drivers/vhost/net.c

   1 /* Copyright (C) 2009 Red Hat, Inc.
   2  * Author: Michael S. Tsirkin <mst@redhat.com>
   3  *
   4  * This work is licensed under the terms of the GNU GPL, version 2.
   5  *
   6  * virtio-net server in host kernel.
   7  */
   8
   9 #include <linux/compat.h>
  10 #include <linux/eventfd.h>
  11 #include <linux/vhost.h>
  12 #include <linux/virtio_net.h>
  13 #include <linux/miscdevice.h>
  14 #include <linux/module.h>
  15 #include <linux/moduleparam.h>
  16 #include <linux/mutex.h>
  17 #include <linux/workqueue.h>
  18 #include <linux/file.h>
  19 #include <linux/slab.h>
  20 #include <linux/vmalloc.h>
  21
  22 #include <linux/net.h>
  23 #include <linux/if_packet.h>
  24 #include <linux/if_arp.h>
  25 #include <linux/if_tun.h>
  26 #include <linux/if_macvlan.h>
  27 #include <linux/if_vlan.h>
  28
  29 #include <net/sock.h>
  30
  31 #include "vhost.h"
  32
  33 static int experimental_zcopytx = 1;
  34 module_param(experimental_zcopytx, int, 0444);
  35 MODULE_PARM_DESC(experimental_zcopytx, "Enable Zero Copy TX;"
  36                                        " 1 -Enable; 0 - Disable");
  37
  38 /* Max number of bytes transferred before requeueing the job.
  39  * Using this limit prevents one virtqueue from starving others. */
  40 #define VHOST_NET_WEIGHT 0x80000
  41
  42 /* MAX number of TX used buffers for outstanding zerocopy */
  43 #define VHOST_MAX_PEND 128
  44 #define VHOST_GOODCOPY_LEN 256
  45
  46 /*
  47  * For transmit, used buffer len is unused; we override it to track buffer
  48  * status internally; used for zerocopy tx only.
  49  */
  50 /* Lower device DMA failed */
  51 #define VHOST_DMA_FAILED_LEN    ((__force __virtio32)3)
  52 /* Lower device DMA done */
  53 #define VHOST_DMA_DONE_LEN      ((__force __virtio32)2)
  54 /* Lower device DMA in progress */
  55 #define VHOST_DMA_IN_PROGRESS   ((__force __virtio32)1)
  56 /* Buffer unused */
  57 #define VHOST_DMA_CLEAR_LEN     ((__force __virtio32)0)
  58
  59 #define VHOST_DMA_IS_DONE(len) ((__force u32)(len) >= (__force u32)VHOST_DMA_DONE_LEN)
  60
  61 enum {
  62         VHOST_NET_FEATURES = VHOST_FEATURES |
  63                          (1ULL << VHOST_NET_F_VIRTIO_NET_HDR) |
  64                          (1ULL << VIRTIO_NET_F_MRG_RXBUF)
  65 };
  66
  67 enum {
  68         VHOST_NET_VQ_RX = 0,
  69         VHOST_NET_VQ_TX = 1,
  70         VHOST_NET_VQ_MAX = 2,
  71 };
  72
  73 struct vhost_net_ubuf_ref {
  74         /* refcount follows semantics similar to kref:
  75          *  0: object is released
  76          *  1: no outstanding ubufs
  77          * >1: outstanding ubufs
  78          */
  79         atomic_t refcount;
  80         wait_queue_head_t wait;
  81         struct vhost_virtqueue *vq;
  82 };
  83
  84 struct vhost_net_virtqueue {
  85         struct vhost_virtqueue vq;
  86         size_t vhost_hlen;
  87         size_t sock_hlen;
  88         /* vhost zerocopy support fields below: */
  89         /* last used idx for outstanding DMA zerocopy buffers */
  90         int upend_idx;
  91         /* first used idx for DMA done zerocopy buffers */
  92         int done_idx;
  93         /* an array of userspace buffers info */
  94         struct ubuf_info *ubuf_info;
  95         /* Reference counting for outstanding ubufs.
  96          * Protected by vq mutex. Writers must also take device mutex. */
  97         struct vhost_net_ubuf_ref *ubufs;
  98 };
  99
 100 struct vhost_net {
 101         struct vhost_dev dev;
 102         struct vhost_net_virtqueue vqs[VHOST_NET_VQ_MAX];
 103         struct vhost_poll poll[VHOST_NET_VQ_MAX];
 104         /* Number of TX recently submitted.
 105          * Protected by tx vq lock. */
 106         unsigned tx_packets;
 107         /* Number of times zerocopy TX recently failed.
