net/vmw_vsock/virtio_transport_common.c

   1 /*
   2  * common code for virtio vsock
   3  *
   4  * Copyright (C) 2013-2015 Red Hat, Inc.
   5  * Author: Asias He <asias@redhat.com>
   6  *         Stefan Hajnoczi <stefanha@redhat.com>
   7  *
   8  * This work is licensed under the terms of the GNU GPL, version 2.
   9  */
  10 #include <linux/spinlock.h>
  11 #include <linux/module.h>
  12 #include <linux/sched/signal.h>
  13 #include <linux/ctype.h>
  14 #include <linux/list.h>
  15 #include <linux/virtio.h>
  16 #include <linux/virtio_ids.h>
  17 #include <linux/virtio_config.h>
  18 #include <linux/virtio_vsock.h>
  19 #include <uapi/linux/vsockmon.h>
  20
  21 #include <net/sock.h>
  22 #include <net/af_vsock.h>
  23
  24 #define CREATE_TRACE_POINTS
  25 #include <trace/events/vsock_virtio_transport_common.h>
  26
  27 /* How long to wait for graceful shutdown of a connection */
  28 #define VSOCK_CLOSE_TIMEOUT (8 * HZ)
  29
  30 static const struct virtio_transport *virtio_transport_get_ops(void)
  31 {
  32         const struct vsock_transport *t = vsock_core_get_transport();
  33
  34         return container_of(t, struct virtio_transport, transport);
  35 }
  36
  37 static struct virtio_vsock_pkt *
  38 virtio_transport_alloc_pkt(struct virtio_vsock_pkt_info *info,
  39                            size_t len,
  40                            u32 src_cid,
  41                            u32 src_port,
  42                            u32 dst_cid,
  43                            u32 dst_port)
  44 {
  45         struct virtio_vsock_pkt *pkt;
  46         int err;
  47
  48         pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
  49         if (!pkt)
  50                 return NULL;
  51
  52         pkt->hdr.type           = cpu_to_le16(info->type);
  53         pkt->hdr.op             = cpu_to_le16(info->op);
  54         pkt->hdr.src_cid        = cpu_to_le64(src_cid);
  55         pkt->hdr.dst_cid        = cpu_to_le64(dst_cid);
  56         pkt->hdr.src_port       = cpu_to_le32(src_port);
  57         pkt->hdr.dst_port       = cpu_to_le32(dst_port);
  58         pkt->hdr.flags          = cpu_to_le32(info->flags);
  59         pkt->len                = len;
  60         pkt->hdr.len            = cpu_to_le32(len);
  61         pkt->reply              = info->reply;
  62         pkt->vsk                = info->vsk;
  63
  64         if (info->msg && len > 0) {
  65                 pkt->buf = kmalloc(len, GFP_KERNEL);
  66                 if (!pkt->buf)
  67                         goto out_pkt;
  68                 err = memcpy_from_msg(pkt->buf, info->msg, len);
  69                 if (err)
  70                         goto out;
  71         }
  72
  73         trace_virtio_transport_alloc_pkt(src_cid, src_port,
  74                                          dst_cid, dst_port,
  75                                          len,
  76                                          info->type,
  77                                          info->op,
  78                                          info->flags);
  79
  80         return pkt;
  81
  82 out:
  83         kfree(pkt->buf);
  84 out_pkt:
  85         kfree(pkt);
  86         return NULL;
  87 }
  88
  89 /* Packet capture */
  90 static struct sk_buff *virtio_transport_build_skb(void *opaque)
  91 {
  92         struct virtio_vsock_pkt *pkt = opaque;
  93         struct af_vsockmon_hdr *hdr;
  94         struct sk_buff *skb;
  95
  96         skb = alloc_skb(sizeof(*hdr) + sizeof(pkt->hdr) + pkt->len,
  97                         GFP_ATOMIC);
  98         if (!skb)
  99                 return NULL;
 100
 101         hdr = skb_put(skb, sizeof(*hdr));
 102
 103         /* pkt->hdr is little-endian so no need to byteswap here */
 104         hdr->src_cid = pkt->hdr.src_cid;
 105         hdr->src_port = pkt->hdr.src_port;
 106         hdr->dst_cid = pkt->hdr.dst_cid;
 107         hdr->dst_port = pkt->hdr.dst_port;
 108
 109         hdr->transport = cpu_to_le16(AF_VSOCK_TRANSPORT_VIRTIO);
 110         hdr->len = cpu_to_le16(sizeof(pkt->hdr));
 111         memset(hdr->reserved, 0, sizeof(hdr->reserved));
 112
 113         switch (le16_to_cpu(pkt->hdr.op)) {
 114         case VIRTIO_VSOCK_OP_REQUEST:
 115         case VIRTIO_VSOCK_OP_RESPONSE:
 116                 hdr->op = cpu_to_le16(AF_VSOCK_OP_CONNECT);
 117                 break;
 118         case VIRTIO_VSOCK_OP_RST:
 119         case VIRTIO_VSOCK_OP_SHUTDOWN:
 120                 hdr->op = cpu_to_le16(AF_VSOCK_OP_DISCONNECT);
 121                 break;
 122         case VIRTIO_VSOCK_OP_RW:
 123                 hdr->op = cpu_to_le16(AF_VSOCK_OP_PAYLOAD);
 124                 break;
 125         case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
 126         case VIRTIO_VSOCK_OP_CREDIT_REQUEST:
 127                 hdr->op = cpu_to_le16(AF_VSOCK_OP_CONTROL);
 128                 break;
 129         default:
 130                 hdr->op = cpu_to_le16(AF_VSOCK_OP_UNKNOWN);
 131                 break;
 132         }
 133
 134         skb_put_data(skb, &pkt->hdr, sizeof(pkt->hdr));
