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 <sys/types.h> 36 #include <unistd.h> 37 #include <fcntl.h> 38 #include <sys/socket.h> 39 40 #include <rte_malloc.h> 41 #include <rte_kvargs.h> 42 #include <rte_ethdev_vdev.h> 43 #include <rte_vdev.h> 44 #include <rte_alarm.h> 45 46 #include "virtio_ethdev.h" 47 #include "virtio_logs.h" 48 #include "virtio_pci.h" 49 #include "virtqueue.h" 50 #include "virtio_rxtx.h" 51 #include "virtio_user/virtio_user_dev.h" 52 53 #define virtio_user_get_dev(hw) \ 54 ((struct virtio_user_dev *)(hw)->virtio_user_dev) 55 56 static void 57 virtio_user_delayed_handler(void *param) 58 { 59 struct virtio_hw *hw = (struct virtio_hw *)param; 60 struct rte_eth_dev *dev = &rte_eth_devices[hw->port_id]; 61 62 rte_intr_callback_unregister(dev->intr_handle, 63 virtio_interrupt_handler, 64 dev); 65 } 66 67 static void 68 virtio_user_read_dev_config(struct virtio_hw *hw, size_t offset, 69 void *dst, int length) 70 { 71 int i; 72 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 73 74 if (offset == offsetof(struct virtio_net_config, mac) && 75 length == ETHER_ADDR_LEN) { 76 for (i = 0; i < ETHER_ADDR_LEN; ++i) 77 ((uint8_t *)dst)[i] = dev->mac_addr[i]; 78 return; 79 } 80 81 if (offset == offsetof(struct virtio_net_config, status)) { 82 char buf[128]; 83 84 if (dev->vhostfd >= 0) { 85 int r; 86 int flags; 87 88 flags = fcntl(dev->vhostfd, F_GETFL); 89 fcntl(dev->vhostfd, F_SETFL, flags | O_NONBLOCK); 90 r = recv(dev->vhostfd, buf, 128, MSG_PEEK); 91 if (r == 0 || (r < 0 && errno != EAGAIN)) { 92 dev->status &= (~VIRTIO_NET_S_LINK_UP); 93 PMD_DRV_LOG(ERR, "virtio-user port %u is down", 94 hw->port_id); 95 /* Only client mode is available now. Once the 96 * connection is broken, it can never be up 97 * again. Besides, this function could be called 98 * in the process of interrupt handling, 99 * callback cannot be unregistered here, set an 100 * alarm to do it. 101 */ 102 rte_eal_alarm_set(1, 103 virtio_user_delayed_handler, 104 (void *)hw); 105 } else { 106 dev->status |= VIRTIO_NET_S_LINK_UP; 107 } 108 fcntl(dev->vhostfd, F_SETFL, flags & (~O_NONBLOCK)); 109 } 110 *(uint16_t *)dst = dev->status; 111 } 112 113 if (offset == offsetof(struct virtio_net_config, max_virtqueue_pairs)) 114 *(uint16_t *)dst = dev->max_queue_pairs; 115 } 116 117 static void 118 virtio_user_write_dev_config(struct virtio_hw *hw, size_t offset, 119 const void *src, int length) 120 { 121 int i; 122 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 123 124 if ((offset == offsetof(struct virtio_net_config, mac)) && 125 (length == ETHER_ADDR_LEN)) 126 for (i = 0; i < ETHER_ADDR_LEN; ++i) 127 dev->mac_addr[i] = ((const uint8_t *)src)[i]; 128 else 129 PMD_DRV_LOG(ERR, "not supported offset=%zu, len=%d", 130 offset, length); 131 } 132 133 static void 134 virtio_user_reset(struct virtio_hw *hw) 135 { 136 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 137 138 if (dev->status & VIRTIO_CONFIG_STATUS_DRIVER_OK) 139 virtio_user_stop_device(dev); 140 } 141 142 static void 143 virtio_user_set_status(struct virtio_hw *hw, uint8_t status) 144 { 145 