 108          * Protected by tx vq lock. */
 109         unsigned tx_zcopy_err;
 110         /* Flush in progress. Protected by tx vq lock. */
 111         bool tx_flush;
 112 };
 113
 114 static unsigned vhost_net_zcopy_mask __read_mostly;
 115
 116 static void vhost_net_enable_zcopy(int vq)
 117 {
 118         vhost_net_zcopy_mask |= 0x1 << vq;
 119 }
 120
 121 static struct vhost_net_ubuf_ref *
 122 vhost_net_ubuf_alloc(struct vhost_virtqueue *vq, bool zcopy)
 123 {
 124         struct vhost_net_ubuf_ref *ubufs;
 125         /* No zero copy backend? Nothing to count. */
 126         if (!zcopy)
 127                 return NULL;
 128         ubufs = kmalloc(sizeof(*ubufs), GFP_KERNEL);
 129         if (!ubufs)
 130                 return ERR_PTR(-ENOMEM);
 131         atomic_set(&ubufs->refcount, 1);
 132         init_waitqueue_head(&ubufs->wait);
 133         ubufs->vq = vq;
 134         return ubufs;
 135 }
 136
 137 static int vhost_net_ubuf_put(struct vhost_net_ubuf_ref *ubufs)
 138 {
 139         int r = atomic_sub_return(1, &ubufs->refcount);
 140         if (unlikely(!r))
 141                 wake_up(&ubufs->wait);
 142         return r;
 143 }
 144
 145 static void vhost_net_ubuf_put_and_wait(struct vhost_net_ubuf_ref *ubufs)
 146 {
 147         vhost_net_ubuf_put(ubufs);
 148         wait_event(ubufs->wait, !atomic_read(&ubufs->refcount));
 149 }
 150
 151 static void vhost_net_ubuf_put_wait_and_free(struct vhost_net_ubuf_ref *ubufs)
 152 {
 153         vhost_net_ubuf_put_and_wait(ubufs);
 154         kfree(ubufs);
 155 }
 156
 157 static void vhost_net_clear_ubuf_info(struct vhost_net *n)
 158 {
 159         int i;
 160
 161         for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
 162                 kfree(n->vqs[i].ubuf_info);
 163                 n->vqs[i].ubuf_info = NULL;
 164         }
 165 }
 166
 167 static int vhost_net_set_ubuf_info(struct vhost_net *n)
 168 {
 169         bool zcopy;
 170         int i;
 171
 172         for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
 173                 zcopy = vhost_net_zcopy_mask & (0x1 << i);
 174                 if (!zcopy)
 175                         continue;
 176                 n->vqs[i].ubuf_info = kmalloc(sizeof(*n->vqs[i].ubuf_info) *
 177                                               UIO_MAXIOV, GFP_KERNEL);
 178                 if  (!n->vqs[i].ubuf_info)
 179                         goto err;
 180         }
 181         return 0;
 182
 183 err:
 184         vhost_net_clear_ubuf_info(n);
 185         return -ENOMEM;
 186 }
 187
 188 static void vhost_net_vq_reset(struct vhost_net *n)
 189 {
 190         int i;
 191
 192         vhost_net_clear_ubuf_info(n);
 193
 194         for (i = 0; i < VHOST_NET_VQ_MAX; i++) {
 195                 n->vqs[i].done_idx = 0;
 196                 n->vqs[i].upend_idx = 0;
 197                 n->vqs[i].ubufs = NULL;
 198                 n->vqs[i].vhost_hlen = 0;
 199                 n->vqs[i].sock_hlen = 0;
 200         }
 201
 202 }
 203
 204 static void vhost_net_tx_packet(struct vhost_net *net)
 205 {
 206         ++net->tx_packets;
 207         if (net->tx_packets < 1024)
 208                 return;
 209         net->tx_packets = 0;
 210         net->tx_zcopy_err = 0;
 211 }
 212
 213 static void vhost_net_tx_err(struct vhost_net *net)
 214 {
 215         ++net->tx_zcopy_err;
 216 }
 217
 218 static bool vhost_net_tx_select_zcopy(struct vhost_net *net)
 219 {
 220         /* TX flush waits for outstanding DMAs to be done.
 221          * Don't start new DMAs.
 222          */
 223         return !net->tx_flush &&
 224                 net->tx_packets / 64 >= net->tx_zcopy_err;
 225 }
 226
 227 static bool vhost_sock_zcopy(struct socket *sock)
 228 {
 229         return unlikely(experimental_zcopytx) &&
 230                 sock_flag(sock->sk, SOCK_ZEROCOPY);
 231 }
 232
 233 /* In case of DMA done not in order in lower device driver for some reason.
 234  * upend_idx is used to track end of used idx, done_idx is used to track head
 235  * of used idx. Once lower device DMA done contiguously, we will signal KVM
 236  * guest used idx.
 237  */
 238 static void vhost_zerocopy_signal_used(struct vhost_net *net,
 239                                        struct vhost_virtqueue *vq)
 240 {
 241         struct vhost_net_virtqueue *nvq =
 242                 container_of(vq, struct vhost_net_virtqueue, vq);
 243         int i, add;
 244         int j = 0;
 245
 246         for (i = nvq->done_idx; i != nvq->upend_idx; i = (i + 1) % UIO_MAXIOV) {
 247                 if (vq->heads[i].len == VHOST_DMA_FAILED_LEN)
 248                         vhost_net_tx_err(net);
 249                 if (VHOST_DMA_IS_DONE(vq->heads[i].len)) {
 250                         vq->heads[i].len = VHOST_DMA_CLEAR_LEN;
 251                         ++j;
 252                 } else
 253                         break;
 254         }
 255         while (j) {
 256                 add = min(UIO_MAXIOV - nvq->done_idx, j);
 257                 vhost_add_used_and_signal_n(vq->dev, vq,
 258                                             &vq->heads[nvq->done_idx], add);
 259                 nvq->done_idx = (nvq->done_idx + add) % UIO_MAXIOV;
 260                 j -= add;
 261         }
 262 }
 263
 264 static void vhost_zerocopy_callback(struct ubuf_info *ubuf, bool success)
 265 {
 266         struct vhost_net_ubuf_ref *ubufs = ubuf->ctx;
 267         struct vhost_virtqueue *vq = ubufs->vq;
 268         int cnt;
 269
 270         rcu_read_lock_bh();
 271
 272         /* set len to mark this desc buffers done DMA */
 273         vq->heads[ubuf->desc].len = success ?
 274                 VHOST_DMA_DONE_LEN : VHOST_DMA_FAILED_LEN;
 275         cnt = vhost_net_ubuf_put(ubufs);
 276
 277         /*
 278          * Trigger polling thread if guest stopped submitting new buffers:
 279          * in this case, the refcount after decrement will eventually reach 1.