 135
 136         if (pkt->len) {
 137                 skb_put_data(skb, pkt->buf, pkt->len);
 138         }
 139
 140         return skb;
 141 }
 142
 143 void virtio_transport_deliver_tap_pkt(struct virtio_vsock_pkt *pkt)
 144 {
 145         vsock_deliver_tap(virtio_transport_build_skb, pkt);
 146 }
 147 EXPORT_SYMBOL_GPL(virtio_transport_deliver_tap_pkt);
 148
 149 static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
 150                                           struct virtio_vsock_pkt_info *info)
 151 {
 152         u32 src_cid, src_port, dst_cid, dst_port;
 153         struct virtio_vsock_sock *vvs;
 154         struct virtio_vsock_pkt *pkt;
 155         u32 pkt_len = info->pkt_len;
 156
 157         src_cid = vm_sockets_get_local_cid();
 158         src_port = vsk->local_addr.svm_port;
 159         if (!info->remote_cid) {
 160                 dst_cid = vsk->remote_addr.svm_cid;
 161                 dst_port = vsk->remote_addr.svm_port;
 162         } else {
 163                 dst_cid = info->remote_cid;
 164                 dst_port = info->remote_port;
 165         }
 166
 167         vvs = vsk->trans;
 168
 169         /* we can send less than pkt_len bytes */
 170         if (pkt_len > VIRTIO_VSOCK_DEFAULT_RX_BUF_SIZE)
 171                 pkt_len = VIRTIO_VSOCK_DEFAULT_RX_BUF_SIZE;
 172
 173         /* virtio_transport_get_credit might return less than pkt_len credit */
 174         pkt_len = virtio_transport_get_credit(vvs, pkt_len);
 175
 176         /* Do not send zero length OP_RW pkt */
 177         if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
 178                 return pkt_len;
 179
 180         pkt = virtio_transport_alloc_pkt(info, pkt_len,
 181                                          src_cid, src_port,
 182                                          dst_cid, dst_port);
 183         if (!pkt) {
 184                 virtio_transport_put_credit(vvs, pkt_len);
 185                 return -ENOMEM;
 186         }
 187
 188         virtio_transport_inc_tx_pkt(vvs, pkt);
 189
 190         return virtio_transport_get_ops()->send_pkt(pkt);
 191 }
 192
 193 static void virtio_transport_inc_rx_pkt(struct virtio_vsock_sock *vvs,
 194                                         struct virtio_vsock_pkt *pkt)
 195 {
 196         vvs->rx_bytes += pkt->len;
 197 }
 198
 199 static void virtio_transport_dec_rx_pkt(struct virtio_vsock_sock *vvs,
 200                                         struct virtio_vsock_pkt *pkt)
 201 {
 202         vvs->rx_bytes -= pkt->len;
 203         vvs->fwd_cnt += pkt->len;
 204 }
 205
 206 void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs, struct virtio_vsock_pkt *pkt)
 207 {
 208         spin_lock_bh(&vvs->tx_lock);
 209         pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
 210         pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
 211         spin_unlock_bh(&vvs->tx_lock);
 212 }
 213 EXPORT_SYMBOL_GPL(virtio_transport_inc_tx_pkt);
 214
 215 u32 virtio_transport_get_credit(struct virtio_vsock_sock *vvs, u32 credit)
 216 {
 217         u32 ret;
 218
 219         spin_lock_bh(&vvs->tx_lock);
 220         ret = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
 221         if (ret > credit)
 222                 ret = credit;
 223         vvs->tx_cnt += ret;
 224         spin_unlock_bh(&vvs->tx_lock);
 225
 226         return ret;
 227 }
 228 EXPORT_SYMBOL_GPL(virtio_transport_get_credit);
 229
 230 void virtio_transport_put_credit(struct virtio_vsock_sock *vvs, u32 credit)
 231 {
 232         spin_lock_bh(&vvs->tx_lock);
 233         vvs->tx_cnt -= credit;
 234         spin_unlock_bh(&vvs->tx_lock);
 235 }
 236 EXPORT_SYMBOL_GPL(virtio_transport_put_credit);
 237
 238 static int virtio_transport_send_credit_update(struct vsock_sock *vsk,
 239                                                int type,
 240                                                struct virtio_vsock_hdr *hdr)
 241 {
 242         struct virtio_vsock_pkt_info info = {
 243                 .op = VIRTIO_VSOCK_OP_CREDIT_UPDATE,
 244                 .type = type,
 245                 .vsk = vsk,
 246         };
 247
 248         return virtio_transport_send_pkt_info(vsk, &info);
 249 }
 250
 251 static ssize_t
 252 virtio_transport_stream_do_dequeue(struct vsock_sock *vsk,
 253                                    struct msghdr *msg,
 254                                    size_t len)
 255 {
 256         struct virtio_vsock_sock *vvs = vsk->trans;
 257         struct virtio_vsock_pkt *pkt;
 258         size_t bytes, total = 0;
 259         int err = -EFAULT;
 260
 261         spin_lock_bh(&vvs->rx_lock);
 262         while (total < len && !list_empty(&vvs->rx_queue)) {
 263                 pkt = list_first_entry(&vvs->rx_queue,
 264                                        struct virtio_vsock_pkt, list);
 265
 266                 bytes = len - total;
 267                 if (bytes > pkt->len - pkt->off)
 268                         bytes = pkt->len - pkt->off;
 269
 270                 /* sk_lock is held by caller so no one else can dequeue.