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 146 147 if (status & VIRTIO_CONFIG_STATUS_DRIVER_OK) 148 virtio_user_start_device(dev); 149 else if (status == VIRTIO_CONFIG_STATUS_RESET) 150 virtio_user_reset(hw); 151 dev->status = status; 152 } 153 154 static uint8_t 155 virtio_user_get_status(struct virtio_hw *hw) 156 { 157 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 158 159 return dev->status; 160 } 161 162 static uint64_t 163 virtio_user_get_features(struct virtio_hw *hw) 164 { 165 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 166 167 /* unmask feature bits defined in vhost user protocol */ 168 return dev->device_features & VIRTIO_PMD_SUPPORTED_GUEST_FEATURES; 169 } 170 171 static void 172 virtio_user_set_features(struct virtio_hw *hw, uint64_t features) 173 { 174 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 175 176 dev->features = features & dev->device_features; 177 } 178 179 static uint8_t 180 virtio_user_get_isr(struct virtio_hw *hw __rte_unused) 181 { 182 /* rxq interrupts and config interrupt are separated in virtio-user, 183 * here we only report config change. 184 */ 185 return VIRTIO_PCI_ISR_CONFIG; 186 } 187 188 static uint16_t 189 virtio_user_set_config_irq(struct virtio_hw *hw __rte_unused, 190 uint16_t vec __rte_unused) 191 { 192 return 0; 193 } 194 195 static uint16_t 196 virtio_user_set_queue_irq(struct virtio_hw *hw __rte_unused, 197 struct virtqueue *vq __rte_unused, 198 uint16_t vec) 199 { 200 /* pretend we have done that */ 201 return vec; 202 } 203 204 /* This function is to get the queue size, aka, number of descs, of a specified 205 * queue. Different with the VHOST_USER_GET_QUEUE_NUM, which is used to get the 206 * max supported queues. 207 */ 208 static uint16_t 209 virtio_user_get_queue_num(struct virtio_hw *hw, uint16_t queue_id __rte_unused) 210 { 211 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 212 213 /* Currently, each queue has same queue size */ 214 return dev->queue_size; 215 } 216 217 static int 218 virtio_user_setup_queue(struct virtio_hw *hw, struct virtqueue *vq) 219 { 220 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 221 uint16_t queue_idx = vq->vq_queue_index; 222 uint64_t desc_addr, avail_addr, used_addr; 223 224 desc_addr = (uintptr_t)vq->vq_ring_virt_mem; 225 avail_addr = desc_addr + vq->vq_nentries * sizeof(struct vring_desc); 226 used_addr = RTE_ALIGN_CEIL(avail_addr + offsetof(struct vring_avail, 227 ring[vq->vq_nentries]), 228 VIRTIO_PCI_VRING_ALIGN); 229 230 dev->vrings[queue_idx].num = vq->vq_nentries; 231 dev->vrings[queue_idx].desc = (void *)(uintptr_t)desc_addr; 232 dev->vrings[queue_idx].avail = (void *)(uintptr_t)avail_addr; 233 dev->vrings[queue_idx].used = (void *)(uintptr_t)used_addr; 234 235 return 0; 236 } 237 238 static void 239 virtio_user_del_queue(struct virtio_hw *hw, struct virtqueue *vq) 240 { 241 /* For legacy devices, write 0 to VIRTIO_PCI_QUEUE_PFN port, QEMU 242 * correspondingly stops the ioeventfds, and reset the status of 243 * the device. 