 280          * We also trigger polling periodically after each 16 packets
 281          * (the value 16 here is more or less arbitrary, it's tuned to trigger
 282          * less than 10% of times).
 283          */
 284         if (cnt <= 1 || !(cnt % 16))
 285                 vhost_poll_queue(&vq->poll);
 286
 287         rcu_read_unlock_bh();
 288 }
 289
 290 /* Expects to be always run from workqueue - which acts as
 291  * read-size critical section for our kind of RCU. */
 292 static void handle_tx(struct vhost_net *net)
 293 {
 294         struct vhost_net_virtqueue *nvq = &net->vqs[VHOST_NET_VQ_TX];
 295         struct vhost_virtqueue *vq = &nvq->vq;
 296         unsigned out, in;
 297         int head;
 298         struct msghdr msg = {
 299                 .msg_name = NULL,
 300                 .msg_namelen = 0,
 301                 .msg_control = NULL,
 302                 .msg_controllen = 0,
 303                 .msg_flags = MSG_DONTWAIT,
 304         };
 305         size_t len, total_len = 0;
 306         int err;
 307         size_t hdr_size;
 308         struct socket *sock;
 309         struct vhost_net_ubuf_ref *uninitialized_var(ubufs);
 310         bool zcopy, zcopy_used;
 311
 312         mutex_lock(&vq->mutex);
 313         sock = vq->private_data;
 314         if (!sock)
 315                 goto out;
 316
 317         vhost_disable_notify(&net->dev, vq);
 318
 319         hdr_size = nvq->vhost_hlen;
 320         zcopy = nvq->ubufs;
 321
 322         for (;;) {
 323                 /* Release DMAs done buffers first */
 324                 if (zcopy)
 325                         vhost_zerocopy_signal_used(net, vq);
 326
 327                 /* If more outstanding DMAs, queue the work.
 328                  * Handle upend_idx wrap around
 329                  */
 330                 if (unlikely((nvq->upend_idx + vq->num - VHOST_MAX_PEND)
 331                               % UIO_MAXIOV == nvq->done_idx))
 332                         break;
 333
 334                 head = vhost_get_vq_desc(vq, vq->iov,
 335                                          ARRAY_SIZE(vq->iov),
 336                                          &out, &in,
 337                                          NULL, NULL);
 338                 /* On error, stop handling until the next kick. */
 339                 if (unlikely(head < 0))
 340                         break;
 341                 /* Nothing new?  Wait for eventfd to tell us they refilled. */
 342                 if (head == vq->num) {
 343                         if (unlikely(vhost_enable_notify(&net->dev, vq))) {
 344                                 vhost_disable_notify(&net->dev, vq);
 345                                 continue;
 346                         }
 347                         break;
 348                 }
 349                 if (in) {
 350                         vq_err(vq, "Unexpected descriptor format for TX: "
 351                                "out %d, int %d\n", out, in);
 352                         break;
 353                 }
 354                 /* Skip header. TODO: support TSO. */
 355                 len = iov_length(vq->iov, out);
 356                 iov_iter_init(&msg.msg_iter, WRITE, vq->iov, out, len);
 357                 iov_iter_advance(&msg.msg_iter, hdr_size);
 358                 /* Sanity check */
 359                 if (!msg_data_left(&msg)) {
 360                         vq_err(vq, "Unexpected header len for TX: "
 361                                "%zd expected %zd\n",
 362                                len, hdr_size);
 363                         break;
 364                 }
 365                 len = msg_data_left(&msg);
 366
 367                 zcopy_used = zcopy && len >= VHOST_GOODCOPY_LEN
 368                                    && (nvq->upend_idx + 1) % UIO_MAXIOV !=
 369                                       nvq->done_idx
 370                                    && vhost_net_tx_select_zcopy(net);
 371
 372                 /* use msg_control to pass vhost zerocopy ubuf info to skb */
 373                 if (zcopy_used) {
 374                         struct ubuf_info *ubuf;
 375                         ubuf = nvq->ubuf_info + nvq->upend_idx;
 376
 377                         vq->heads[nvq->upend_idx].id = cpu_to_vhost32(vq, head);
 378                         vq->heads[nvq->upend_idx].len = VHOST_DMA_IN_PROGRESS;
 379                         ubuf->callback = vhost_zerocopy_callback;
 380                         ubuf->ctx = nvq->ubufs;
 381                         ubuf->desc = nvq->upend_idx;
 382                         msg.msg_control = ubuf;
 383                         msg.msg_controllen = sizeof(ubuf);
 384                         ubufs = nvq->ubufs;
 385                         atomic_inc(&ubufs->refcount);
 386                         nvq->upend_idx = (nvq->upend_idx + 1) % UIO_MAXIOV;
 387                 } else {
 388                         msg.msg_control = NULL;
 389                         ubufs = NULL;
 390                 }
 391                 /* TODO: Check specific error and bomb out unless ENOBUFS? */
 392                 err = sock->ops->sendmsg(sock, &msg, len);
 393                 if (unlikely(err < 0)) {
 394                         if (zcopy_used) {
 395                                 vhost_net_ubuf_put(ubufs);
 396                                 nvq->upend_idx = ((unsigned)nvq->upend_idx - 1)
 397                                         % UIO_MAXIOV;
 398                         }
 399                         vhost_discard_vq_desc(vq, 1);
 400                         break;
 401                 }
 402                 if (err != len)
 403                         pr_debug("Truncated TX packet: "
 404                                  " len %d != %zd\n", err, len);
 405                 if (!zcopy_used)
 406                         vhost_add_used_and_signal(&net->dev, vq, head, 0);
 407                 else
 408                         vhost_zerocopy_signal_used(net, vq);
 409                 total_len += len;
 410                 vhost_net_tx_packet(net);
 411                 if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
 412                         vhost_poll_queue(&vq->poll);
 413                         break;
 414                 }
 415         }
 416 out:
 417         mutex_unlock(&vq->mutex);
 418 }
 419
 420 static int peek_head_len(struct sock *sk)
 421 {
 422         struct sk_buff *head;
 423         int len = 0;
 424         unsigned long flags;
 425
 426         spin_lock_irqsave(&sk->sk_receive_queue.lock, flags);
 427         head = skb_peek(&sk->sk_receive_queue);
 428         if (likely(head)) {
 429                 len = head->len;
 430                 if (skb_vlan_tag_present(head))
 431                         len += VLAN_HLEN;
 432         }
 433
 434         spin_unlock_irqrestore(&sk->sk_receive_queue.lock, flags);
 435         return len;
 436 }
 437
 438 /* This is a multi-buffer version of vhost_get_desc, that works if
 439  *      vq has read descriptors only.
