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 <stdint.h> 35 #include <stdio.h> 36 #include <fcntl.h> 37 #include <string.h> 38 #include <errno.h> 39 #include <sys/mman.h> 40 #include <unistd.h> 41 #include <sys/eventfd.h> 42 #include <sys/types.h> 43 #include <sys/stat.h> 44 #include <unistd.h> 45 46 #include "vhost.h" 47 #include "virtio_user_dev.h" 48 #include "../virtio_ethdev.h" 49 50 static int 51 virtio_user_create_queue(struct virtio_user_dev *dev, uint32_t queue_sel) 52 { 53 /* Of all per virtqueue MSGs, make sure VHOST_SET_VRING_CALL come 54 * firstly because vhost depends on this msg to allocate virtqueue 55 * pair. 56 */ 57 int callfd; 58 struct vhost_vring_file file; 59 60 /* May use invalid flag, but some backend leverages kickfd and callfd as 61 * criteria to judge if dev is alive. so finally we use real event_fd. 62 */ 63 callfd = eventfd(0, EFD_CLOEXEC | EFD_NONBLOCK); 64 if (callfd < 0) { 65 PMD_DRV_LOG(ERR, "callfd error, %s", strerror(errno)); 66 return -1; 67 } 68 file.index = queue_sel; 69 file.fd = callfd; 70 dev->ops->send_request(dev, VHOST_USER_SET_VRING_CALL, &file); 71 dev->callfds[queue_sel] = callfd; 72 73 return 0; 74 } 75 76 static int 77 virtio_user_kick_queue(struct virtio_user_dev *dev, uint32_t queue_sel) 78 { 79 int kickfd; 80 struct vhost_vring_file file; 81 struct vhost_vring_state state; 82 struct vring *vring = &dev->vrings[queue_sel]; 83 struct vhost_vring_addr addr = { 84 .index = queue_sel, 85 .desc_user_addr = (uint64_t)(uintptr_t)vring->desc, 86 .avail_user_addr = (uint64_t)(uintptr_t)vring->avail, 87 .used_user_addr = (uint64_t)(uintptr_t)vring->used, 88 .log_guest_addr = 0, 89 .flags = 0, /* disable log */ 90 }; 91 92 state.index = queue_sel; 93 state.num = vring->num; 94 dev->ops->send_request(dev, VHOST_USER_SET_VRING_NUM, &state); 95 96 state.index = queue_sel; 97 state.num = 0; /* no reservation */ 98 dev->ops->send_request(dev, VHOST_USER_SET_VRING_BASE, &state); 99 100 dev->ops->send_request(dev, VHOST_USER_SET_VRING_ADDR, &addr); 101 102 /* Of all per virtqueue MSGs, make sure VHOST_USER_SET_VRING_KICK comes 103 * lastly because vhost depends on this msg to judge if 104 * virtio is ready. 105 */ 106 kickfd = eventfd(0, EFD_CLOEXEC | EFD_NONBLOCK); 107 if (kickfd < 0) { 108 PMD_DRV_LOG(ERR, "kickfd error, %s", strerror(errno)); 109 return -1; 110 } 111 file.index = queue_sel; 112 file.fd = kickfd; 113 dev->ops->send_request(dev, VHOST_USER_SET_VRING_KICK, &file); 114 dev->kickfds[queue_sel] = kickfd; 115 116 return 0; 117 } 118 119 static int 120 virtio_user_queue_setup(struct virtio_user_dev *dev, 121 int (*fn)(struct virtio_user_dev *, uint32_t)) 122 { 123 uint32_t i, queue_sel; 124 125 for (i = 0; i < dev->max_queue_pairs; ++i) { 126 queue_sel = 2 * i + VTNET_SQ_RQ_QUEUE_IDX; 127 if (fn(dev, queue_sel) < 0) { 128 PMD_DRV_LOG(INFO, "setup rx vq fails: %u", i); 129 return -1; 130 } 131 } 132 for (i = 0; i < dev->max_queue_pairs; ++i) { 133 queue_sel = 2 * i + VTNET_SQ_TQ_QUEUE_IDX; 134 if (fn(dev, queue_sel) < 0) { 135 PMD_DRV_LOG(INFO, "setup tx vq fails: %u", i); 136 return -1; 137 } 138 } 139 140 return 0; 141 } 142 143 int 144 virtio_user_start_device(struct virtio_user_dev *dev) 145 { 146 uint64_t features; 147 int ret; 148 149 /* Step 0: tell vhost to create queues */ 150 if (virtio_user_queue_setup(dev, virtio_user_create_queue) < 0) 151 goto error; 152 153 /* Step 1: set features */ 154 features = dev->features; 155 /* Strip VIRTIO_NET_F_MAC, as