 271                  * Unlock rx_lock since memcpy_to_msg() may sleep.
 272                  */
 273                 spin_unlock_bh(&vvs->rx_lock);
 274
 275                 err = memcpy_to_msg(msg, pkt->buf + pkt->off, bytes);
 276                 if (err)
 277                         goto out;
 278
 279                 spin_lock_bh(&vvs->rx_lock);
 280
 281                 total += bytes;
 282                 pkt->off += bytes;
 283                 if (pkt->off == pkt->len) {
 284                         virtio_transport_dec_rx_pkt(vvs, pkt);
 285                         list_del(&pkt->list);
 286                         virtio_transport_free_pkt(pkt);
 287                 }
 288         }
 289         spin_unlock_bh(&vvs->rx_lock);
 290
 291         /* Send a credit pkt to peer */
 292         virtio_transport_send_credit_update(vsk, VIRTIO_VSOCK_TYPE_STREAM,
 293                                             NULL);
 294
 295         return total;
 296
 297 out:
 298         if (total)
 299                 err = total;
 300         return err;
 301 }
 302
 303 ssize_t
 304 virtio_transport_stream_dequeue(struct vsock_sock *vsk,
 305                                 struct msghdr *msg,
 306                                 size_t len, int flags)
 307 {
 308         if (flags & MSG_PEEK)
 309                 return -EOPNOTSUPP;
 310
 311         return virtio_transport_stream_do_dequeue(vsk, msg, len);
 312 }
 313 EXPORT_SYMBOL_GPL(virtio_transport_stream_dequeue);
 314
 315 int
 316 virtio_transport_dgram_dequeue(struct vsock_sock *vsk,
 317                                struct msghdr *msg,
 318                                size_t len, int flags)
 319 {
 320         return -EOPNOTSUPP;
 321 }
 322 EXPORT_SYMBOL_GPL(virtio_transport_dgram_dequeue);
 323
 324 s64 virtio_transport_stream_has_data(struct vsock_sock *vsk)
 325 {
 326         struct virtio_vsock_sock *vvs = vsk->trans;
 327         s64 bytes;
 328
 329         spin_lock_bh(&vvs->rx_lock);
 330         bytes = vvs->rx_bytes;
 331         spin_unlock_bh(&vvs->rx_lock);
 332
 333         return bytes;
 334 }
 335 EXPORT_SYMBOL_GPL(virtio_transport_stream_has_data);
 336
 337 static s64 virtio_transport_has_space(struct vsock_sock *vsk)
 338 {
 339         struct virtio_vsock_sock *vvs = vsk->trans;
 340         s64 bytes;
 341
 342         bytes = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
 343         if (bytes < 0)
 344                 bytes = 0;
 345
 346         return bytes;
 347 }
 348
 349 s64 virtio_transport_stream_has_space(struct vsock_sock *vsk)
 350 {
 351         struct virtio_vsock_sock *vvs = vsk->trans;
 352         s64 bytes;
 353
 354         spin_lock_bh(&vvs->tx_lock);
 355         bytes = virtio_transport_has_space(vsk);
 356         spin_unlock_bh(&vvs->tx_lock);
 357
 358         return bytes;
 359 }
 360 EXPORT_SYMBOL_GPL(virtio_transport_stream_has_space);
 361
 362 int virtio_transport_do_socket_init(struct vsock_sock *vsk,
 363                                     struct vsock_sock *psk)
 364 {
 365         struct virtio_vsock_sock *vvs;
 366
 367         vvs = kzalloc(sizeof(*vvs), GFP_KERNEL);
 368         if (!vvs)
 369                 return -ENOMEM;
 370
 371         vsk->trans = vvs;
 372         vvs->vsk = vsk;
 373         if (psk) {
 374                 struct virtio_vsock_sock *ptrans = psk->trans;
 375
 376                 vvs->buf_size   = ptrans->buf_size;
 377                 vvs->buf_size_min = ptrans->buf_size_min;
 378                 vvs->buf_size_max = ptrans->buf_size_max;
 379                 vvs->peer_buf_alloc = ptrans->peer_buf_alloc;
 380         } else {
 381                 vvs->buf_size = VIRTIO_VSOCK_DEFAULT_BUF_SIZE;
 382                 vvs->buf_size_min = VIRTIO_VSOCK_DEFAULT_MIN_BUF_SIZE;
 383                 vvs->buf_size_max = VIRTIO_VSOCK_DEFAULT_MAX_BUF_SIZE;
 384         }
 385
 386         vvs->buf_alloc = vvs->buf_size;
 387
 388         spin_lock_init(&vvs->rx_lock);
 389         spin_lock_init(&vvs->tx_lock);
 390         INIT_LIST_HEAD(&vvs->rx_queue);
 391
 392         return 0;
 393 }
 394 EXPORT_SYMBOL_GPL(virtio_transport_do_socket_init);
 395
 396 u64 virtio_transport_get_buffer_size(struct vsock_sock *vsk)
 397 {
 398         struct virtio_vsock_sock *vvs = vsk->trans;
 399
 400         return vvs->buf_size;
 401 }
 402 EXPORT_SYMBOL_GPL(virtio_transport_get_buffer_size);
 403
 404 u64 virtio_transport_get_min_buffer_size(struct vsock_sock *vsk)
 405 {