 440  * @vq          - the relevant virtqueue
 441  * @datalen     - data length we'll be reading
 442  * @iovcount    - returned count of io vectors we fill
 443  * @log         - vhost log
 444  * @log_num     - log offset
 445  * @quota       - headcount quota, 1 for big buffer
 446  *      returns number of buffer heads allocated, negative on error
 447  */
 448 static int get_rx_bufs(struct vhost_virtqueue *vq,
 449                        struct vring_used_elem *heads,
 450                        int datalen,
 451                        unsigned *iovcount,
 452                        struct vhost_log *log,
 453                        unsigned *log_num,
 454                        unsigned int quota)
 455 {
 456         unsigned int out, in;
 457         int seg = 0;
 458         int headcount = 0;
 459         unsigned d;
 460         int r, nlogs = 0;
 461         /* len is always initialized before use since we are always called with
 462          * datalen > 0.
 463          */
 464         u32 uninitialized_var(len);
 465
 466         while (datalen > 0 && headcount < quota) {
 467                 if (unlikely(seg >= UIO_MAXIOV)) {
 468                         r = -ENOBUFS;
 469                         goto err;
 470                 }
 471                 r = vhost_get_vq_desc(vq, vq->iov + seg,
 472                                       ARRAY_SIZE(vq->iov) - seg, &out,
 473                                       &in, log, log_num);
 474                 if (unlikely(r < 0))
 475                         goto err;
 476
 477                 d = r;
 478                 if (d == vq->num) {
 479                         r = 0;
 480                         goto err;
 481                 }
 482                 if (unlikely(out || in <= 0)) {
 483                         vq_err(vq, "unexpected descriptor format for RX: "
 484                                 "out %d, in %d\n", out, in);
 485                         r = -EINVAL;
 486                         goto err;
 487                 }
 488                 if (unlikely(log)) {
 489                         nlogs += *log_num;
 490                         log += *log_num;
 491                 }
 492                 heads[headcount].id = cpu_to_vhost32(vq, d);
 493                 len = iov_length(vq->iov + seg, in);
 494                 heads[headcount].len = cpu_to_vhost32(vq, len);
 495                 datalen -= len;
 496                 ++headcount;
 497                 seg += in;
 498         }
 499         heads[headcount - 1].len = cpu_to_vhost32(vq, len + datalen);
 500         *iovcount = seg;
 501         if (unlikely(log))
 502                 *log_num = nlogs;
 503
 504         /* Detect overrun */
 505         if (unlikely(datalen > 0)) {
 506                 r = UIO_MAXIOV + 1;
 507                 goto err;
 508         }
 509         return headcount;
 510 err:
 511         vhost_discard_vq_desc(vq, headcount);
 512         return r;
 513 }
 514
 515 /* Expects to be always run from workqueue - which acts as
 516  * read-size critical section for our kind of RCU. */
 517 static void handle_rx(struct vhost_net *net)
 518 {
 519         struct vhost_net_virtqueue *nvq = &net->vqs[VHOST_NET_VQ_RX];
 520         struct vhost_virtqueue *vq = &nvq->vq;
 521         unsigned uninitialized_var(in), log;
 522         struct vhost_log *vq_log;
 523         struct msghdr msg = {
 524                 .msg_name = NULL,
 525                 .msg_namelen = 0,
 526                 .msg_control = NULL, /* FIXME: get and handle RX aux data. */
 527                 .msg_controllen = 0,
 528                 .msg_flags = MSG_DONTWAIT,
 529         };
 530         struct virtio_net_hdr hdr = {
 531                 .flags = 0,
 532                 .gso_type = VIRTIO_NET_HDR_GSO_NONE
 533         };
 534         size_t total_len = 0;
 535         int err, mergeable;
 536         s16 headcount;
 537         size_t vhost_hlen, sock_hlen;
 538         size_t vhost_len, sock_len;
 539         struct socket *sock;
 540         struct iov_iter fixup;
 541         __virtio16 num_buffers;
 542
 543         mutex_lock(&vq->mutex);
 544         sock = vq->private_data;
 545         if (!sock)
 546                 goto out;
 547         vhost_disable_notify(&net->dev, vq);
 548
 549         vhost_hlen = nvq->vhost_hlen;
 550         sock_hlen = nvq->sock_hlen;
 551
 552         vq_log = unlikely(vhost_has_feature(vq, VHOST_F_LOG_ALL)) ?
