ceph/src/spdk/lib/vhost/rte_vhost/vhost.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
   5  *   All rights reserved.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  11  *     * Redistributions of source code must retain the above copyright
  12  *       notice, this list of conditions and the following disclaimer.
  13  *     * Redistributions in binary form must reproduce the above copyright
  14  *       notice, this list of conditions and the following disclaimer in
  15  *       the documentation and/or other materials provided with the
  16  *       distribution.
  17  *     * Neither the name of Intel Corporation nor the names of its
  18  *       contributors may be used to endorse or promote products derived
  19  *       from this software without specific prior written permission.
  20  *
  21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 #include <linux/vhost.h>
  35 #include <linux/virtio_net.h>
  36 #include <stddef.h>
  37 #include <stdint.h>
  38 #include <stdlib.h>
  39 #ifdef RTE_LIBRTE_VHOST_NUMA
  40 #include <numaif.h>
  41 #endif
  42
  43 #include <rte_ethdev.h>
  44 #include <rte_log.h>
  45 #include <rte_string_fns.h>
  46 #include <rte_memory.h>
  47 #include <rte_malloc.h>
  48 #include <rte_virtio_net.h>
  49
  50 #include "vhost.h"
  51
  52 #define VHOST_USER_F_PROTOCOL_FEATURES  30
  53
  54 /* Features supported by this lib. */
  55 #define VHOST_SUPPORTED_FEATURES ((1ULL << VIRTIO_NET_F_MRG_RXBUF) | \
  56                                 (1ULL << VIRTIO_NET_F_CTRL_VQ) | \
  57                                 (1ULL << VIRTIO_NET_F_CTRL_RX) | \
  58                                 (1ULL << VIRTIO_NET_F_GUEST_ANNOUNCE) | \
  59                                 (VHOST_SUPPORTS_MQ)            | \
  60                                 (1ULL << VIRTIO_F_VERSION_1)   | \
  61                                 (1ULL << VHOST_F_LOG_ALL)      | \
  62                                 (1ULL << VHOST_USER_F_PROTOCOL_FEATURES) | \
  63                                 (1ULL << VIRTIO_NET_F_HOST_TSO4) | \
  64                                 (1ULL << VIRTIO_NET_F_HOST_TSO6) | \
  65                                 (1ULL << VIRTIO_NET_F_CSUM)    | \
  66                                 (1ULL << VIRTIO_NET_F_GUEST_CSUM) | \
  67                                 (1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
  68                                 (1ULL << VIRTIO_NET_F_GUEST_TSO6))
  69
  70 uint64_t VHOST_FEATURES = VHOST_SUPPORTED_FEATURES;
  71
  72 struct virtio_net *vhost_devices[MAX_VHOST_DEVICE];
  73
  74 /* device ops to add/remove device to/from data core. */
  75 struct virtio_net_device_ops const *notify_ops;
  76
  77 struct virtio_net *
  78 get_device(int vid)
  79 {
  80         struct virtio_net *dev = vhost_devices[vid];
  81
  82         if (unlikely(!dev)) {
  83                 RTE_LOG(ERR, VHOST_CONFIG,
  84                         "(%d) device not found.\n", vid);
  85         }
  86
  87         return dev;
  88 }
  89
  90 static void
  91 cleanup_vq(struct vhost_virtqueue *vq, int destroy)
  92 {
  93         if ((vq->callfd >= 0) && (destroy != 0))
  94                 close(vq->callfd);
  95         if (vq->kickfd >= 0)
  96                 close(vq->kickfd);
  97 }
  98
  99 /*
 100  * Unmap any memory, close any file descriptors and
 101  * free any memory owned by a device.
 102  */
 103 void
 104 cleanup_device(struct virtio_net *dev, int destroy)
 105 {
 106         uint32_t i;
 107
 108         vhost_backend_cleanup(dev);
 109
 110         for (i = 0; i < dev->virt_qp_nb; i++) {
 111                 cleanup_vq(dev->virtqueue[i * VIRTIO_QNUM + VIRTIO_RXQ], destroy);
 112                 cleanup_vq(dev->virtqueue[i * VIRTIO_QNUM + VIRTIO_TXQ], destroy);
 113         }
 114 }
 115
 116 /*
 117  * Release virtqueues and device memory.
