUMBO_SUBPKTLEN;
 362         len = skb->len - offset;
 363         if (subpacket < sp->nr_subpackets - 1)
 364                 len = RXRPC_JUMBO_DATALEN;
 365         else if (sp->rx_flags & RXRPC_SKB_INCL_LAST)
 366                 last = true;
 367
 368         if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
 369                 ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
 370                 if (ret < 0)
 371                         return ret;
 372                 *_annotation |= RXRPC_RX_ANNO_VERIFIED;
 373         }
 374
 375         *_offset = offset;
 376         *_len = len;
 377         *_last = last;
 378         call->security->locate_data(call, skb, _offset, _len);
 379         return 0;
 380 }
 381
 382 /*
 383  * Deliver messages to a call.  This keeps processing packets until the buffer
 384  * is filled and we find either more DATA (returns 0) or the end of the DATA
 385  * (returns 1).  If more packets are required, it returns -EAGAIN.
 386  */
 387 static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
 388                               struct msghdr *msg, struct iov_iter *iter,
 389                               size_t len, int flags, size_t *_offset)
 390 {
 391         struct rxrpc_skb_priv *sp;
 392         struct sk_buff *skb;
 393         rxrpc_serial_t serial;
 394         rxrpc_seq_t hard_ack, top, seq;
 395         size_t remain;
 396         bool rx_pkt_last;
 397         unsigned int rx_pkt_offset, rx_pkt_len;
 398         int ix, copy, ret = -EAGAIN, ret2;
 399
 400         if (test_and_clear_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags) &&
 401             call->ackr_reason)
 402                 rxrpc_send_ack_packet(call, false, NULL);
 403
 404         rx_pkt_offset = call->rx_pkt_offset;
 405         rx_pkt_len = call->rx_pkt_len;
 406         rx_pkt_last = call->rx_pkt_last;
 407
 408         if (call->state >= RXRPC_CALL_SERVER_ACK_REQUEST) {
 409                 seq = call->rx_hard_ack;
 410                 ret = 1;
 411                 goto done;
 412         }
 413
 414         /* Barriers against rxrpc_input_data(). */
 415         hard_ack = call->rx_hard_ack;
 416         seq = hard_ack + 1;
 417
 418         while (top = smp_load_acquire(&call->rx_top),
 419                before_eq(seq, top)
 420                ) {
 421                 ix = seq & RXRPC_RXTX_BUFF_MASK;
 422                 skb = call->rxtx_buffer[ix];
 423                 if (!skb) {
 424                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_hole, seq,
 425                                             rx_pkt_offset, rx_pkt_len, 0);
 426                         break;
 427                 }
 428                 smp_rmb();
 429                 rxrpc_see_skb(skb, rxrpc_skb_seen);
 430                 sp = rxrpc_skb(skb);
 431
 432                 if (!(flags & MSG_PEEK)) {
 433                         serial = sp->hdr.serial;
 434                         serial += call->rxtx_annotations[ix] & RXRPC_RX_ANNO_SUBPACKET;
 435                         trace_rxrpc_receive(call, rxrpc_receive_front,
 436                                             serial, seq);
 437                 }
 438
 439                 if (msg)
 440                         sock_recv_timestamp(msg, sock->sk, skb);
 441
 442                 if (rx_pkt_offset == 0) {
 443                         ret2 = rxrpc_locate_data(call, skb,
 444                                                  &call->rxtx_annotations[ix],
 445                                                  &rx_pkt_offset, &rx_pkt_len,
 446                                                  &rx_pkt_last);
 447                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_next, seq,
 448                                             rx_pkt_offset, rx_pkt_len, ret2);
 449                         if (ret2 < 0) {
 450                                 ret = ret2;
 451                                 goto out;
 452                         }
 453                 } else {
 454                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_cont, seq,
 455                                             rx_pkt_offset, rx_pkt_len, 0);
 456                 }
 457
 458                 /* We have to handle short, empty and used-up DATA packets. */
 459                 remain = len - *_offset;
 460                 copy = rx_pkt_len;
 461                 if (copy > remain)
 462                         copy = remain;
 463                 if (copy > 0) {
 464                         ret2 = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
 465                                                       copy);
 466                         if (ret2 < 0) {
 467                                 ret = ret2;
 468                                 goto out;
 469                         }