 553                 vq->log : NULL;
 554         mergeable = vhost_has_feature(vq, VIRTIO_NET_F_MRG_RXBUF);
 555
 556         while ((sock_len = peek_head_len(sock->sk))) {
 557                 sock_len += sock_hlen;
 558                 vhost_len = sock_len + vhost_hlen;
 559                 headcount = get_rx_bufs(vq, vq->heads, vhost_len,
 560                                         &in, vq_log, &log,
 561                                         likely(mergeable) ? UIO_MAXIOV : 1);
 562                 /* On error, stop handling until the next kick. */
 563                 if (unlikely(headcount < 0))
 564                         break;
 565                 /* On overrun, truncate and discard */
 566                 if (unlikely(headcount > UIO_MAXIOV)) {
 567                         iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1);
 568                         err = sock->ops->recvmsg(sock, &msg,
 569                                                  1, MSG_DONTWAIT | MSG_TRUNC);
 570                         pr_debug("Discarded rx packet: len %zd\n", sock_len);
 571                         continue;
 572                 }
 573                 /* OK, now we need to know about added descriptors. */
 574                 if (!headcount) {
 575                         if (unlikely(vhost_enable_notify(&net->dev, vq))) {
 576                                 /* They have slipped one in as we were
 577                                  * doing that: check again. */
 578                                 vhost_disable_notify(&net->dev, vq);
 579                                 continue;
 580                         }
 581                         /* Nothing new?  Wait for eventfd to tell us
 582                          * they refilled. */
 583                         break;
 584                 }
 585                 /* We don't need to be notified again. */
 586                 iov_iter_init(&msg.msg_iter, READ, vq->iov, in, vhost_len);
 587                 fixup = msg.msg_iter;
 588                 if (unlikely((vhost_hlen))) {
 589                         /* We will supply the header ourselves
 590                          * TODO: support TSO.
 591                          */
 592                         iov_iter_advance(&msg.msg_iter, vhost_hlen);
 593                 }
 594                 err = sock->ops->recvmsg(sock, &msg,
 595                                          sock_len, MSG_DONTWAIT | MSG_TRUNC);
 596                 /* Userspace might have consumed the packet meanwhile:
 597                  * it's not supposed to do this usually, but might be hard
 598                  * to prevent. Discard data we got (if any) and keep going. */
 599                 if (unlikely(err != sock_len)) {
 600                         pr_debug("Discarded rx packet: "
 601                                  " len %d, expected %zd\n", err, sock_len);
 602                         vhost_discard_vq_desc(vq, headcount);
 603                         continue;
 604                 }
 605                 /* Supply virtio_net_hdr if VHOST_NET_F_VIRTIO_NET_HDR */
 606                 if (unlikely(vhost_hlen)) {
 607                         if (copy_to_iter(&hdr, sizeof(hdr),
 608                                          &fixup) != sizeof(hdr)) {
 609                                 vq_err(vq, "Unable to write vnet_hdr "
 610                                        "at addr %p\n", vq->iov->iov_base);
 611                                 break;
 612                         }
 613                 } else {
 614                         /* Header came from socket; we'll need to patch
 615                          * ->num_buffers over if VIRTIO_NET_F_MRG_RXBUF
 616                          */
 617                         iov_iter_advance(&fixup, sizeof(hdr));
 618                 }
 619                 /* TODO: Should check and handle checksum. */
 620
 621                 num_buffers = cpu_to_vhost16(vq, headcount);
 622                 if (likely(mergeable) &&
 623                     copy_to_iter(&num_buffers, sizeof num_buffers,
 624                                  &fixup) != sizeof num_buffers) {
 625                         vq_err(vq, "Failed num_buffers write");
 626                         vhost_discard_vq_desc(vq, headcount);
 627                         break;
 628                 }
 629                 vhost_add_used_and_signal_n(&net->dev, vq, vq->heads,
 630                                             headcount);
 631                 if (unlikely(vq_log))
 632                         vhost_log_write(vq, vq_log, log, vhost_len);
 633                 total_len += vhost_len;
 634                 if (unlikely(total_len >= VHOST_NET_WEIGHT)) {
 635                         vhost_poll_queue(&vq->poll);
 636                         break;
 637                 }
 638         }
 639 out:
 640         mutex_unlock(&vq->mutex);
 641 }
 642
 643 static void handle_tx_kick(struct vhost_work *work)
 644 {
 645         struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
 646                                                   poll.work);
 647         struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);
 648
 649         handle_tx(net);
 650 }
 651
 652 static void handle_rx_kick(struct vhost_work *work)
 653 {
 654         struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
 655                                                   poll.work);
 656         struct vhost_net *net = container_of(vq->dev, struct vhost_net, dev);
 657
 658         handle_rx(net);
 659 }
 660
 661 static void handle_tx_net(struct vhost_work *work)
 662 {
 663         struct vhost_net *net = container_of(work, struct vhost_net,
 664                                              poll[VHOST_NET_VQ_TX].work);