 118  */
 119 static void
 120 free_device(struct virtio_net *dev)
 121 {
 122         uint32_t i;
 123         struct vhost_virtqueue *rxq, *txq;
 124
 125         for (i = 0; i < dev->virt_qp_nb; i++) {
 126                 rxq = dev->virtqueue[i * VIRTIO_QNUM + VIRTIO_RXQ];
 127                 txq = dev->virtqueue[i * VIRTIO_QNUM + VIRTIO_TXQ];
 128
 129                 rte_free(rxq->shadow_used_ring);
 130                 rte_free(txq->shadow_used_ring);
 131
 132                 /* rxq and txq are allocated together as queue-pair */
 133                 rte_free(rxq);
 134         }
 135
 136         rte_free(dev);
 137 }
 138
 139 static void
 140 init_vring_queue(struct vhost_virtqueue *vq, int qp_idx)
 141 {
 142         memset(vq, 0, sizeof(struct vhost_virtqueue));
 143
 144         vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
 145         vq->callfd = VIRTIO_UNINITIALIZED_EVENTFD;
 146
 147         /* Backends are set to -1 indicating an inactive device. */
 148         vq->backend = -1;
 149
 150         /* always set the default vq pair to enabled */
 151         if (qp_idx == 0)
 152                 vq->enabled = 1;
 153
 154         TAILQ_INIT(&vq->zmbuf_list);
 155 }
 156
 157 static void
 158 init_vring_queue_pair(struct virtio_net *dev, uint32_t qp_idx)
 159 {
 160         uint32_t base_idx = qp_idx * VIRTIO_QNUM;
 161
 162         init_vring_queue(dev->virtqueue[base_idx + VIRTIO_RXQ], qp_idx);
 163         init_vring_queue(dev->virtqueue[base_idx + VIRTIO_TXQ], qp_idx);
 164 }
 165
 166 static void
 167 reset_vring_queue(struct vhost_virtqueue *vq, int qp_idx)
 168 {
 169         int callfd;
 170
 171         callfd = vq->callfd;
 172         init_vring_queue(vq, qp_idx);
 173         vq->callfd = callfd;
 174 }
 175
 176 static void
 177 reset_vring_queue_pair(struct virtio_net *dev, uint32_t qp_idx)
 178 {
 179         uint32_t base_idx = qp_idx * VIRTIO_QNUM;
 180
 181         reset_vring_queue(dev->virtqueue[base_idx + VIRTIO_RXQ], qp_idx);
 182         reset_vring_queue(dev->virtqueue[base_idx + VIRTIO_TXQ], qp_idx);
 183 }
 184
 185 int
 186 alloc_vring_queue_pair(struct virtio_net *dev, uint32_t qp_idx)
 187 {
 188         struct vhost_virtqueue *virtqueue = NULL;
 189         uint32_t virt_rx_q_idx = qp_idx * VIRTIO_QNUM + VIRTIO_RXQ;
 190         uint32_t virt_tx_q_idx = qp_idx * VIRTIO_QNUM + VIRTIO_TXQ;
 191
 192         virtqueue = rte_malloc(NULL,
 193                                sizeof(struct vhost_virtqueue) * VIRTIO_QNUM, 0);
 194         if (virtqueue == NULL) {
 195                 RTE_LOG(ERR, VHOST_CONFIG,
 196                         "Failed to allocate memory for virt qp:%d.\n", qp_idx);
 197                 return -1;
 198         }
 199
 200         dev->virtqueue[virt_rx_q_idx] = virtqueue;
 201         dev->virtqueue[virt_tx_q_idx] = virtqueue + VIRTIO_TXQ;
 202
 203         init_vring_queue_pair(dev, qp_idx);
 204
 205         dev->virt_qp_nb += 1;
 206
 207         return 0;
 208 }
 209
 210 /*
 211  * Reset some variables in device structure, while keeping few
 212  * others untouched, such as vid, ifname, virt_qp_nb: they
 213  * should be same unless the device is removed.
 214  */
 215 void
 216 reset_device(struct virtio_net *dev)
 217 {
 218         uint32_t i;
 219
 220         dev->features = 0;
 221         dev->protocol_features = 0;
 222         dev->flags = 0;
 223
 224         for (i = 0; i < dev->virt_qp_nb; i++)
 225                 reset_vring_queue_pair(dev, i);
 226 }
 227
 228 /*
 229  * Invoked when there is a new vhost-user connection established (when
 230  * there is a new virtio device being attached).
 231  */
 232 int
 233 vhost_new_device(void)
 234 {
 235         struct virtio_net *dev;
 236         int i;
 237
 238         dev = rte_zmalloc(NULL, sizeof(struct virtio_net), 0);
 239         if (dev == NULL) {
 240                 RTE_LOG(ERR, VHOST_CONFIG,
 241                         "Failed to allocate memory for new dev.\n");
 242                 return -1;
 243         }
 244
 245         for (i = 0; i < MAX_VHOST_DEVICE; i++) {
 246                 if (vhost_devices[i] == NULL)
 247                         break;
 248         }
 249         if (i == MAX_VHOST_DEVICE) {
 250                 RTE_LOG(ERR, VHOST_CONFIG,
 251                         "Failed to find a free slot for new device.\n");
 252                 return -1;
 253         }
 254
 255         vhost_devices[i] = dev;
 256         dev->vid = i;
 257
 258         return i;
 259 }
 260
 261 /*
 262  * Invoked when there is the vhost-user connection is broken (when
 263  * the virtio device is being detached).