 470
 471                         /* handle piecemeal consumption of data packets */
 472                         rx_pkt_offset += copy;
 473                         rx_pkt_len -= copy;
 474                         *_offset += copy;
 475                 }
 476
 477                 if (rx_pkt_len > 0) {
 478                         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_full, seq,
 479                                             rx_pkt_offset, rx_pkt_len, 0);
 480                         ASSERTCMP(*_offset, ==, len);
 481                         ret = 0;
 482                         break;
 483                 }
 484
 485                 /* The whole packet has been transferred. */
 486                 if (!(flags & MSG_PEEK))
 487                         rxrpc_rotate_rx_window(call);
 488                 rx_pkt_offset = 0;
 489                 rx_pkt_len = 0;
 490
 491                 if (rx_pkt_last) {
 492                         ASSERTCMP(seq, ==, READ_ONCE(call->rx_top));
 493                         ret = 1;
 494                         goto out;
 495                 }
 496
 497                 seq++;
 498         }
 499
 500 out:
 501         if (!(flags & MSG_PEEK)) {
 502                 call->rx_pkt_offset = rx_pkt_offset;
 503                 call->rx_pkt_len = rx_pkt_len;
 504                 call->rx_pkt_last = rx_pkt_last;
 505         }
 506 done:
 507         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_data_return, seq,
 508                             rx_pkt_offset, rx_pkt_len, ret);
 509         if (ret == -EAGAIN)
 510                 set_bit(RXRPC_CALL_RX_UNDERRUN, &call->flags);
 511         return ret;
 512 }
 513
 514 /*
 515  * Receive a message from an RxRPC socket
 516  * - we need to be careful about two or more threads calling recvmsg
 517  *   simultaneously
 518  */
 519 int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
 520                   int flags)
 521 {
 522         struct rxrpc_call *call;
 523         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 524         struct list_head *l;
 525         size_t copied = 0;
 526         long timeo;
 527         int ret;
 528
 529         DEFINE_WAIT(wait);
 530
 531         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_enter, 0, 0, 0, 0);
 532
 533         if (flags & (MSG_OOB | MSG_TRUNC))
 534                 return -EOPNOTSUPP;
 535
 536         timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
 537
 538 try_again:
 539         lock_sock(&rx->sk);
 540
 541         /* Return immediately if a client socket has no outstanding calls */
 542         if (RB_EMPTY_ROOT(&rx->calls) &&
 543             list_empty(&rx->recvmsg_q) &&
 544             rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
 545                 release_sock(&rx->sk);
 546                 return -ENODATA;
 547         }
 548
 549         if (list_empty(&rx->recvmsg_q)) {
 550                 ret = -EWOULDBLOCK;
 551                 if (timeo == 0) {
 552                         call = NULL;
 553                         goto error_no_call;
 554                 }
 555
 556                 release_sock(&rx->sk);
 557
 558                 /* Wait for something to happen */
 559                 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
 560                                           TASK_INTERRUPTIBLE);
 561                 ret = sock_error(&rx->sk);
 562                 if (ret)
 563                         goto wait_error;
 564
 565                 if (list_empty(&rx->recvmsg_q)) {
 566                         if (signal_pending(current))
 567                                 goto wait_interrupted;
 568                         trace_rxrpc_recvmsg(NULL, rxrpc_recvmsg_wait,
 569                                             0, 0, 0, 0);
 570                         timeo = schedule_timeout(timeo);
 571                 }
 572                 finish_wait(sk_sleep(&rx->sk), &wait);
 573                 goto try_again;
 574         }
 575
 576         /* Find the next call and dequeue it if we're not just peeking.  If we
 577          * do dequeue it, that comes with a ref that we will need to release.
 578          */
 579         write_lock_bh(&rx->recvmsg_lock);
 580         l = rx->recvmsg_q.next;
 581         call = list_entry(l, struct rxrpc_call, recvmsg_link);
 582         if (!(flags & MSG_PEEK))
 583                 list_del_init(&call->recvmsg_link);
 584         else
 585                 rxrpc_get_call(call, rxrpc_call_got);
 586         write_unlock_bh(&rx->recvmsg_lock);
 587
 588         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_dequeue, 0, 0, 0, 0);
 589
 590         /* We're going to drop the socket lock, so we need to lock the call
 591          * against interference by sendmsg.