244 * For modern devices, set queue desc, avail, used in PCI bar to 0, 245 * not see any more behavior in QEMU. 246 * 247 * Here we just care about what information to deliver to vhost-user 248 * or vhost-kernel. So we just close ioeventfd for now. 249 */ 250 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 251 252 close(dev->callfds[vq->vq_queue_index]); 253 close(dev->kickfds[vq->vq_queue_index]); 254 } 255 256 static void 257 virtio_user_notify_queue(struct virtio_hw *hw, struct virtqueue *vq) 258 { 259 uint64_t buf = 1; 260 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 261 262 if (hw->cvq && (hw->cvq->vq == vq)) { 263 virtio_user_handle_cq(dev, vq->vq_queue_index); 264 return; 265 } 266 267 if (write(dev->kickfds[vq->vq_queue_index], &buf, sizeof(buf)) < 0) 268 PMD_DRV_LOG(ERR, "failed to kick backend: %s", 269 strerror(errno)); 270 } 271 272 const struct virtio_pci_ops virtio_user_ops = { 273 .read_dev_cfg = virtio_user_read_dev_config, 274 .write_dev_cfg = virtio_user_write_dev_config, 275 .reset = virtio_user_reset, 276 .get_status = virtio_user_get_status, 277 .set_status = virtio_user_set_status, 278 .get_features = virtio_user_get_features, 279 .set_features = virtio_user_set_features, 280 .get_isr = virtio_user_get_isr, 281 .set_config_irq = virtio_user_set_config_irq, 282 .set_queue_irq = virtio_user_set_queue_irq, 283 .get_queue_num = virtio_user_get_queue_num, 284 .setup_queue = virtio_user_setup_queue, 285 .del_queue = virtio_user_del_queue, 286 .notify_queue = virtio_user_notify_queue, 287 }; 288 289 static const char *valid_args[] = { 290 #define VIRTIO_USER_ARG_QUEUES_NUM "queues" 291 VIRTIO_USER_ARG_QUEUES_NUM, 292 #define VIRTIO_USER_ARG_CQ_NUM "cq" 293 VIRTIO_USER_ARG_CQ_NUM, 294 #define VIRTIO_USER_ARG_MAC "mac" 295 VIRTIO_USER_ARG_MAC, 296 #define VIRTIO_USER_ARG_PATH "path" 297 VIRTIO_USER_ARG_PATH, 298 #define VIRTIO_USER_ARG_QUEUE_SIZE "queue_size" 299 VIRTIO_USER_ARG_QUEUE_SIZE, 300 #define VIRTIO_USER_ARG_INTERFACE_NAME "iface" 301 VIRTIO_USER_ARG_INTERFACE_NAME, 302 NULL 303 }; 304 305 #define VIRTIO_USER_DEF_CQ_EN 0 306 #define VIRTIO_USER_DEF_Q_NUM 1 307 #define VIRTIO_USER_DEF_Q_SZ 256 308 309 static int 310 get_string_arg(const char *key __rte_unused, 311 const char *value, void *extra_args) 312 { 313 if (!value || !extra_args) 314 return -EINVAL; 315 316 *(char **)extra_args = strdup(value); 317 318 if (!*(char **)extra_args) 319 return -ENOMEM; 320 321 return 0; 322 } 323 324 static int 325 get_integer_arg(const char *key __rte_unused, 326 const char *value, void *extra_args) 327 { 328 if (!value || !extra_args) 329 return -EINVAL; 330 331 *(uint64_t *)extra_args = strtoull(value, NULL, 0); 332 333 return 0; 334 } 335 336 static struct rte_vdev_driver virtio_user_driver; 337 338 static struct rte_eth_dev * 339 virtio_user_eth_dev_alloc(struct rte_vdev_device *vdev) 340 { 341 struct rte_eth_dev *eth_dev; 342 struct rte_eth_dev_data *data; 343 struct virtio_hw *hw; 344 struct virtio_user_dev *dev; 345 346 eth_dev = rte_eth_vdev_allocate(vdev, sizeof(*hw)); 347 if (!eth_dev) { 348 PMD_INIT_LOG(ERR, "cannot alloc rte_eth_dev"); 349 return NULL; 350 } 351 352 data = eth_dev->data; 353 hw = eth_dev->data->dev_private; 354 355 dev = rte_zmalloc(NULL, sizeof(*dev), 0); 356 if (!dev) { 357 PMD_INIT_LOG(ERR, "malloc virtio_user_dev failed"); 358 rte_eth_dev_release_port(eth_dev); 359 rte_free(hw); 360 return NULL; 361 } 362 363 hw->port_id = data->port_id; 364 dev->port_id = data->port_id; 365 virtio_hw_internal[hw->port_id].vtpci_ops = &virtio_user_ops; 366 /* 367 * MSIX is required to enable LSC (see virtio_init_device). 