 406         struct virtio_vsock_sock *vvs = vsk->trans;
 407
 408         return vvs->buf_size_min;
 409 }
 410 EXPORT_SYMBOL_GPL(virtio_transport_get_min_buffer_size);
 411
 412 u64 virtio_transport_get_max_buffer_size(struct vsock_sock *vsk)
 413 {
 414         struct virtio_vsock_sock *vvs = vsk->trans;
 415
 416         return vvs->buf_size_max;
 417 }
 418 EXPORT_SYMBOL_GPL(virtio_transport_get_max_buffer_size);
 419
 420 void virtio_transport_set_buffer_size(struct vsock_sock *vsk, u64 val)
 421 {
 422         struct virtio_vsock_sock *vvs = vsk->trans;
 423
 424         if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
 425                 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
 426         if (val < vvs->buf_size_min)
 427                 vvs->buf_size_min = val;
 428         if (val > vvs->buf_size_max)
 429                 vvs->buf_size_max = val;
 430         vvs->buf_size = val;
 431         vvs->buf_alloc = val;
 432 }
 433 EXPORT_SYMBOL_GPL(virtio_transport_set_buffer_size);
 434
 435 void virtio_transport_set_min_buffer_size(struct vsock_sock *vsk, u64 val)
 436 {
 437         struct virtio_vsock_sock *vvs = vsk->trans;
 438
 439         if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
 440                 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
 441         if (val > vvs->buf_size)
 442                 vvs->buf_size = val;
 443         vvs->buf_size_min = val;
 444 }
 445 EXPORT_SYMBOL_GPL(virtio_transport_set_min_buffer_size);
 446
 447 void virtio_transport_set_max_buffer_size(struct vsock_sock *vsk, u64 val)
 448 {
 449         struct virtio_vsock_sock *vvs = vsk->trans;
 450
 451         if (val > VIRTIO_VSOCK_MAX_BUF_SIZE)
 452                 val = VIRTIO_VSOCK_MAX_BUF_SIZE;
 453         if (val < vvs->buf_size)
 454                 vvs->buf_size = val;
 455         vvs->buf_size_max = val;
 456 }
 457 EXPORT_SYMBOL_GPL(virtio_transport_set_max_buffer_size);
 458
 459 int
 460 virtio_transport_notify_poll_in(struct vsock_sock *vsk,
 461                                 size_t target,
 462                                 bool *data_ready_now)
 463 {
 464         if (vsock_stream_has_data(vsk))
 465                 *data_ready_now = true;
 466         else
 467                 *data_ready_now = false;
 468
 469         return 0;
 470 }
 471 EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_in);
 472
 473 int
 474 virtio_transport_notify_poll_out(struct vsock_sock *vsk,
 475                                  size_t target,
 476                                  bool *space_avail_now)
 477 {
 478         s64 free_space;
 479
 480         free_space = vsock_stream_has_space(vsk);
 481         if (free_space > 0)
 482                 *space_avail_now = true;
 483         else if (free_space == 0)
 484                 *space_avail_now = false;
 485
 486         return 0;
 487 }
 488 EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_out);
 489
 490 int virtio_transport_notify_recv_init(struct vsock_sock *vsk,
 491         size_t target, struct vsock_transport_recv_notify_data *data)
 492 {
 493         return 0;
 494 }
 495 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_init);
 496
 497 int virtio_transport_notify_recv_pre_block(struct vsock_sock *vsk,
 498         size_t target, struct vsock_transport_recv_notify_data *data)
 499 {
 500         return 0;
 501 }
 502 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_block);
 503
 504 int virtio_transport_notify_recv_pre_dequeue(struct vsock_sock *vsk,
 505         size_t target, struct vsock_transport_recv_notify_data *data)
 506 {
 507         return 0;
 508 }
 509 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_dequeue);
 510
 511 int virtio_transport_notify_recv_post_dequeue(struct vsock_sock *vsk,
 512         size_t target, ssize_t copied, bool data_read,
 513         struct vsock_transport_recv_notify_data *data)
 514 {
 515         return 0;
 516 }
 517 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_post_dequeue);
 518
 519 int virtio_transport_notify_send_init(struct vsock_sock *vsk,
 520         struct vsock_transport_send_notify_data *data)
 521 {
 522         return 0;
 523 }
 524 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_init);
 525
 526 int virtio_transport_notify_send_pre_block(struct vsock_sock *vsk,
 527         struct vsock_transport_send_notify_data *data)
 528 {
 529         return 0;
 530 }
 531 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_block);
 532