 665         handle_tx(net);
 666 }
 667
 668 static void handle_rx_net(struct vhost_work *work)
 669 {
 670         struct vhost_net *net = container_of(work, struct vhost_net,
 671                                              poll[VHOST_NET_VQ_RX].work);
 672         handle_rx(net);
 673 }
 674
 675 static int vhost_net_open(struct inode *inode, struct file *f)
 676 {
 677         struct vhost_net *n;
 678         struct vhost_dev *dev;
 679         struct vhost_virtqueue **vqs;
 680         int i;
 681
 682         n = kmalloc(sizeof *n, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
 683         if (!n) {
 684                 n = vmalloc(sizeof *n);
 685                 if (!n)
 686                         return -ENOMEM;
 687         }
 688         vqs = kmalloc(VHOST_NET_VQ_MAX * sizeof(*vqs), GFP_KERNEL);
 689         if (!vqs) {
 690                 kvfree(n);
 691                 return -ENOMEM;
 692         }
 693
 694         dev = &n->dev;
 695         vqs[VHOST_NET_VQ_TX] = &n->vqs[VHOST_NET_VQ_TX].vq;
 696         vqs[VHOST_NET_VQ_RX] = &n->vqs[VHOST_NET_VQ_RX].vq;
 697         n->vqs[VHOST_NET_VQ_TX].vq.handle_kick = handle_tx_kick;
 698         n->vqs[VHOST_NET_VQ_RX].vq.handle_kick = handle_rx_kick;
 699         for (i = 0; i < VHOST_NET_VQ_MAX; i++) {
 700                 n->vqs[i].ubufs = NULL;
 701                 n->vqs[i].ubuf_info = NULL;
 702                 n->vqs[i].upend_idx = 0;
 703                 n->vqs[i].done_idx = 0;
 704                 n->vqs[i].vhost_hlen = 0;
 705                 n->vqs[i].sock_hlen = 0;
 706         }
 707         vhost_dev_init(dev, vqs, VHOST_NET_VQ_MAX);
 708
 709         vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT, dev);
 710         vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN, dev);
 711
 712         f->private_data = n;
 713
 714         return 0;
 715 }
 716
 717 static void vhost_net_disable_vq(struct vhost_net *n,
 718                                  struct vhost_virtqueue *vq)
 719 {
 720         struct vhost_net_virtqueue *nvq =
 721                 container_of(vq, struct vhost_net_virtqueue, vq);
 722         struct vhost_poll *poll = n->poll + (nvq - n->vqs);
 723         if (!vq->private_data)
 724                 return;
 725         vhost_poll_stop(poll);
 726 }
 727
 728 static int vhost_net_enable_vq(struct vhost_net *n,
 729                                 struct vhost_virtqueue *vq)
 730 {
 731         struct vhost_net_virtqueue *nvq =
 732                 container_of(vq, struct vhost_net_virtqueue, vq);
 733         struct vhost_poll *poll = n->poll + (nvq - n->vqs);
 734         struct socket *sock;
 735
 736         sock = vq->private_data;
 737         if (!sock)
 738                 return 0;
 739
 740         return vhost_poll_start(poll, sock->file);
 741 }
 742
 743 static struct socket *vhost_net_stop_vq(struct vhost_net *n,
 744                                         struct vhost_virtqueue *vq)
 745 {
 746         struct socket *sock;
 747
 748         mutex_lock(&vq->mutex);
 749         sock = vq->private_data;
 750         vhost_net_disable_vq(n, vq);
 751         vq->private_data = NULL;
 752         mutex_unlock(&vq->mutex);
 753         return sock;
 754 }
 755
 756 static void vhost_net_stop(struct vhost_net *n, struct socket **tx_sock,
 757                            struct socket **rx_sock)
 758 {
 759         *tx_sock = vhost_net_stop_vq(n, &n->vqs[VHOST_NET_VQ_TX].vq);
 760         *rx_sock = vhost_net_stop_vq(n, &n->vqs[VHOST_NET_VQ_RX].vq);
 761 }
 762
 763 static void vhost_net_flush_vq(struct vhost_net *n, int index)
 764 {
 765         vhost_poll_flush(n->poll + index);
 766         vhost_poll_flush(&n->vqs[index].vq.poll);
 767 }
 768
 769 static void vhost_net_flush(struct vhost_net *n)
 770 {
 771         vhost_net_flush_vq(n, VHOST_NET_VQ_TX);
 772         vhost_net_flush_vq(n, VHOST_NET_VQ_RX);
 773         if (n->vqs[VHOST_NET_VQ_TX].ubufs) {
 774                 mutex_lock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
 775                 n->tx_flush = true;
 776                 mutex_unlock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
 777                 /* Wait for all lower device DMAs done. */
 778                 vhost_net_ubuf_put_and_wait(n->vqs[VHOST_NET_VQ_TX].ubufs);
 779                 mutex_lock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
 780                 n->tx_flush = false;
 781                 atomic_set(&n->vqs[VHOST_NET_VQ_TX].ubufs->refcount, 1);
 782                 mutex_unlock(&n->vqs[VHOST_NET_VQ_TX].vq.mutex);
 783         }
 784 }
 785
 786 static int vhost_net_release(struct inode *inode, struct file *f)
 787 {
 788         struct vhost_net *n = f->private_data;
 789         struct socket *tx_sock;
 790         struct socket *rx_sock;
 791
 792         vhost_net_stop(n, &tx_sock, &rx_sock);
 793         vhost_net_flush(n);
 794         vhost_dev_stop(&n->dev);
 795         vhost_dev_cleanup(&n->dev, false);
 796         vhost_net_vq_reset(n);
 797         if (tx_sock)
 798                 sockfd_put(tx_sock);
 799         if (rx_sock)
 800                 sockfd_put(rx_sock);
 801         /* Make sure no callbacks are outstanding */
 802         synchronize_rcu_bh();
 803         /* We do an extra flush before freeing memory,
 804          * since jobs can re-queue themselves. */
 805         vhost_net_flush(n);
 806         kfree(n->dev.vqs);
 807         kvfree(n);
 808         return 0;
 809 }
 810
 811 static struct socket *get_raw_socket(int fd)
 812 {
 813         struct {
 814                 struct sockaddr_ll sa;
 815                 char  buf[MAX_ADDR_LEN];
 816         } uaddr;
 817         int uaddr_len = sizeof uaddr, r;
 818         struct socket *sock = sockfd_lookup(fd, &r);
