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 if (fcntl(dev->vhostfd, F_SETFL, 90 flags | O_NONBLOCK) == -1) { 91 PMD_DRV_LOG(ERR, "error setting O_NONBLOCK flag"); 92 return; 93 } 94 r = recv(dev->vhostfd, buf, 128, MSG_PEEK); 95 if (r == 0 || (r < 0 && errno != EAGAIN)) { 96 dev->status &= (~VIRTIO_NET_S_LINK_UP); 97 PMD_DRV_LOG(ERR, "virtio-user port %u is down", 98 hw->port_id); 99 /* Only client mode is available now. Once the 100 * connection is broken, it can never be up 101 * again. Besides, this function could be called 102 * in the process of interrupt handling, 103 * callback cannot be unregistered here, set an 104 * alarm to do it. 105 */ 106 rte_eal_alarm_set(1, 107 virtio_user_delayed_handler, 108 (void *)hw); 109 } else { 110 dev->status |= VIRTIO_NET_S_LINK_UP; 111 } 112 fcntl(dev->vhostfd, F_SETFL, flags & (~O_NONBLOCK)); 113 } 114 *(uint16_t *)dst = dev->status; 115 } 116 117 if (offset == offsetof(struct virtio_net_config, max_virtqueue_pairs)) 118 *(uint16_t *)dst = dev->max_queue_pairs; 119 } 120 121 static void 122 virtio_user_write_dev_config(struct virtio_hw *hw, size_t offset, 123 const void *src, int length) 124 { 125 int i; 126 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 127 128 if ((offset == offsetof(struct virtio_net_config, mac)) && 129 (length == ETHER_ADDR_LEN)) 130 for (i = 0; i < ETHER_ADDR_LEN; ++i) 131 dev->mac_addr[i] = ((const uint8_t *)src)[i]; 132 else 133 PMD_DRV_LOG(ERR, "not supported offset=%zu, len=%d", 134 offset, length); 135 } 136 137 static void 138 virtio_user_reset(struct virtio_hw *hw) 139 { 140 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 141 142 if (dev->status & VIRTIO_CONFIG_STATUS_DRIVER_OK) 143 virtio_user_stop_device(dev); 144 } 145 146 static void 147 virtio_user_set_status(struct virtio_hw *hw, uint8_t status) 148 { 149 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 150 151 if (status & VIRTIO_CONFIG_STATUS_DRIVER_OK) 152 virtio_user_start_device(dev); 153 else if (status == VIRTIO_CONFIG_STATUS_RESET) 154 virtio_user_reset(hw); 155 dev->status = status; 156 } 157 158 static uint8_t 159 virtio_user_get_status(struct virtio_hw *hw) 160 { 161 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 162 163 return dev->status; 164 } 165 166 static uint64_t 167 virtio_user_get_features(struct virtio_hw *hw) 168 { 169 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 170 171 /* unmask feature bits defined in vhost user protocol */ 172 return dev->device_features & VIRTIO_PMD_SUPPORTED_GUEST_FEATURES; 173 } 174 175 static void 176 virtio_user_set_features(struct virtio_hw *hw, uint64_t features) 177 { 178 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 179 180 dev->features = features & dev->device_features; 181 } 182 183 static uint8_t 184 virtio_user_get_isr(struct virtio_hw *hw __rte_unused) 185 { 186 /* rxq interrupts and config interrupt are separated in virtio-user, 187 * here we only report config change. 188 */ 189 return VIRTIO_PCI_ISR_CONFIG; 190 } 191 192 static uint16_t 193 virtio_user_set_config_irq(struct virtio_hw *hw __rte_unused, 194 uint16_t vec __rte_unused) 195 { 196 return 0; 197 } 198 199 static uint16_t 200 virtio_user_set_queue_irq(struct virtio_hw *hw __rte_unused, 201 struct virtqueue *vq __rte_unused, 202 uint16_t vec) 203 { 204 /* pretend we have done that */ 205 return vec; 206 } 207 208 /* This function is to get the queue size, aka, number of descs, of a specified 209 * queue. Different with the VHOST_USER_GET_QUEUE_NUM, which is used to get the 210 * max supported queues. 