 533 int virtio_transport_notify_send_pre_enqueue(struct vsock_sock *vsk,
 534         struct vsock_transport_send_notify_data *data)
 535 {
 536         return 0;
 537 }
 538 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_enqueue);
 539
 540 int virtio_transport_notify_send_post_enqueue(struct vsock_sock *vsk,
 541         ssize_t written, struct vsock_transport_send_notify_data *data)
 542 {
 543         return 0;
 544 }
 545 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_post_enqueue);
 546
 547 u64 virtio_transport_stream_rcvhiwat(struct vsock_sock *vsk)
 548 {
 549         struct virtio_vsock_sock *vvs = vsk->trans;
 550
 551         return vvs->buf_size;
 552 }
 553 EXPORT_SYMBOL_GPL(virtio_transport_stream_rcvhiwat);
 554
 555 bool virtio_transport_stream_is_active(struct vsock_sock *vsk)
 556 {
 557         return true;
 558 }
 559 EXPORT_SYMBOL_GPL(virtio_transport_stream_is_active);
 560
 561 bool virtio_transport_stream_allow(u32 cid, u32 port)
 562 {
 563         return true;
 564 }
 565 EXPORT_SYMBOL_GPL(virtio_transport_stream_allow);
 566
 567 int virtio_transport_dgram_bind(struct vsock_sock *vsk,
 568                                 struct sockaddr_vm *addr)
 569 {
 570         return -EOPNOTSUPP;
 571 }
 572 EXPORT_SYMBOL_GPL(virtio_transport_dgram_bind);
 573
 574 bool virtio_transport_dgram_allow(u32 cid, u32 port)
 575 {
 576         return false;
 577 }
 578 EXPORT_SYMBOL_GPL(virtio_transport_dgram_allow);
 579
 580 int virtio_transport_connect(struct vsock_sock *vsk)
 581 {
 582         struct virtio_vsock_pkt_info info = {
 583                 .op = VIRTIO_VSOCK_OP_REQUEST,
 584                 .type = VIRTIO_VSOCK_TYPE_STREAM,
 585                 .vsk = vsk,
 586         };
 587
 588         return virtio_transport_send_pkt_info(vsk, &info);
 589 }
 590 EXPORT_SYMBOL_GPL(virtio_transport_connect);
 591
 592 int virtio_transport_shutdown(struct vsock_sock *vsk, int mode)
 593 {
 594         struct virtio_vsock_pkt_info info = {
 595                 .op = VIRTIO_VSOCK_OP_SHUTDOWN,
 596                 .type = VIRTIO_VSOCK_TYPE_STREAM,
 597                 .flags = (mode & RCV_SHUTDOWN ?
 598                           VIRTIO_VSOCK_SHUTDOWN_RCV : 0) |
 599                          (mode & SEND_SHUTDOWN ?
 600                           VIRTIO_VSOCK_SHUTDOWN_SEND : 0),
 601                 .vsk = vsk,
 602         };
 603
 604         return virtio_transport_send_pkt_info(vsk, &info);
 605 }
 606 EXPORT_SYMBOL_GPL(virtio_transport_shutdown);
 607
 608 int
 609 virtio_transport_dgram_enqueue(struct vsock_sock *vsk,
 610                                struct sockaddr_vm *remote_addr,
 611                                struct msghdr *msg,
 612                                size_t dgram_len)
 613 {
 614         return -EOPNOTSUPP;
 615 }
 616 EXPORT_SYMBOL_GPL(virtio_transport_dgram_enqueue);
 617
 618 ssize_t
 619 virtio_transport_stream_enqueue(struct vsock_sock *vsk,
 620                                 struct msghdr *msg,
 621                                 size_t len)
 622 {
 623         struct virtio_vsock_pkt_info info = {
 624                 .op = VIRTIO_VSOCK_OP_RW,
 625                 .type = VIRTIO_VSOCK_TYPE_STREAM,
 626                 .msg = msg,
 627                 .pkt_len = len,
 628                 .vsk = vsk,
 629         };
 630
 631         return virtio_transport_send_pkt_info(vsk, &info);
 632 }
 633 EXPORT_SYMBOL_GPL(virtio_transport_stream_enqueue);
 634
 635 void virtio_transport_destruct(struct vsock_sock *vsk)
 636 {
 637         struct virtio_vsock_sock *vvs = vsk->trans;
 638
 639         kfree(vvs);
 640 }
 641 EXPORT_SYMBOL_GPL(virtio_transport_destruct);
 642
 643 static int virtio_transport_reset(struct vsock_sock *vsk,
 644                                   struct virtio_vsock_pkt *pkt)
 645 {
 646         struct virtio_vsock_pkt_info info = {
 647                 .op = VIRTIO_VSOCK_OP_RST,
 648                 .type = VIRTIO_VSOCK_TYPE_STREAM,
 649                 .reply = !!pkt,
 650                 .vsk = vsk,
 651         };
 652
 653         /* Send RST only if the original pkt is not a RST pkt */
 654         if (pkt && le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
 655                 return 0;
 656
 657         return virtio_transport_send_pkt_info(vsk, &info);
 658 }
 659
 660 /* Normally packets are associated with a socket.  There may be no socket if an
 661  * attempt was made to connect to a socket that does not exist.
