4 * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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.
41 #include <sys/eventfd.h>
42 #include <sys/types.h>
46 #include "virtio_user_dev.h"
47 #include "../virtio_ethdev.h"
50 virtio_user_create_queue(struct virtio_user_dev *dev, uint32_t queue_sel)
52 /* Of all per virtqueue MSGs, make sure VHOST_SET_VRING_CALL come
53 * firstly because vhost depends on this msg to allocate virtqueue
56 struct vhost_vring_file file;
58 file.index = queue_sel;
59 file.fd = dev->callfds[queue_sel];
60 dev->ops->send_request(dev, VHOST_USER_SET_VRING_CALL, &file);
66 virtio_user_kick_queue(struct virtio_user_dev *dev, uint32_t queue_sel)
68 struct vhost_vring_file file;
69 struct vhost_vring_state state;
70 struct vring *vring = &dev->vrings[queue_sel];
71 struct vhost_vring_addr addr = {
73 .desc_user_addr = (uint64_t)(uintptr_t)vring->desc,
74 .avail_user_addr = (uint64_t)(uintptr_t)vring->avail,
75 .used_user_addr = (uint64_t)(uintptr_t)vring->used,
77 .flags = 0, /* disable log */
80 state.index = queue_sel;
81 state.num = vring->num;
82 dev->ops->send_request(dev, VHOST_USER_SET_VRING_NUM, &state);
84 state.index = queue_sel;
85 state.num = 0; /* no reservation */
86 dev->ops->send_request(dev, VHOST_USER_SET_VRING_BASE, &state);
88 dev->ops->send_request(dev, VHOST_USER_SET_VRING_ADDR, &addr);
90 /* Of all per virtqueue MSGs, make sure VHOST_USER_SET_VRING_KICK comes
91 * lastly because vhost depends on this msg to judge if
94 file.index = queue_sel;
95 file.fd = dev->kickfds[queue_sel];
96 dev->ops->send_request(dev, VHOST_USER_SET_VRING_KICK, &file);
102 virtio_user_queue_setup(struct virtio_user_dev *dev,
103 int (*fn)(struct virtio_user_dev *, uint32_t))
105 uint32_t i, queue_sel;
107 for (i = 0; i < dev->max_queue_pairs; ++i) {
108 queue_sel = 2 * i + VTNET_SQ_RQ_QUEUE_IDX;
109 if (fn(dev, queue_sel) < 0) {
110 PMD_DRV_LOG(INFO, "setup rx vq fails: %u", i);
114 for (i = 0; i < dev->max_queue_pairs; ++i) {
115 queue_sel = 2 * i + VTNET_SQ_TQ_QUEUE_IDX;
116 if (fn(dev, queue_sel) < 0) {
117 PMD_DRV_LOG(INFO, "setup tx vq fails: %u", i);
126 virtio_user_start_device(struct virtio_user_dev *dev)
131 /* Step 0: tell vhost to create queues */
132 if (virtio_user_queue_setup(dev, virtio_user_create_queue) < 0)
135 /* Step 1: set features */
136 features = dev->features;
137 /* Strip VIRTIO_NET_F_MAC, as MAC address is handled in vdev init */
138 features &= ~(1ull << VIRTIO_NET_F_MAC);
139 /* Strip VIRTIO_NET_F_CTRL_VQ, as devices do not really need to know */
140 features &= ~(1ull << VIRTIO_NET_F_CTRL_VQ);
141 features &= ~(1ull << VIRTIO_NET_F_STATUS);
142 ret = dev->ops->send_request(dev, VHOST_USER_SET_FEATURES, &features);
145 PMD_DRV_LOG(INFO, "set features: %" PRIx64, features);
147 /* Step 2: share memory regions */
148 ret = dev->ops->send_request(dev, VHOST_USER_SET_MEM_TABLE, NULL);
152 /* Step 3: kick queues */
153 if (virtio_user_queue_setup(dev, virtio_user_kick_queue) < 0)
156 /* Step 4: enable queues
157 * we enable the 1st queue pair by default.