211 */ 212 static uint16_t 213 virtio_user_get_queue_num(struct virtio_hw *hw, uint16_t queue_id __rte_unused) 214 { 215 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 216 217 /* Currently, each queue has same queue size */ 218 return dev->queue_size; 219 } 220 221 static int 222 virtio_user_setup_queue(struct virtio_hw *hw, struct virtqueue *vq) 223 { 224 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 225 uint16_t queue_idx = vq->vq_queue_index; 226 uint64_t desc_addr, avail_addr, used_addr; 227 228 desc_addr = (uintptr_t)vq->vq_ring_virt_mem; 229 avail_addr = desc_addr + vq->vq_nentries * sizeof(struct vring_desc); 230 used_addr = RTE_ALIGN_CEIL(avail_addr + offsetof(struct vring_avail, 231 ring[vq->vq_nentries]), 232 VIRTIO_PCI_VRING_ALIGN); 233 234 dev->vrings[queue_idx].num = vq->vq_nentries; 235 dev->vrings[queue_idx].desc = (void *)(uintptr_t)desc_addr; 236 dev->vrings[queue_idx].avail = (void *)(uintptr_t)avail_addr; 237 dev->vrings[queue_idx].used = (void *)(uintptr_t)used_addr; 238 239 return 0; 240 } 241 242 static void 243 virtio_user_del_queue(struct virtio_hw *hw, struct virtqueue *vq) 244 { 245 /* For legacy devices, write 0 to VIRTIO_PCI_QUEUE_PFN port, QEMU 246 * correspondingly stops the ioeventfds, and reset the status of 247 * the device. 248 * For modern devices, set queue desc, avail, used in PCI bar to 0, 249 * not see any more behavior in QEMU. 250 * 251 * Here we just care about what information to deliver to vhost-user 252 * or vhost-kernel. So we just close ioeventfd for now. 253 */ 254 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 255 256 close(dev->callfds[vq->vq_queue_index]); 257 close(dev->kickfds[vq->vq_queue_index]); 258 } 259 260 static void 261 virtio_user_notify_queue(struct virtio_hw *hw, struct virtqueue *vq) 262 { 263 uint64_t buf = 1; 264 struct virtio_user_dev *dev = virtio_user_get_dev(hw); 265 266 if (hw->cvq && (hw->cvq->vq == vq)) { 267 virtio_user_handle_cq(dev, vq->vq_queue_index); 268 return; 269 } 270 271 if (write(dev->kickfds[vq->vq_queue_index], &buf, sizeof(buf)) < 0) 272 PMD_DRV_LOG(ERR, "failed to kick backend: %s", 273 strerror(errno)); 274 } 275 276 const struct virtio_pci_ops virtio_user_ops = { 277 .read_dev_cfg = virtio_user_read_dev_config, 278 .write_dev_cfg = virtio_user_write_dev_config, 279 .reset = virtio_user_reset, 280 .get_status = virtio_user_get_status, 281 .set_status = virtio_user_set_status, 282 .get_features = virtio_user_get_features, 283 .set_features = virtio_user_set_features, 284 .get_isr = virtio_user_get_isr, 285 .set_config_irq = virtio_user_set_config_irq, 286 .set_queue_irq = virtio_user_set_queue_irq, 287 .get_queue_num = virtio_user_get_queue_num, 288 .setup_queue = virtio_user_setup_queue, 289 .del_queue = virtio_user_del_queue, 290 .notify_queue = virtio_user_notify_queue, 291 }; 292 293 static const char *valid_args[] = { 294 #define VIRTIO_USER_ARG_QUEUES_NUM "queues" 295 VIRTIO_USER_ARG_QUEUES_NUM, 296 #define VIRTIO_USER_ARG_CQ_NUM "cq" 297 VIRTIO_USER_ARG_CQ_NUM, 298 #define VIRTIO_USER_ARG_MAC "mac" 299 VIRTIO_USER_ARG_MAC, 300 #define VIRTIO_USER_ARG_PATH "path" 301 VIRTIO_USER_ARG_PATH, 302 #define VIRTIO_USER_ARG_QUEUE_SIZE "queue_size" 303 VIRTIO_USER_ARG_QUEUE_SIZE, 304 #define VIRTIO_USER_ARG_INTERFACE_NAME "iface" 305 VIRTIO_USER_ARG_INTERFACE_NAME, 306 NULL 307 }; 308 309 #define VIRTIO_USER_DEF_CQ_EN 0 310 #define VIRTIO_USER_DEF_Q_NUM 1 311 #define VIRTIO_USER_DEF_Q_SZ 256 312 313 static int 314 get_string_arg(const char *key __rte_unused, 315 const char *value, void *extra_args) 316 { 317 if (!value || !extra_args) 318 return -EINVAL; 319 320 *(char **)extra_args = strdup(value); 321 322 if (!