 662  */
 663 static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt)
 664 {
 665         struct virtio_vsock_pkt_info info = {
 666                 .op = VIRTIO_VSOCK_OP_RST,
 667                 .type = le16_to_cpu(pkt->hdr.type),
 668                 .reply = true,
 669         };
 670
 671         /* Send RST only if the original pkt is not a RST pkt */
 672         if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
 673                 return 0;
 674
 675         pkt = virtio_transport_alloc_pkt(&info, 0,
 676                                          le64_to_cpu(pkt->hdr.dst_cid),
 677                                          le32_to_cpu(pkt->hdr.dst_port),
 678                                          le64_to_cpu(pkt->hdr.src_cid),
 679                                          le32_to_cpu(pkt->hdr.src_port));
 680         if (!pkt)
 681                 return -ENOMEM;
 682
 683         return virtio_transport_get_ops()->send_pkt(pkt);
 684 }
 685
 686 static void virtio_transport_wait_close(struct sock *sk, long timeout)
 687 {
 688         if (timeout) {
 689                 DEFINE_WAIT_FUNC(wait, woken_wake_function);
 690
 691                 add_wait_queue(sk_sleep(sk), &wait);
 692
 693                 do {
 694                         if (sk_wait_event(sk, &timeout,
 695                                           sock_flag(sk, SOCK_DONE), &wait))
 696                                 break;
 697                 } while (!signal_pending(current) && timeout);
 698
 699                 remove_wait_queue(sk_sleep(sk), &wait);
 700         }
 701 }
 702
 703 static void virtio_transport_do_close(struct vsock_sock *vsk,
 704                                       bool cancel_timeout)
 705 {
 706         struct sock *sk = sk_vsock(vsk);
 707
 708         sock_set_flag(sk, SOCK_DONE);
 709         vsk->peer_shutdown = SHUTDOWN_MASK;
 710         if (vsock_stream_has_data(vsk) <= 0)
 711                 sk->sk_state = TCP_CLOSING;
 712         sk->sk_state_change(sk);
 713
 714         if (vsk->close_work_scheduled &&
 715             (!cancel_timeout || cancel_delayed_work(&vsk->close_work))) {
 716                 vsk->close_work_scheduled = false;
 717
 718                 vsock_remove_sock(vsk);
 719
 720                 /* Release refcnt obtained when we scheduled the timeout */
 721                 sock_put(sk);
 722         }
 723 }
 724
 725 static void virtio_transport_close_timeout(struct work_struct *work)
 726 {
 727         struct vsock_sock *vsk =
 728                 container_of(work, struct vsock_sock, close_work.work);
 729         struct sock *sk = sk_vsock(vsk);
 730
 731         sock_hold(sk);
 732         lock_sock(sk);
 733
 734         if (!sock_flag(sk, SOCK_DONE)) {
 735                 (void)virtio_transport_reset(vsk, NULL);
 736
 737                 virtio_transport_do_close(vsk, false);
 738         }
 739
 740         vsk->close_work_scheduled = false;
 741
 742         release_sock(sk);
 743         sock_put(sk);
 744 }
 745
 746 /* User context, vsk->sk is locked */
 747 static bool virtio_transport_close(struct vsock_sock *vsk)
 748 {
 749         struct sock *sk = &vsk->sk;
 750
 751         if (!(sk->sk_state == TCP_ESTABLISHED ||
 752               sk->sk_state == TCP_CLOSING))
 753                 return true;
 754
 755         /* Already received SHUTDOWN from peer, reply with RST */
 756         if ((vsk->peer_shutdown & SHUTDOWN_MASK) == SHUTDOWN_MASK) {
 757                 (void)virtio_transport_reset(vsk, NULL);
 758                 return true;
 759         }
 760
 761         if ((sk->sk_shutdown & SHUTDOWN_MASK) != SHUTDOWN_MASK)
 762                 (void)virtio_transport_shutdown(vsk, SHUTDOWN_MASK);
 763
 764         if (sock_flag(sk, SOCK_LINGER) && !(current->flags & PF_EXITING))
 765                 virtio_transport_wait_close(sk, sk->sk_lingertime);
 766
 767         if (sock_flag(sk, SOCK_DONE)) {
 768                 return true;
 769         }
 770
 771         sock_hold(sk);
 772         INIT_DELAYED_WORK(&vsk->close_work,
 773                           virtio_transport_close_timeout);
 774         vsk->close_work_scheduled = true;
 775         schedule_delayed_work(&vsk->close_work, VSOCK_CLOSE_TIMEOUT);
 776         return false;
 777 }
 778
 779 void virtio_transport_release(struct vsock_sock *vsk)
 780 {
 781         struct sock *sk = &vsk->sk;
 782         bool remove_sock = true;
 783
 784         lock_sock(sk);
 785         if (sk->sk_type == SOCK_STREAM)
 786                 remove_sock = virtio_transport_close(vsk);
 787         release_sock(sk);
 788
 789         if (remove_sock)
 790                 vsock_remove_sock(vsk);
 791 }
 792 EXPORT_SYMBOL_GPL(virtio_transport_release);
 793
 794 static int
 795 virtio_transport_recv_connecting(struct sock *sk,
 796                                  struct virtio_vsock_pkt *pkt)
 797 {
 798         struct vsock_sock *vsk = vsock_sk(sk);
 799         int err;
 800         int skerr;
 801
 802         switch (le16_to_cpu(pkt->hdr.op)) {
 803         case VIRTIO_VSOCK_OP_RESPONSE:
 804                 sk->sk_state = TCP_ESTABLISHED;
 805                 sk->sk_socket->state = SS_CONNECTED;
 806                 vsock_insert_connected(vsk);
 807                 sk->sk_state_change(sk);
 808                 break;
 809         case VIRTIO_VSOCK_OP_INVALID:
 810                 break;
 811         case VIRTIO_VSOCK_OP_RST:
 812                 skerr = ECONNRESET;
 813                 err = 0;
 814                 goto destroy;
 815         default:
 816                 skerr = EPROTO;