159 dev->ops->enable_qp(dev, 0, 1);
163 /* TODO: free resource here or caller to check */
167 int virtio_user_stop_device(struct virtio_user_dev *dev)
171 for (i = 0; i < dev->max_queue_pairs; ++i)
172 dev->ops->enable_qp(dev, i, 0);
178 parse_mac(struct virtio_user_dev *dev, const char *mac)
181 uint32_t tmp[ETHER_ADDR_LEN];
186 r = sscanf(mac, "%x:%x:%x:%x:%x:%x", &tmp[0],
187 &tmp[1], &tmp[2], &tmp[3], &tmp[4], &tmp[5]);
188 if (r == ETHER_ADDR_LEN) {
189 for (i = 0; i < ETHER_ADDR_LEN; ++i)
190 dev->mac_addr[i] = (uint8_t)tmp[i];
191 dev->mac_specified = 1;
193 /* ignore the wrong mac, use random mac */
194 PMD_DRV_LOG(ERR, "wrong format of mac: %s", mac);
199 is_vhost_user_by_type(const char *path)
203 if (stat(path, &sb) == -1)
206 return S_ISSOCK(sb.st_mode);
210 virtio_user_dev_init_notify(struct virtio_user_dev *dev)
216 for (i = 0; i < VIRTIO_MAX_VIRTQUEUES; ++i) {
217 if (i >= dev->max_queue_pairs * 2) {
218 dev->kickfds[i] = -1;
219 dev->callfds[i] = -1;
223 /* May use invalid flag, but some backend uses kickfd and
224 * callfd as criteria to judge if dev is alive. so finally we
227 callfd = eventfd(0, EFD_CLOEXEC | EFD_NONBLOCK);
229 PMD_DRV_LOG(ERR, "callfd error, %s", strerror(errno));
232 kickfd = eventfd(0, EFD_CLOEXEC | EFD_NONBLOCK);
234 PMD_DRV_LOG(ERR, "kickfd error, %s", strerror(errno));
237 dev->callfds[i] = callfd;
238 dev->kickfds[i] = kickfd;
241 if (i < VIRTIO_MAX_VIRTQUEUES) {
242 for (j = 0; j <= i; ++j) {
243 close(dev->callfds[j]);
244 close(dev->kickfds[j]);
254 virtio_user_fill_intr_handle(struct virtio_user_dev *dev)
257 struct rte_eth_dev *eth_dev = &rte_eth_devices[dev->port_id];
259 if (!eth_dev->intr_handle) {
260 eth_dev->intr_handle = malloc(sizeof(*eth_dev->intr_handle));
261 if (!eth_dev->intr_handle) {
262 PMD_DRV_LOG(ERR, "fail to allocate intr_handle");
265 memset(eth_dev->intr_handle, 0, sizeof(*eth_dev->intr_handle));
268 for (i = 0; i < dev->max_queue_pairs; ++i)
269 eth_dev->intr_handle->efds[i] = dev->callfds[i];
270 eth_dev->intr_handle->nb_efd = dev->max_queue_pairs;
271 eth_dev->intr_handle->max_intr = dev->max_queue_pairs + 1;
272 eth_dev->intr_handle->type = RTE_INTR_HANDLE_VDEV;
273 if (dev->vhostfd >= 0)
274 eth_dev->intr_handle->fd = dev->vhostfd;
280 virtio_user_dev_setup(struct virtio_user_dev *dev)
285 dev->vhostfds = NULL;
288 if (is_vhost_user_by_type(dev->path)) {
289 dev->ops = &ops_user;
291 dev->ops = &ops_kernel;
293 dev->vhostfds = malloc(dev->max_queue_pairs * sizeof(int));
294 dev->tapfds = malloc(dev->max_queue_pairs * sizeof(int));
295 if (!dev->vhostfds || !dev->tapfds) {
296 PMD_INIT_LOG(ERR, "Failed to malloc");
300 for (q = 0; q < dev->max_queue_pairs; ++q) {
301 dev->vhostfds[q] = -1;
306 if (dev->ops->setup(dev) < 0)
309 if (virtio_user_dev_init_notify(dev) < 0)
312 if (virtio_user_fill_intr_handle(dev) < 0)
318 /* Use below macro to filter features from vhost backend */
319 #define VIRTIO_USER_SUPPORTED_FEATURES \
320 (1ULL << VIRTIO_NET_F_MAC | \
321 1ULL << VIRTIO_NET_F_STATUS | \
322 1ULL << VIRTIO_NET_F_MQ | \
323 1ULL << VIRTIO_NET_F_CTRL_MAC_ADDR | \
324 1ULL << VIRTIO_NET_F_CTRL_VQ | \
325 1ULL << VIRTIO_NET_F_CTRL_RX | \
326 1ULL << VIRTIO_NET_F_CTRL_VLAN | \
327 1ULL << VIRTIO_NET_F_CSUM | \
328 1ULL << VIRTIO_NET_F_HOST_TSO4 | \
329 1ULL << VIRTIO_NET_F_HOST_TSO6 | \
330 1ULL << VIRTIO_NET_F_MRG_RXBUF | \
331 1ULL << VIRTIO_RING_F_INDIRECT_DESC | \
332 1ULL << VIRTIO_NET_F_GUEST_CSUM | \