 592          */
 593         if (!mutex_trylock(&call->user_mutex)) {
 594                 ret = -EWOULDBLOCK;
 595                 if (flags & MSG_DONTWAIT)
 596                         goto error_requeue_call;
 597                 ret = -ERESTARTSYS;
 598                 if (mutex_lock_interruptible(&call->user_mutex) < 0)
 599                         goto error_requeue_call;
 600         }
 601
 602         release_sock(&rx->sk);
 603
 604         if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
 605                 BUG();
 606
 607         if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
 608                 if (flags & MSG_CMSG_COMPAT) {
 609                         unsigned int id32 = call->user_call_ID;
 610
 611                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
 612                                        sizeof(unsigned int), &id32);
 613                 } else {
 614                         unsigned long idl = call->user_call_ID;
 615
 616                         ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
 617                                        sizeof(unsigned long), &idl);
 618                 }
 619                 if (ret < 0)
 620                         goto error_unlock_call;
 621         }
 622
 623         if (msg->msg_name) {
 624                 struct sockaddr_rxrpc *srx = msg->msg_name;
 625                 size_t len = sizeof(call->peer->srx);
 626
 627                 memcpy(msg->msg_name, &call->peer->srx, len);
 628                 srx->srx_service = call->service_id;
 629                 msg->msg_namelen = len;
 630         }
 631
 632         switch (READ_ONCE(call->state)) {
 633         case RXRPC_CALL_SERVER_ACCEPTING:
 634                 ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
 635                 break;
 636         case RXRPC_CALL_CLIENT_RECV_REPLY:
 637         case RXRPC_CALL_SERVER_RECV_REQUEST:
 638         case RXRPC_CALL_SERVER_ACK_REQUEST:
 639                 ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
 640                                          flags, &copied);
 641                 if (ret == -EAGAIN)
 642                         ret = 0;
 643
 644                 if (after(call->rx_top, call->rx_hard_ack) &&
 645                     call->rxtx_buffer[(call->rx_hard_ack + 1) & RXRPC_RXTX_BUFF_MASK])
 646                         rxrpc_notify_socket(call);
 647                 break;
 648         default:
 649                 ret = 0;
 650                 break;
 651         }
 652
 653         if (ret < 0)
 654                 goto error_unlock_call;
 655
 656         if (call->state == RXRPC_CALL_COMPLETE) {
 657                 ret = rxrpc_recvmsg_term(call, msg);
 658                 if (ret < 0)
 659                         goto error_unlock_call;
 660                 if (!(flags & MSG_PEEK))
 661                         rxrpc_release_call(rx, call);
 662                 msg->msg_flags |= MSG_EOR;
 663                 ret = 1;
 664         }
 665
 666         if (ret == 0)
 667                 msg->msg_flags |= MSG_MORE;
 668         else
 669                 msg->msg_flags &= ~MSG_MORE;
 670         ret = copied;
 671
 672 error_unlock_call:
 673         mutex_unlock(&call->user_mutex);
 674         rxrpc_put_call(call, rxrpc_call_put);
 675         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
 676         return ret;
 677
 678 error_requeue_call:
 679         if (!(flags & MSG_PEEK)) {
 680                 write_lock_bh(&rx->recvmsg_lock);
 681                 list_add(&call->recvmsg_link, &rx->recvmsg_q);
 682                 write_unlock_bh(&rx->recvmsg_lock);
 683                 trace_rxrpc_recvmsg(call, rxrpc_recvmsg_requeue, 0, 0, 0, 0);
 684         } else {
 685                 rxrpc_put_call(call, rxrpc_call_put);
 686         }
 687 error_no_call:
 688         release_sock(&rx->sk);
 689 error_trace:
 690         trace_rxrpc_recvmsg(call, rxrpc_recvmsg_return, 0, 0, 0, ret);
 691         return ret;
 692
 693 wait_interrupted:
 694         ret = sock_intr_errno(timeo);
 695 wait_error:
 696         finish_wait(sk_sleep(&rx->sk), &wait);
 697         call = NULL;
 698         goto error_trace;
 699 }
 700
 701 /**
 702  * rxrpc_kernel_recv_data - Allow a kernel service to receive data/info
 703  * @sock: The socket that the call exists on
 704  * @call: The call to send data through
 705  * @iter: The buffer to receive into
 706  * @want_more: True if more data is expected to be read
 707  * @_abort: Where the abort code is stored if -ECONNABORTED is returned
 708  * @_service: Where to store the actual service ID (may be upgraded)
 709  *
 710  * Allow a kernel service to receive data and pick up information about the
 711  * state of a call.  Returns 0 if got what was asked for and there's more
 712  * available, 1 if we got what was asked for and we're at the end of the data
 713  * and -EAGAIN if we need more data.
 714  *
 715  * Note that we may return -EAGAIN to drain empty packets at the end of the
 716  * data, even if we've already copied over the requested data.
 717  *
 718  * *_abort should also be initialised to 0.