 817                 err = -EINVAL;
 818                 goto destroy;
 819         }
 820         return 0;
 821
 822 destroy:
 823         virtio_transport_reset(vsk, pkt);
 824         sk->sk_state = TCP_CLOSE;
 825         sk->sk_err = skerr;
 826         sk->sk_error_report(sk);
 827         return err;
 828 }
 829
 830 static int
 831 virtio_transport_recv_connected(struct sock *sk,
 832                                 struct virtio_vsock_pkt *pkt)
 833 {
 834         struct vsock_sock *vsk = vsock_sk(sk);
 835         struct virtio_vsock_sock *vvs = vsk->trans;
 836         int err = 0;
 837
 838         switch (le16_to_cpu(pkt->hdr.op)) {
 839         case VIRTIO_VSOCK_OP_RW:
 840                 pkt->len = le32_to_cpu(pkt->hdr.len);
 841                 pkt->off = 0;
 842
 843                 spin_lock_bh(&vvs->rx_lock);
 844                 virtio_transport_inc_rx_pkt(vvs, pkt);
 845                 list_add_tail(&pkt->list, &vvs->rx_queue);
 846                 spin_unlock_bh(&vvs->rx_lock);
 847
 848                 sk->sk_data_ready(sk);
 849                 return err;
 850         case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
 851                 sk->sk_write_space(sk);
 852                 break;
 853         case VIRTIO_VSOCK_OP_SHUTDOWN:
 854                 if (le32_to_cpu(pkt->hdr.flags) & VIRTIO_VSOCK_SHUTDOWN_RCV)
 855                         vsk->peer_shutdown |= RCV_SHUTDOWN;
 856                 if (le32_to_cpu(pkt->hdr.flags) & VIRTIO_VSOCK_SHUTDOWN_SEND)
 857                         vsk->peer_shutdown |= SEND_SHUTDOWN;
 858                 if (vsk->peer_shutdown == SHUTDOWN_MASK &&
 859                     vsock_stream_has_data(vsk) <= 0)
 860                         sk->sk_state = TCP_CLOSING;
 861                 if (le32_to_cpu(pkt->hdr.flags))
 862                         sk->sk_state_change(sk);
 863                 break;
 864         case VIRTIO_VSOCK_OP_RST:
 865                 virtio_transport_do_close(vsk, true);
 866                 break;
 867         default:
 868                 err = -EINVAL;
 869                 break;
 870         }
 871
 872         virtio_transport_free_pkt(pkt);
 873         return err;
 874 }
 875
 876 static void
 877 virtio_transport_recv_disconnecting(struct sock *sk,
 878                                     struct virtio_vsock_pkt *pkt)
 879 {
 880         struct vsock_sock *vsk = vsock_sk(sk);
 881
 882         if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
 883                 virtio_transport_do_close(vsk, true);
 884 }
 885
 886 static int
 887 virtio_transport_send_response(struct vsock_sock *vsk,
 888                                struct virtio_vsock_pkt *pkt)
 889 {
 890         struct virtio_vsock_pkt_info info = {
 891                 .op = VIRTIO_VSOCK_OP_RESPONSE,
 892                 .type = VIRTIO_VSOCK_TYPE_STREAM,
 893                 .remote_cid = le64_to_cpu(pkt->hdr.src_cid),
 894                 .remote_port = le32_to_cpu(pkt->hdr.src_port),
 895                 .reply = true,
 896                 .vsk = vsk,
 897         };
 898
 899         return virtio_transport_send_pkt_info(vsk, &info);
 900 }
 901
 902 /* Handle server socket */
 903 static int
 904 virtio_transport_recv_listen(struct sock *sk, struct virtio_vsock_pkt *pkt)
 905 {
 906         struct vsock_sock *vsk = vsock_sk(sk);
 907         struct vsock_sock *vchild;
 908         struct sock *child;
 909
 910         if (le16_to_cpu(pkt->hdr.op) != VIRTIO_VSOCK_OP_REQUEST) {
 911                 virtio_transport_reset(vsk, pkt);
 912                 return -EINVAL;
 913         }
 914
 915         if (sk_acceptq_is_full(sk)) {
 916                 virtio_transport_reset(vsk, pkt);
 917                 return -ENOMEM;
 918         }
 919
 920         child = __vsock_create(sock_net(sk), NULL, sk, GFP_KERNEL,
 921                                sk->sk_type, 0);
 922         if (!child) {
 923                 virtio_transport_reset(vsk, pkt);
 924                 return -ENOMEM;
 925         }
 926
 927         sk->sk_ack_backlog++;
 928
 929         lock_sock_nested(child, SINGLE_DEPTH_NESTING);
 930
 931         child->sk_state = TCP_ESTABLISHED;
 932
 933         vchild = vsock_sk(child);
 934         vsock_addr_init(&vchild->local_addr, le64_to_cpu(pkt->hdr.dst_cid),
 935                         le32_to_cpu(pkt->hdr.dst_port));
 936         vsock_addr_init(&vchild->remote_addr, le64_to_cpu(pkt->hdr.src_cid),
 937                         le32_to_cpu(pkt->hdr.src_port));
 938
 939         vsock_insert_connected(vchild);
 940         vsock_enqueue_accept(sk, child);
 941         virtio_transport_send_response(vchild, pkt);
 942
 943         release_sock(child);
 944
 945         sk->sk_data_ready(sk);
 946         return 0;
 947 }
 948
 949 static bool virtio_transport_space_update(struct sock *sk,
 950                                           struct virtio_vsock_pkt *pkt)
 951 {
 952         struct vsock_sock *vsk = vsock_sk(sk);
 953         struct virtio_vsock_sock *vvs = vsk->trans;
 954         bool space_available;
 955
 956         /* buf_alloc and fwd_cnt is always included in the hdr */
 957         spin_lock_bh(&vvs->tx_lock);
 958         vvs->peer_buf_alloc = le32_to_cpu(pkt->hdr.buf_alloc);
 959         vvs->peer_fwd_cnt = le32_to_cpu(pkt->hdr.fwd_cnt);
 960         space_available = virtio_transport_has_space(vsk);
 961         spin_unlock_bh(&vvs->tx_lock);
 962         return space_available;
 963 }
 964
 965 /* We are under the virtio-vsock's vsock->rx_lock or vhost-vsock's vq->mutex
 966  * lock.