 819
 820         if (!sock)
 821                 return ERR_PTR(-ENOTSOCK);
 822
 823         /* Parameter checking */
 824         if (sock->sk->sk_type != SOCK_RAW) {
 825                 r = -ESOCKTNOSUPPORT;
 826                 goto err;
 827         }
 828
 829         r = sock->ops->getname(sock, (struct sockaddr *)&uaddr.sa,
 830                                &uaddr_len, 0);
 831         if (r)
 832                 goto err;
 833
 834         if (uaddr.sa.sll_family != AF_PACKET) {
 835                 r = -EPFNOSUPPORT;
 836                 goto err;
 837         }
 838         return sock;
 839 err:
 840         sockfd_put(sock);
 841         return ERR_PTR(r);
 842 }
 843
 844 static struct socket *get_tap_socket(int fd)
 845 {
 846         struct file *file = fget(fd);
 847         struct socket *sock;
 848
 849         if (!file)
 850                 return ERR_PTR(-EBADF);
 851         sock = tun_get_socket(file);
 852         if (!IS_ERR(sock))
 853                 return sock;
 854         sock = macvtap_get_socket(file);
 855         if (IS_ERR(sock))
 856                 fput(file);
 857         return sock;
 858 }
 859
 860 static struct socket *get_socket(int fd)
 861 {
 862         struct socket *sock;
 863
 864         /* special case to disable backend */
 865         if (fd == -1)
 866                 return NULL;
 867         sock = get_raw_socket(fd);
 868         if (!IS_ERR(sock))
 869                 return sock;
 870         sock = get_tap_socket(fd);
 871         if (!IS_ERR(sock))
 872                 return sock;
 873         return ERR_PTR(-ENOTSOCK);
 874 }
 875
 876 static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
 877 {
 878         struct socket *sock, *oldsock;
 879         struct vhost_virtqueue *vq;
 880         struct vhost_net_virtqueue *nvq;
 881         struct vhost_net_ubuf_ref *ubufs, *oldubufs = NULL;
 882         int r;
 883
 884         mutex_lock(&n->dev.mutex);
 885         r = vhost_dev_check_owner(&n->dev);
 886         if (r)
 887                 goto err;
 888
 889         if (index >= VHOST_NET_VQ_MAX) {
 890                 r = -ENOBUFS;
 891                 goto err;
 892         }
 893         vq = &n->vqs[index].vq;
 894         nvq = &n->vqs[index];
 895         mutex_lock(&vq->mutex);
 896
 897         /* Verify that ring has been setup correctly. */
 898         if (!vhost_vq_access_ok(vq)) {
 899                 r = -EFAULT;
 900                 goto err_vq;
 901         }
 902         sock = get_socket(fd);
 903         if (IS_ERR(sock)) {
 904                 r = PTR_ERR(sock);
 905                 goto err_vq;
 906         }
 907
 908         /* start polling new socket */
 909         oldsock = vq->private_data;
 910         if (sock != oldsock) {
 911                 ubufs = vhost_net_ubuf_alloc(vq,
 912                                              sock && vhost_sock_zcopy(sock));
 913                 if (IS_ERR(ubufs)) {
 914                         r = PTR_ERR(ubufs);
 915                         goto err_ubufs;
 916                 }
 917
 918                 vhost_net_disable_vq(n, vq);
 919                 vq->private_data = sock;
 920                 r = vhost_init_used(vq);
 921                 if (r)
 922                         goto err_used;
 923                 r = vhost_net_enable_vq(n, vq);
 924                 if (r)
 925                         goto err_used;
 926
 927                 oldubufs = nvq->ubufs;
 928                 nvq->ubufs = ubufs;
 929
 930                 n->tx_packets = 0;
 931                 n->tx_zcopy_err = 0;
 932                 n->tx_flush = false;
 933         }
 934
 935         mutex_unlock(&vq->mutex);
 936
 937         if (oldubufs) {
 938                 vhost_net_ubuf_put_wait_and_free(oldubufs);
 939                 mutex_lock(&vq->mutex);
 940                 vhost_zerocopy_signal_used(n, vq);
 941                 mutex_unlock(&vq->mutex);
 942         }
 943
 944         if (oldsock) {
 945                 vhost_net_flush_vq(n, index);
 946                 sockfd_put(oldsock);
 947         }
 948
 949         mutex_unlock(&n->dev.mutex);
 950         return 0;
 951
 952 err_used:
 953         vq->private_data = oldsock;
 954         vhost_net_enable_vq(n, vq);
 955         if (ubufs)
 956                 vhost_net_ubuf_put_wait_and_free(ubufs);
 957 err_ubufs:
 958         sockfd_put(sock);
 959 err_vq:
 960         mutex_unlock(&vq->mutex);
 961 err:
 962         mutex_unlock(&n->dev.mutex);
 963         return r;
 964 }
 965
 966 static long vhost_net_reset_owner(struct vhost_net *n)
 967 {
 968         struct socket *tx_sock = NULL;
 969         struct socket *rx_sock = NULL;
 970         long err;
 971         struct vhost_memory *memory;
 972
 973         mutex_lock(&n->dev.mutex);
 974         err = vhost_dev_check_owner(&n->dev);
 975         if (err)
 976                 goto done;
 977         memory = vhost_dev_reset_owner_prepare();
 978         if (!memory) {
 979                 err = -ENOMEM;
 980                 goto done;
 981         }
 982         vhost_net_stop(n, &tx_sock, &rx_sock);
 983         vhost_net_flush(n);
 984         vhost_dev_reset_owner(&n->dev, memory);
 985         vhost_net_vq_reset(n);
 986 done:
 987         mutex_unlock(&n->dev.mutex);
 988         if (tx_sock)
 989                 sockfd_put(tx_sock);
 990         if (rx_sock)
 991                 sockfd_put(rx_sock);
 992         return err;
 993 }
 994
 995 static int vhost_net_set_features(struct vhost_net *n, u64 features)
 996 {
 997         size_t vhost_hlen, sock_hlen, hdr_len;
 998         int i;
 999
1000         hdr_len = (features & ((1ULL << VIRTIO_NET_F_MRG_RXBUF) |
1001                                (1ULL << VIRTIO_F_VERSION_1))) ?