 719  */
 720 int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
 721                            struct iov_iter *iter,
 722                            bool want_more, u32 *_abort, u16 *_service)
 723 {
 724         size_t offset = 0;
 725         int ret;
 726
 727         _enter("{%d,%s},%zu,%d",
 728                call->debug_id, rxrpc_call_states[call->state],
 729                iov_iter_count(iter), want_more);
 730
 731         ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
 732
 733         mutex_lock(&call->user_mutex);
 734
 735         switch (READ_ONCE(call->state)) {
 736         case RXRPC_CALL_CLIENT_RECV_REPLY:
 737         case RXRPC_CALL_SERVER_RECV_REQUEST:
 738         case RXRPC_CALL_SERVER_ACK_REQUEST:
 739                 ret = rxrpc_recvmsg_data(sock, call, NULL, iter,
 740                                          iov_iter_count(iter), 0,
 741                                          &offset);
 742                 if (ret < 0)
 743                         goto out;
 744
 745                 /* We can only reach here with a partially full buffer if we
 746                  * have reached the end of the data.  We must otherwise have a
 747                  * full buffer or have been given -EAGAIN.
 748                  */
 749                 if (ret == 1) {
 750                         if (iov_iter_count(iter) > 0)
 751                                 goto short_data;
 752                         if (!want_more)
 753                                 goto read_phase_complete;
 754                         ret = 0;
 755                         goto out;
 756                 }
 757
 758                 if (!want_more)
 759                         goto excess_data;
 760                 goto out;
 761
 762         case RXRPC_CALL_COMPLETE:
 763                 goto call_complete;
 764
 765         default:
 766                 ret = -EINPROGRESS;
 767                 goto out;
 768         }
 769
 770 read_phase_complete:
 771         ret = 1;
 772 out:
 773         switch (call->ackr_reason) {
 774         case RXRPC_ACK_IDLE:
 775                 break;
 776         case RXRPC_ACK_DELAY:
 777                 if (ret != -EAGAIN)
 778                         break;
 779                 /* Fall through */
 780         default:
 781                 rxrpc_send_ack_packet(call, false, NULL);
 782         }
 783
 784         if (_service)
 785                 *_service = call->service_id;
 786         mutex_unlock(&call->user_mutex);
 787         _leave(" = %d [%zu,%d]", ret, iov_iter_count(iter), *_abort);
 788         return ret;
 789
 790 short_data:
 791         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("short_data"));
 792         ret = -EBADMSG;
 793         goto out;
 794 excess_data:
 795         trace_rxrpc_rx_eproto(call, 0, tracepoint_string("excess_data"));
 796         ret = -EMSGSIZE;
 797         goto out;
 798 call_complete:
 799         *_abort = call->abort_code;
 800         ret = call->error;
 801         if (call->completion == RXRPC_CALL_SUCCEEDED) {
 802                 ret = 1;
 803                 if (iov_iter_count(iter) > 0)
 804                         ret = -ECONNRESET;
 805         }
 806         goto out;
 807 }
 808 EXPORT_SYMBOL(rxrpc_kernel_recv_data);
 809
 810 /**
 811  * rxrpc_kernel_get_reply_time - Get timestamp on first reply packet
 812  * @sock: The socket that the call exists on
 813  * @call: The call to query
 814  * @_ts: Where to put the timestamp
 815  *
 816  * Retrieve the timestamp from the first DATA packet of the reply if it is
 817  * in the ring.  Returns true if successful, false if not.
 818  */
 819 bool rxrpc_kernel_get_reply_time(struct socket *sock, struct rxrpc_call *call,
 820                                  ktime_t *_ts)
 821 {
 822         struct sk_buff *skb;
 823         rxrpc_seq_t hard_ack, top, seq;
 824         bool success = false;
 825
 826         mutex_lock(&call->user_mutex);
 827
 828         if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_RECV_REPLY)
 829                 goto out;
 830
 831         hard_ack = call->rx_hard_ack;
 832         if (hard_ack != 0)
 833                 goto out;
 834
 835         seq = hard_ack + 1;
 836         top = smp_load_acquire(&call->rx_top);
 837         if (after(seq, top))
 838                 goto out;
 839
 840         skb = call->rxtx_buffer[seq & RXRPC_RXTX_BUFF_MASK];
 841         if (!skb)
 842                 goto out;
 843
 844         *_ts = skb_get_ktime(skb);
 845         success = true;
 846
 847 out:
 848         mutex_unlock(&call->user_mutex);
 849         return success;
 850 }
 851 EXPORT_SYMBOL(rxrpc_kernel_get_reply_time);