MAC address is handled in vdev init */ 156 features &= ~(1ull << VIRTIO_NET_F_MAC); 157 /* Strip VIRTIO_NET_F_CTRL_VQ, as devices do not really need to know */ 158 features &= ~(1ull << VIRTIO_NET_F_CTRL_VQ); 159 ret = dev->ops->send_request(dev, VHOST_USER_SET_FEATURES, &features); 160 if (ret < 0) 161 goto error; 162 PMD_DRV_LOG(INFO, "set features: %" PRIx64, features); 163 164 /* Step 2: share memory regions */ 165 ret = dev->ops->send_request(dev, VHOST_USER_SET_MEM_TABLE, NULL); 166 if (ret < 0) 167 goto error; 168 169 /* Step 3: kick queues */ 170 if (virtio_user_queue_setup(dev, virtio_user_kick_queue) < 0) 171 goto error; 172 173 /* Step 4: enable queues 174 * we enable the 1st queue pair by default. 175 */ 176 dev->ops->enable_qp(dev, 0, 1); 177 178 return 0; 179 error: 180 /* TODO: free resource here or caller to check */ 181 return -1; 182 } 183 184 int virtio_user_stop_device(struct virtio_user_dev *dev) 185 { 186 uint32_t i; 187 188 for (i = 0; i < dev->max_queue_pairs * 2; ++i) { 189 close(dev->callfds[i]); 190 close(dev->kickfds[i]); 191 } 192 193 for (i = 0; i < dev->max_queue_pairs; ++i) 194 dev->ops->enable_qp(dev, i, 0); 195 196 free(dev->ifname); 197 dev->ifname = NULL; 198 199 return 0; 200 } 201 202 static inline void 203 parse_mac(struct virtio_user_dev *dev, const char *mac) 204 { 205 int i, r; 206 uint32_t tmp[ETHER_ADDR_LEN]; 207 208 if (!mac) 209 return; 210 211 r = sscanf(mac, "%x:%x:%x:%x:%x:%x", &tmp[0], 212 &tmp[1], &tmp[2], &tmp[3], &tmp[4], &tmp[5]); 213 if (r == ETHER_ADDR_LEN) { 214 for (i = 0; i < ETHER_ADDR_LEN; ++i) 215 dev->mac_addr[i] = (uint8_t)tmp[i]; 216 dev->mac_specified = 1; 217 } else { 218 /* ignore the wrong mac, use random mac */ 219 PMD_DRV_LOG(ERR, "wrong format of mac: %s", mac); 220 } 221 } 222 223 static int 224 is_vhost_user_by_type(const char *path) 225 { 226 struct stat sb; 227 228 if (stat(path, &sb) == -1) 229 return 0; 230 231 return S_ISSOCK(sb.st_mode); 232 } 233 234 static int 235 virtio_user_dev_setup(struct virtio_user_dev *dev) 236 { 237 uint32_t i, q; 238 239 dev->vhostfd = -1; 240 for (i = 0; i < VIRTIO_MAX_VIRTQUEUES * 2 + 1; ++i) { 241 dev->kickfds[i] = -1; 242 dev->callfds[i] = -1; 243 } 244 245 dev->vhostfds = NULL; 246 dev->tapfds = NULL; 247 248 if (is_vhost_user_by_type(dev->path)) { 249 dev->ops = &ops_user; 250 } else { 251 dev->ops = &ops_kernel; 252 253 dev->vhostfds = malloc(dev->max_queue_pairs * sizeof(int)); 254 dev->tapfds = malloc(dev->max_queue_pairs * sizeof(int)); 255 if (!dev->vhostfds || !dev->tapfds) { 256 PMD_INIT_LOG(ERR, "Failed to malloc"); 257 return -1; 258 } 259 260 for (q = 0; q < dev->max_queue_pairs; ++q) { 261 dev->vhostfds[q] = -1; 262 dev->tapfds[q] = -1; 263 } 264 } 265 266 return dev->ops->setup(dev); 267 } 268 269 int 270 virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues, 271 int cq, int queue_size, const char *mac) 272 { 273 snprintf(dev->path, PATH_MAX, "%s", path); 274 dev->max_queue_pairs = queues; 275 dev->queue_pairs = 1; /* mq disabled by default */ 276 dev->queue_size = queue_size; 277 dev->mac_specified = 0; 278 parse_mac(dev, mac); 279 280 if (virtio_user_dev_setup(dev) < 0) { 281 PMD_INIT_LOG(ERR, "backend set up fails"); 282 return -1; 283 } 284 if (dev->ops->send_request(dev, VHOST_USER_SET_OWNER, NULL) < 0) { 285 PMD_INIT_LOG(ERR, "set_owner fails: %s", strerror(errno)); 286 return -1; 287 } 288 289 if (dev->ops->send_request(dev, VHOST_USER_GET_FEATURES, 290 &dev->device_features) < 