*(char **)extra_args) 323 return -ENOMEM; 324 325 return 0; 326 } 327 328 static int 329 get_integer_arg(const char *key __rte_unused, 330 const char *value, void *extra_args) 331 { 332 if (!value || !extra_args) 333 return -EINVAL; 334 335 *(uint64_t *)extra_args = strtoull(value, NULL, 0); 336 337 return 0; 338 } 339 340 static struct rte_vdev_driver virtio_user_driver; 341 342 static struct rte_eth_dev * 343 virtio_user_eth_dev_alloc(struct rte_vdev_device *vdev) 344 { 345 struct rte_eth_dev *eth_dev; 346 struct rte_eth_dev_data *data; 347 struct virtio_hw *hw; 348 struct virtio_user_dev *dev; 349 350 eth_dev = rte_eth_vdev_allocate(vdev, sizeof(*hw)); 351 if (!eth_dev) { 352 PMD_INIT_LOG(ERR, "cannot alloc rte_eth_dev"); 353 return NULL; 354 } 355 356 data = eth_dev->data; 357 hw = eth_dev->data->dev_private; 358 359 dev = rte_zmalloc(NULL, sizeof(*dev), 0); 360 if (!dev) { 361 PMD_INIT_LOG(ERR, "malloc virtio_user_dev failed"); 362 rte_eth_dev_release_port(eth_dev); 363 rte_free(hw); 364 return NULL; 365 } 366 367 hw->port_id = data->port_id; 368 dev->port_id = data->port_id; 369 virtio_hw_internal[hw->port_id].vtpci_ops = &virtio_user_ops; 370 /* 371 * MSIX is required to enable LSC (see virtio_init_device). 372 * Here just pretend that we support msix. 373 */ 374 hw->use_msix = 1; 375 hw->modern = 0; 376 hw->use_simple_rx = 0; 377 hw->use_simple_tx = 0; 378 hw->virtio_user_dev = dev; 379 data->dev_flags = RTE_ETH_DEV_DETACHABLE; 380 return eth_dev; 381 } 382 383 static void 384 virtio_user_eth_dev_free(struct rte_eth_dev *eth_dev) 385 { 386 struct rte_eth_dev_data *data = eth_dev->data; 387 struct virtio_hw *hw = data->dev_private; 388 389 rte_free(hw->virtio_user_dev); 390 rte_free(hw); 391 rte_eth_dev_release_port(eth_dev); 392 } 393 394 /* Dev initialization routine. Invoked once for each virtio vdev at 395 * EAL init time, see rte_bus_probe(). 396 * Returns 0 on success. 397 */ 398 static int 399 virtio_user_pmd_probe(struct rte_vdev_device *dev) 400 { 401 struct rte_kvargs *kvlist = NULL; 402 struct rte_eth_dev *eth_dev; 403 struct virtio_hw *hw; 404 uint64_t queues = VIRTIO_USER_DEF_Q_NUM; 405 uint64_t cq = VIRTIO_USER_DEF_CQ_EN; 406 uint64_t queue_size = VIRTIO_USER_DEF_Q_SZ; 407 char *path = NULL; 408 char *ifname = NULL; 409 char *mac_addr = NULL; 410 int ret = -1; 411 412 kvlist = rte_kvargs_parse(rte_vdev_device_args(dev), valid_args); 413 if (!kvlist) { 414 PMD_INIT_LOG(ERR, "error when parsing param"); 415 goto end; 416 } 417 418 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_PATH) == 1) { 419 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_PATH, 420 &get_string_arg, &path) < 0) { 421 PMD_INIT_LOG(ERR, "error to parse %s", 422 VIRTIO_USER_ARG_PATH); 423 goto end; 424 } 425 } else { 426 PMD_INIT_LOG(ERR, "arg %s is mandatory for virtio_user", 427 VIRTIO_USER_ARG_QUEUE_SIZE); 428 goto end; 429 } 430 431 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME) == 1) { 432 if (is_vhost_user_by_type(path)) { 433 PMD_INIT_LOG(ERR, 434 "arg %s applies only to vhost-kernel backend", 435 VIRTIO_USER_ARG_INTERFACE_NAME); 436 goto end; 437 } 438 439 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME, 440 &get_string_arg, &ifname) < 0) { 441 PMD_INIT_LOG(ERR, "error to parse %s", 442 VIRTIO_USER_ARG_INTERFACE_NAME); 443 goto end; 444 } 445 } 446 447 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_MAC) == 1) { 448 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_MAC, 