368 * Here just pretend that we support msix. 369 */ 370 hw->use_msix = 1; 371 hw->modern = 0; 372 hw->use_simple_rxtx = 0; 373 hw->virtio_user_dev = dev; 374 data->dev_flags = RTE_ETH_DEV_DETACHABLE; 375 return eth_dev; 376 } 377 378 static void 379 virtio_user_eth_dev_free(struct rte_eth_dev *eth_dev) 380 { 381 struct rte_eth_dev_data *data = eth_dev->data; 382 struct virtio_hw *hw = data->dev_private; 383 384 rte_free(hw->virtio_user_dev); 385 rte_free(hw); 386 rte_eth_dev_release_port(eth_dev); 387 } 388 389 /* Dev initialization routine. Invoked once for each virtio vdev at 390 * EAL init time, see rte_bus_probe(). 391 * Returns 0 on success. 392 */ 393 static int 394 virtio_user_pmd_probe(struct rte_vdev_device *dev) 395 { 396 struct rte_kvargs *kvlist = NULL; 397 struct rte_eth_dev *eth_dev; 398 struct virtio_hw *hw; 399 uint64_t queues = VIRTIO_USER_DEF_Q_NUM; 400 uint64_t cq = VIRTIO_USER_DEF_CQ_EN; 401 uint64_t queue_size = VIRTIO_USER_DEF_Q_SZ; 402 char *path = NULL; 403 char *ifname = NULL; 404 char *mac_addr = NULL; 405 int ret = -1; 406 407 kvlist = rte_kvargs_parse(rte_vdev_device_args(dev), valid_args); 408 if (!kvlist) { 409 PMD_INIT_LOG(ERR, "error when parsing param"); 410 goto end; 411 } 412 413 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_PATH) == 1) { 414 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_PATH, 415 &get_string_arg, &path) < 0) { 416 PMD_INIT_LOG(ERR, "error to parse %s", 417 VIRTIO_USER_ARG_PATH); 418 goto end; 419 } 420 } else { 421 PMD_INIT_LOG(ERR, "arg %s is mandatory for virtio_user", 422 VIRTIO_USER_ARG_QUEUE_SIZE); 423 goto end; 424 } 425 426 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME) == 1) { 427 if (is_vhost_user_by_type(path)) { 428 PMD_INIT_LOG(ERR, 429 "arg %s applies only to vhost-kernel backend", 430 VIRTIO_USER_ARG_INTERFACE_NAME); 431 goto end; 432 } 433 434 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME, 435 &get_string_arg, &ifname) < 0) { 436 PMD_INIT_LOG(ERR, "error to parse %s", 437 VIRTIO_USER_ARG_INTERFACE_NAME); 438 goto end; 439 } 440 } 441 442 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_MAC) == 1) { 443 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_MAC, 444 &get_string_arg, &mac_addr) < 0) { 445 PMD_INIT_LOG(ERR, "error to parse %s", 446 VIRTIO_USER_ARG_MAC); 447 goto end; 448 } 449 } 450 451 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE) == 1) { 452 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE, 453 &get_integer_arg, &queue_size) < 0) { 454 PMD_INIT_LOG(ERR, "error to parse %s", 455 VIRTIO_USER_ARG_QUEUE_SIZE); 456 goto end; 457 } 458 } 459 460 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUES_NUM) == 