0) { 291 PMD_INIT_LOG(ERR, "get_features failed: %s", strerror(errno)); 292 return -1; 293 } 294 if (dev->mac_specified) 295 dev->device_features |= (1ull << VIRTIO_NET_F_MAC); 296 297 if (cq) { 298 /* device does not really need to know anything about CQ, 299 * so if necessary, we just claim to support CQ 300 */ 301 dev->device_features |= (1ull << VIRTIO_NET_F_CTRL_VQ); 302 } else { 303 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_VQ); 304 /* Also disable features depends on VIRTIO_NET_F_CTRL_VQ */ 305 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_RX); 306 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_VLAN); 307 dev->device_features &= ~(1ull << VIRTIO_NET_F_GUEST_ANNOUNCE); 308 dev->device_features &= ~(1ull << VIRTIO_NET_F_MQ); 309 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_MAC_ADDR); 310 } 311 312 return 0; 313 } 314 315 void 316 virtio_user_dev_uninit(struct virtio_user_dev *dev) 317 { 318 uint32_t i; 319 320 virtio_user_stop_device(dev); 321 322 close(dev->vhostfd); 323 324 if (dev->vhostfds) { 325 for (i = 0; i < dev->max_queue_pairs; ++i) 326 close(dev->vhostfds[i]); 327 free(dev->vhostfds); 328 free(dev->tapfds); 329 } 330 } 331 332 static uint8_t 333 virtio_user_handle_mq(struct virtio_user_dev *dev, uint16_t q_pairs) 334 { 335 uint16_t i; 336 uint8_t ret = 0; 337 338 if (q_pairs > dev->max_queue_pairs) { 339 PMD_INIT_LOG(ERR, "multi-q config %u, but only %u supported", 340 q_pairs, dev->max_queue_pairs); 341 return -1; 342 } 343 344 for (i = 0; i < q_pairs; ++i) 345 ret |= dev->ops->enable_qp(dev, i, 1); 346 for (i = q_pairs; i < dev->max_queue_pairs; ++i) 347 ret |= dev->ops->enable_qp(dev, i, 0); 348 349 dev->queue_pairs = q_pairs; 350 351 return ret; 352 } 353 354 static uint32_t 355 virtio_user_handle_ctrl_msg(struct virtio_user_dev *dev, struct vring *vring, 356 uint16_t idx_hdr) 357 { 358 struct virtio_net_ctrl_hdr *hdr; 359 virtio_net_ctrl_ack status = ~0; 360 uint16_t i, idx_data, idx_status; 361 uint32_t n_descs = 0; 362 363 /* locate desc for header, data, and status */ 364 idx_data = vring->desc[idx_hdr].next; 365 n_descs++; 366 367 i = idx_data; 368 while (vring->desc[i].flags == VRING_DESC_F_NEXT) { 369 i = vring->desc[i].next; 370 n_descs++; 371 } 372 373 /* locate desc for status */ 374 idx_status = i; 375 n_descs++; 376 377 hdr = (void *)(uintptr_t)vring->desc[idx_hdr].addr; 378 if (hdr->class == VIRTIO_NET_CTRL_MQ && 379 hdr->cmd == VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET) { 380 uint16_t queues; 381 382 queues = *(uint16_t *)(uintptr_t)vring->desc[idx_data].addr; 383 status = virtio_user_handle_mq(dev, queues); 384 } 385 386 /* Update status */ 387 *(virtio_net_ctrl_ack *)(uintptr_t)vring->desc[idx_status].addr = status; 388 389 return n_descs; 390 } 391 392 void 393 virtio_user_handle_cq(struct virtio_user_dev *dev, uint16_t queue_idx) 394 { 395 uint16_t avail_idx, desc_idx; 396 struct vring_used_elem *uep; 397 uint32_t n_descs; 398 struct vring *vring = &dev->vrings[queue_idx]; 399 400 /* Consume avail ring, using used ring idx as first one */ 401 while (vring->used->idx != vring->avail->idx) { 402 avail_idx = (vring->used->idx) & (vring->num - 1); 403 desc_idx = vring->avail->ring[avail_idx]; 404 405 n_descs = virtio_user_handle_ctrl_msg(dev, vring, desc_idx); 406 407 /* Update used ring */ 408 uep = &vring->used->ring[avail_idx]; 409 uep->id = avail_idx; 410 uep->len = n_descs; 411 412 vring->used->idx++; 413 } 414 } 415