1002                         sizeof(struct virtio_net_hdr_mrg_rxbuf) :
1003                         sizeof(struct virtio_net_hdr);
1004         if (features & (1 << VHOST_NET_F_VIRTIO_NET_HDR)) {
1005                 /* vhost provides vnet_hdr */
1006                 vhost_hlen = hdr_len;
1007                 sock_hlen = 0;
1008         } else {
1009                 /* socket provides vnet_hdr */
1010                 vhost_hlen = 0;
1011                 sock_hlen = hdr_len;
1012         }
1013         mutex_lock(&n->dev.mutex);
1014         if ((features & (1 << VHOST_F_LOG_ALL)) &&
1015             !vhost_log_access_ok(&n->dev)) {
1016                 mutex_unlock(&n->dev.mutex);
1017                 return -EFAULT;
1018         }
1019         for (i = 0; i < VHOST_NET_VQ_MAX; ++i) {
1020                 mutex_lock(&n->vqs[i].vq.mutex);
1021                 n->vqs[i].vq.acked_features = features;
1022                 n->vqs[i].vhost_hlen = vhost_hlen;
1023                 n->vqs[i].sock_hlen = sock_hlen;
1024                 mutex_unlock(&n->vqs[i].vq.mutex);
1025         }
1026         mutex_unlock(&n->dev.mutex);
1027         return 0;
1028 }
1029
1030 static long vhost_net_set_owner(struct vhost_net *n)
1031 {
1032         int r;
1033
1034         mutex_lock(&n->dev.mutex);
1035         if (vhost_dev_has_owner(&n->dev)) {
1036                 r = -EBUSY;
1037                 goto out;
1038         }
1039         r = vhost_net_set_ubuf_info(n);
1040         if (r)
1041                 goto out;
1042         r = vhost_dev_set_owner(&n->dev);
1043         if (r)
1044                 vhost_net_clear_ubuf_info(n);
1045         vhost_net_flush(n);
1046 out:
1047         mutex_unlock(&n->dev.mutex);
1048         return r;
1049 }
1050
1051 static long vhost_net_ioctl(struct file *f, unsigned int ioctl,
1052                             unsigned long arg)
1053 {
1054         struct vhost_net *n = f->private_data;
1055         void __user *argp = (void __user *)arg;
1056         u64 __user *featurep = argp;
1057         struct vhost_vring_file backend;
1058         u64 features;
1059         int r;
1060
1061         switch (ioctl) {
1062         case VHOST_NET_SET_BACKEND:
1063                 if (copy_from_user(&backend, argp, sizeof backend))
1064                         return -EFAULT;
1065                 return vhost_net_set_backend(n, backend.index, backend.fd);
1066         case VHOST_GET_FEATURES:
1067                 features = VHOST_NET_FEATURES;
1068                 if (copy_to_user(featurep, &features, sizeof features))
1069                         return -EFAULT;
1070                 return 0;
1071         case VHOST_SET_FEATURES:
1072                 if (copy_from_user(&features, featurep, sizeof features))
1073                         return -EFAULT;
1074                 if (features & ~VHOST_NET_FEATURES)
1075                         return -EOPNOTSUPP;
1076                 return vhost_net_set_features(n, features);
1077         case VHOST_RESET_OWNER:
1078                 return vhost_net_reset_owner(n);
1079         case VHOST_SET_OWNER:
1080                 return vhost_net_set_owner(n);
1081         default:
1082                 mutex_lock(&n->dev.mutex);
1083                 r = vhost_dev_ioctl(&n->dev, ioctl, argp);
1084                 if (r == -ENOIOCTLCMD)
1085                         r = vhost_vring_ioctl(&n->dev, ioctl, argp);
1086                 else
1087                         vhost_net_flush(n);
1088                 mutex_unlock(&n->dev.mutex);
1089                 return r;
1090         }
1091 }
1092
1093 #ifdef CONFIG_COMPAT
1094 static long vhost_net_compat_ioctl(struct file *f, unsigned int ioctl,
1095                                    unsigned long arg)
1096 {
1097         return vhost_net_ioctl(f, ioctl, (unsigned long)compat_ptr(arg));
1098 }
1099 #endif
1100
1101 static const struct file_operations vhost_net_fops = {
1102         .owner          = THIS_MODULE,
1103         .release        = vhost_net_release,
1104         .unlocked_ioctl = vhost_net_ioctl,
1105 #ifdef CONFIG_COMPAT
1106         .compat_ioctl   = vhost_net_compat_ioctl,
1107 #endif
1108         .open           = vhost_net_open,
1109         .llseek         = noop_llseek,
1110 };
1111
1112 static struct miscdevice vhost_net_misc = {
1113         .minor = VHOST_NET_MINOR,
1114         .name = "vhost-net",
1115         .fops = &vhost_net_fops,
1116 };
1117
1118 static int vhost_net_init(void)
1119 {
1120         if (experimental_zcopytx)
1121                 vhost_net_enable_zcopy(VHOST_NET_VQ_TX);
1122         return misc_register(&vhost_net_misc);
1123 }
1124 module_init(vhost_net_init);
1125
1126 static void vhost_net_exit(void)
1127 {
1128         misc_deregister(&vhost_net_misc);
1129 }
1130 module_exit(vhost_net_exit);
1131
1132 MODULE_VERSION("0.0.1");
1133 MODULE_LICENSE("GPL v2");
1134 MODULE_AUTHOR("Michael S. Tsirkin");
1135 MODULE_DESCRIPTION("Host kernel accelerator for virtio net");
1136 MODULE_ALIAS_MISCDEV(VHOST_NET_MINOR);
1137 MODULE_ALIAS("devname:vhost-net");