 264  */
 265 void
 266 vhost_destroy_device(int vid)
 267 {
 268         struct virtio_net *dev = get_device(vid);
 269
 270         if (dev == NULL)
 271                 return;
 272
 273         if (dev->flags & VIRTIO_DEV_RUNNING) {
 274                 dev->flags &= ~VIRTIO_DEV_RUNNING;
 275                 notify_ops->destroy_device(vid);
 276         }
 277
 278         cleanup_device(dev, 1);
 279         free_device(dev);
 280
 281         vhost_devices[vid] = NULL;
 282 }
 283
 284 void
 285 vhost_set_ifname(int vid, const char *if_name, unsigned int if_len)
 286 {
 287         struct virtio_net *dev;
 288         unsigned int len;
 289
 290         dev = get_device(vid);
 291         if (dev == NULL)
 292                 return;
 293
 294         len = if_len > sizeof(dev->ifname) ?
 295                 sizeof(dev->ifname) : if_len;
 296
 297         strncpy(dev->ifname, if_name, len);
 298         dev->ifname[sizeof(dev->ifname) - 1] = '\0';
 299 }
 300
 301 void
 302 vhost_enable_dequeue_zero_copy(int vid)
 303 {
 304         struct virtio_net *dev = get_device(vid);
 305
 306         if (dev == NULL)
 307                 return;
 308
 309         dev->dequeue_zero_copy = 1;
 310 }
 311
 312 int
 313 rte_vhost_get_numa_node(int vid)
 314 {
 315 #ifdef RTE_LIBRTE_VHOST_NUMA
 316         struct virtio_net *dev = get_device(vid);
 317         int numa_node;
 318         int ret;
 319
 320         if (dev == NULL)
 321                 return -1;
 322
 323         ret = get_mempolicy(&numa_node, NULL, 0, dev,
 324                             MPOL_F_NODE | MPOL_F_ADDR);
 325         if (ret < 0) {
 326                 RTE_LOG(ERR, VHOST_CONFIG,
 327                         "(%d) failed to query numa node: %d\n", vid, ret);
 328                 return -1;
 329         }
 330
 331         return numa_node;
 332 #else
 333         RTE_SET_USED(vid);
 334         return -1;
 335 #endif
 336 }
 337
 338 uint32_t
 339 rte_vhost_get_queue_num(int vid)
 340 {
 341         struct virtio_net *dev = get_device(vid);
 342
 343         if (dev == NULL)
 344                 return 0;
 345
 346         return dev->virt_qp_nb;
 347 }
 348
 349 int
 350 rte_vhost_get_ifname(int vid, char *buf, size_t len)
 351 {
 352         struct virtio_net *dev = get_device(vid);
 353
 354         if (dev == NULL)
 355                 return -1;
 356
 357         len = RTE_MIN(len, sizeof(dev->ifname));
 358
 359         strncpy(buf, dev->ifname, len);
 360         buf[len - 1] = '\0';
 361
 362         return 0;
 363 }
 364
 365 uint16_t
 366 rte_vhost_avail_entries(int vid, uint16_t queue_id)
 367 {
 368         struct virtio_net *dev;
 369         struct vhost_virtqueue *vq;
 370
 371         dev = get_device(vid);
 372         if (!dev)
 373                 return 0;
 374
 375         vq = dev->virtqueue[queue_id];
 376         if (!vq->enabled)
 377                 return 0;
 378
 379         return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx;
 380 }
 381
 382 int
 383 rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
 384 {
 385         struct virtio_net *dev = get_device(vid);
 386
 387         if (dev == NULL)
 388                 return -1;
 389
 390         if (enable) {
 391                 RTE_LOG(ERR, VHOST_CONFIG,
 392                         "guest notification isn't supported.\n");
 393                 return -1;
 394         }
 395
 396         dev->virtqueue[queue_id]->used->flags = VRING_USED_F_NO_NOTIFY;
 397         return 0;
 398 }
 399
 400 uint64_t rte_vhost_feature_get(void)
 401 {
 402         return VHOST_FEATURES;
 403 }
 404
 405 int rte_vhost_feature_disable(uint64_t feature_mask)
 406 {
 407         VHOST_FEATURES = VHOST_FEATURES & ~feature_mask;
 408         return 0;
 409 }
 410
 411 int rte_vhost_feature_enable(uint64_t feature_mask)
 412 {
 413         if ((feature_mask & VHOST_SUPPORTED_FEATURES) == feature_mask) {
 414                 VHOST_FEATURES = VHOST_FEATURES | feature_mask;
 415                 return 0;
 416         }
 417         return -1;
 418 }
 419
 420 /*
 421  * Register ops so that we can add/remove device to data core.
 422  */
 423 int
 424 rte_vhost_driver_callback_register(struct virtio_net_device_ops const * const ops)
 425 {
 426         notify_ops = ops;
 427
 428         return 0;
 429 }