1) { 461 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUES_NUM, 462 &get_integer_arg, &queues) < 0) { 463 PMD_INIT_LOG(ERR, "error to parse %s", 464 VIRTIO_USER_ARG_QUEUES_NUM); 465 goto end; 466 } 467 } 468 469 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_CQ_NUM) == 1) { 470 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_CQ_NUM, 471 &get_integer_arg, &cq) < 0) { 472 PMD_INIT_LOG(ERR, "error to parse %s", 473 VIRTIO_USER_ARG_CQ_NUM); 474 goto end; 475 } 476 } else if (queues > 1) { 477 cq = 1; 478 } 479 480 if (queues > 1 && cq == 0) { 481 PMD_INIT_LOG(ERR, "multi-q requires ctrl-q"); 482 goto end; 483 } 484 485 if (queues > VIRTIO_MAX_VIRTQUEUE_PAIRS) { 486 PMD_INIT_LOG(ERR, "arg %s %" PRIu64 " exceeds the limit %u", 487 VIRTIO_USER_ARG_QUEUES_NUM, queues, 488 VIRTIO_MAX_VIRTQUEUE_PAIRS); 489 goto end; 490 } 491 492 if (rte_eal_process_type() == RTE_PROC_PRIMARY) { 493 eth_dev = virtio_user_eth_dev_alloc(dev); 494 if (!eth_dev) { 495 PMD_INIT_LOG(ERR, "virtio_user fails to alloc device"); 496 goto end; 497 } 498 499 hw = eth_dev->data->dev_private; 500 if (virtio_user_dev_init(hw->virtio_user_dev, path, queues, cq, 501 queue_size, mac_addr, &ifname) < 0) { 502 PMD_INIT_LOG(ERR, "virtio_user_dev_init fails"); 503 virtio_user_eth_dev_free(eth_dev); 504 goto end; 505 } 506 } else { 507 eth_dev = rte_eth_dev_attach_secondary(rte_vdev_device_name(dev)); 508 if (!eth_dev) 509 goto end; 510 } 511 512 /* previously called by rte_pci_probe() for physical dev */ 513 if (eth_virtio_dev_init(eth_dev) < 0) { 514 PMD_INIT_LOG(ERR, "eth_virtio_dev_init fails"); 515 virtio_user_eth_dev_free(eth_dev); 516 goto end; 517 } 518 ret = 0; 519 520 end: 521 if (kvlist) 522 rte_kvargs_free(kvlist); 523 if (path) 524 free(path); 525 if (mac_addr) 526 free(mac_addr); 527 if (ifname) 528 free(ifname); 529 return ret; 530 } 531 532 /** Called by rte_eth_dev_detach() */ 533 static int 534 virtio_user_pmd_remove(struct rte_vdev_device *vdev) 535 { 536 const char *name; 537 struct rte_eth_dev *eth_dev; 538 struct virtio_hw *hw; 539 struct virtio_user_dev *dev; 540 541 if (!vdev) 542 return -EINVAL; 543 544 name = rte_vdev_device_name(vdev); 545 PMD_DRV_LOG(INFO, "Un-Initializing %s", name); 546 eth_dev = rte_eth_dev_allocated(name); 547 if (!eth_dev) 548 return -ENODEV; 549 550 /* make sure the device is stopped, queues freed */ 551 rte_eth_dev_close(eth_dev->data->port_id); 552 553 hw = eth_dev->data->dev_private; 554 dev = hw->virtio_user_dev; 555 virtio_user_dev_uninit(dev); 556 557 rte_free(eth_dev->data->dev_private); 558 rte_eth_dev_release_port(eth_dev); 559 560 return 0; 561 } 562 563 static struct rte_vdev_driver virtio_user_driver = { 564 .probe = virtio_user_pmd_probe, 565 .remove = virtio_user_pmd_remove, 566 }; 567 568 RTE_PMD_REGISTER_VDEV(net_virtio_user, virtio_user_driver); 569 RTE_PMD_REGISTER_ALIAS(net_virtio_user, virtio_user); 570 RTE_PMD_REGISTER_PARAM_STRING(net_virtio_user, 571 "path=<path> " 572 "mac=<mac addr> " 573 "cq=<int> " 574 "queue_size=<int> " 575 "queues=<int> " 576 "iface=<string>"); 577