 967  */
 968 void virtio_transport_recv_pkt(struct virtio_vsock_pkt *pkt)
 969 {
 970         struct sockaddr_vm src, dst;
 971         struct vsock_sock *vsk;
 972         struct sock *sk;
 973         bool space_available;
 974
 975         vsock_addr_init(&src, le64_to_cpu(pkt->hdr.src_cid),
 976                         le32_to_cpu(pkt->hdr.src_port));
 977         vsock_addr_init(&dst, le64_to_cpu(pkt->hdr.dst_cid),
 978                         le32_to_cpu(pkt->hdr.dst_port));
 979
 980         trace_virtio_transport_recv_pkt(src.svm_cid, src.svm_port,
 981                                         dst.svm_cid, dst.svm_port,
 982                                         le32_to_cpu(pkt->hdr.len),
 983                                         le16_to_cpu(pkt->hdr.type),
 984                                         le16_to_cpu(pkt->hdr.op),
 985                                         le32_to_cpu(pkt->hdr.flags),
 986                                         le32_to_cpu(pkt->hdr.buf_alloc),
 987                                         le32_to_cpu(pkt->hdr.fwd_cnt));
 988
 989         if (le16_to_cpu(pkt->hdr.type) != VIRTIO_VSOCK_TYPE_STREAM) {
 990                 (void)virtio_transport_reset_no_sock(pkt);
 991                 goto free_pkt;
 992         }
 993
 994         /* The socket must be in connected or bound table
 995          * otherwise send reset back
 996          */
 997         sk = vsock_find_connected_socket(&src, &dst);
 998         if (!sk) {
 999                 sk = vsock_find_bound_socket(&dst);
1000                 if (!sk) {
1001                         (void)virtio_transport_reset_no_sock(pkt);
1002                         goto free_pkt;
1003                 }
1004         }
1005
1006         vsk = vsock_sk(sk);
1007
1008         space_available = virtio_transport_space_update(sk, pkt);
1009
1010         lock_sock(sk);
1011
1012         /* Update CID in case it has changed after a transport reset event */
1013         vsk->local_addr.svm_cid = dst.svm_cid;
1014
1015         if (space_available)
1016                 sk->sk_write_space(sk);
1017
1018         switch (sk->sk_state) {
1019         case TCP_LISTEN:
1020                 virtio_transport_recv_listen(sk, pkt);
1021                 virtio_transport_free_pkt(pkt);
1022                 break;
1023         case TCP_SYN_SENT:
1024                 virtio_transport_recv_connecting(sk, pkt);
1025                 virtio_transport_free_pkt(pkt);
1026                 break;
1027         case TCP_ESTABLISHED:
1028                 virtio_transport_recv_connected(sk, pkt);
1029                 break;
1030         case TCP_CLOSING:
1031                 virtio_transport_recv_disconnecting(sk, pkt);
1032                 virtio_transport_free_pkt(pkt);
1033                 break;
1034         default:
1035                 virtio_transport_free_pkt(pkt);
1036                 break;
1037         }
1038         release_sock(sk);
1039
1040         /* Release refcnt obtained when we fetched this socket out of the
1041          * bound or connected list.
1042          */
1043         sock_put(sk);
1044         return;
1045
1046 free_pkt:
1047         virtio_transport_free_pkt(pkt);
1048 }
1049 EXPORT_SYMBOL_GPL(virtio_transport_recv_pkt);
1050
1051 void virtio_transport_free_pkt(struct virtio_vsock_pkt *pkt)
1052 {
1053         kfree(pkt->buf);
1054         kfree(pkt);
1055 }
1056 EXPORT_SYMBOL_GPL(virtio_transport_free_pkt);
1057
1058 MODULE_LICENSE("GPL v2");
1059 MODULE_AUTHOR("Asias He");
1060 MODULE_DESCRIPTION("common code for virtio vsock");