449 &get_string_arg, &mac_addr) < 0) { 450 PMD_INIT_LOG(ERR, "error to parse %s", 451 VIRTIO_USER_ARG_MAC); 452 goto end; 453 } 454 } 455 456 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE) == 1) { 457 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE, 458 &get_integer_arg, &queue_size) < 0) { 459 PMD_INIT_LOG(ERR, "error to parse %s", 460 VIRTIO_USER_ARG_QUEUE_SIZE); 461 goto end; 462 } 463 } 464 465 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUES_NUM) == 1) { 466 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUES_NUM, 467 &get_integer_arg, &queues) < 0) { 468 PMD_INIT_LOG(ERR, "error to parse %s", 469 VIRTIO_USER_ARG_QUEUES_NUM); 470 goto end; 471 } 472 } 473 474 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_CQ_NUM) == 1) { 475 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_CQ_NUM, 476 &get_integer_arg, &cq) < 0) { 477 PMD_INIT_LOG(ERR, "error to parse %s", 478 VIRTIO_USER_ARG_CQ_NUM); 479 goto end; 480 } 481 } else if (queues > 1) { 482 cq = 1; 483 } 484 485 if (queues > 1 && cq == 0) { 486 PMD_INIT_LOG(ERR, "multi-q requires ctrl-q"); 487 goto end; 488 } 489 490 if (queues > VIRTIO_MAX_VIRTQUEUE_PAIRS) { 491 PMD_INIT_LOG(ERR, "arg %s %" PRIu64 " exceeds the limit %u", 492 VIRTIO_USER_ARG_QUEUES_NUM, queues, 493 VIRTIO_MAX_VIRTQUEUE_PAIRS); 494 goto end; 495 } 496 497 if (rte_eal_process_type() == RTE_PROC_PRIMARY) { 498 eth_dev = virtio_user_eth_dev_alloc(dev); 499 if (!eth_dev) { 500 PMD_INIT_LOG(ERR, "virtio_user fails to alloc device"); 501 goto end; 502 } 503 504 hw = eth_dev->data->dev_private; 505 if (virtio_user_dev_init(hw->virtio_user_dev, path, queues, cq, 506 queue_size, mac_addr, &ifname) < 0) { 507 PMD_INIT_LOG(ERR, "virtio_user_dev_init fails"); 508 virtio_user_eth_dev_free(eth_dev); 509 goto end; 510 } 511 } else { 512 eth_dev = rte_eth_dev_attach_secondary(rte_vdev_device_name(dev)); 513 if (!eth_dev) 514 goto end; 515 } 516 517 /* previously called by rte_pci_probe() for physical dev */ 518 if (eth_virtio_dev_init(eth_dev) < 0) { 519 PMD_INIT_LOG(ERR, "eth_virtio_dev_init fails"); 520 virtio_user_eth_dev_free(eth_dev); 521 goto end; 522 } 523 ret = 0; 524 525 end: 526 if (kvlist) 527 rte_kvargs_free(kvlist); 528 if (path) 529 free(path); 530 if (mac_addr) 531 free(mac_addr); 532 if (ifname) 533 free(ifname); 534 return ret; 535 } 536 537 /** Called by rte_eth_dev_detach() */ 538 static int 539 virtio_user_pmd_remove(struct rte_vdev_device *vdev) 540 { 541 const char *name; 542 struct rte_eth_dev *eth_dev; 543 struct virtio_hw *hw; 544 struct virtio_user_dev *dev; 545 546 if (!vdev) 547 return -EINVAL; 548 549 name = rte_vdev_device_name(vdev); 550 PMD_DRV_LOG(INFO, "Un-Initializing %s", name); 551 eth_dev = rte_eth_dev_allocated(name); 552 if (!eth_dev) 553 return -ENODEV; 554 555 /* make sure the device is stopped, queues freed */ 556 rte_eth_dev_close(eth_dev->data->port_id); 557 558 hw = eth_dev->data->dev_private; 559 dev = hw->virtio_user_dev; 560 virtio_user_dev_uninit(dev); 561 562 rte_free(eth_dev->data->dev_private); 563 rte_eth_dev_release_port(eth_dev); 564 565 return 0; 566 } 567 568 static struct rte_vdev_driver virtio_user_driver = { 569 .probe = virtio_user_pmd_probe, 570 .remove = virtio_user_pmd_remove, 571 }; 572 573 RTE_PMD_REGISTER_VDEV(net_virtio_user, virtio_user_driver); 574 RTE_PMD_REGISTER_ALIAS(net_virtio_user, virtio_user); 575 RTE_PMD_REGISTER_PARAM_STRING(net_virtio_user, 576 "path=<path> " 577 "mac=<mac addr> " 578 "cq=<int> " 579 "queue_size=<int> " 580 "queues=<int> " 581 "iface=<string>"); 582