333 1ULL << VIRTIO_NET_F_GUEST_TSO4 | \
334 1ULL << VIRTIO_NET_F_GUEST_TSO6 | \
335 1ULL << VIRTIO_F_VERSION_1)
338 virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
339 int cq, int queue_size, const char *mac, char **ifname)
341 snprintf(dev->path, PATH_MAX, "%s", path);
342 dev->max_queue_pairs = queues;
343 dev->queue_pairs = 1; /* mq disabled by default */
344 dev->queue_size = queue_size;
345 dev->mac_specified = 0;
349 dev->ifname = *ifname;
353 if (virtio_user_dev_setup(dev) < 0) {
354 PMD_INIT_LOG(ERR, "backend set up fails");
357 if (dev->ops->send_request(dev, VHOST_USER_SET_OWNER, NULL) < 0) {
358 PMD_INIT_LOG(ERR, "set_owner fails: %s", strerror(errno));
362 if (dev->ops->send_request(dev, VHOST_USER_GET_FEATURES,
363 &dev->device_features) < 0) {
364 PMD_INIT_LOG(ERR, "get_features failed: %s", strerror(errno));
367 if (dev->mac_specified)
368 dev->device_features |= (1ull << VIRTIO_NET_F_MAC);
371 /* device does not really need to know anything about CQ,
372 * so if necessary, we just claim to support CQ
374 dev->device_features |= (1ull << VIRTIO_NET_F_CTRL_VQ);
376 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_VQ);
377 /* Also disable features depends on VIRTIO_NET_F_CTRL_VQ */
378 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_RX);
379 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_VLAN);
380 dev->device_features &= ~(1ull << VIRTIO_NET_F_GUEST_ANNOUNCE);
381 dev->device_features &= ~(1ull << VIRTIO_NET_F_MQ);
382 dev->device_features &= ~(1ull << VIRTIO_NET_F_CTRL_MAC_ADDR);
385 /* The backend will not report this feature, we add it explicitly */
386 if (is_vhost_user_by_type(dev->path))
387 dev->device_features |= (1ull << VIRTIO_NET_F_STATUS);
389 dev->device_features &= VIRTIO_USER_SUPPORTED_FEATURES;
395 virtio_user_dev_uninit(struct virtio_user_dev *dev)
399 virtio_user_stop_device(dev);
401 for (i = 0; i < dev->max_queue_pairs * 2; ++i) {
402 close(dev->callfds[i]);
403 close(dev->kickfds[i]);
409 for (i = 0; i < dev->max_queue_pairs; ++i)
410 close(dev->vhostfds[i]);
419 virtio_user_handle_mq(struct virtio_user_dev *dev, uint16_t q_pairs)
424 if (q_pairs > dev->max_queue_pairs) {
425 PMD_INIT_LOG(ERR, "multi-q config %u, but only %u supported",
426 q_pairs, dev->max_queue_pairs);
430 for (i = 0; i < q_pairs; ++i)
431 ret |= dev->ops->enable_qp(dev, i, 1);
432 for (i = q_pairs; i < dev->max_queue_pairs; ++i)
433 ret |= dev->ops->enable_qp(dev, i, 0);
435 dev->queue_pairs = q_pairs;
441 virtio_user_handle_ctrl_msg(struct virtio_user_dev *dev, struct vring *vring,
444 struct virtio_net_ctrl_hdr *hdr;
445 virtio_net_ctrl_ack status = ~0;
446 uint16_t i, idx_data, idx_status;
447 uint32_t n_descs = 0;
449 /* locate desc for header, data, and status */
450 idx_data = vring->desc[idx_hdr].next;
454 while (vring->desc[i].flags == VRING_DESC_F_NEXT) {
455 i = vring->desc[i].next;
459 /* locate desc for status */
463 hdr = (void *)(uintptr_t)vring->desc[idx_hdr].addr;
464 if (hdr->class == VIRTIO_NET_CTRL_MQ &&
465 hdr->cmd == VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET) {
468 queues = *(uint16_t *)(uintptr_t)vring->desc[idx_data].addr;
469 status = virtio_user_handle_mq(dev, queues);
473 *(virtio_net_ctrl_ack *)(uintptr_t)vring->desc[idx_status].addr = status;
479 virtio_user_handle_cq(struct virtio_user_dev *dev, uint16_t queue_idx)
481 uint16_t avail_idx, desc_idx;
482 struct vring_used_elem *uep;
484 struct vring *vring = &dev->vrings[queue_idx];
486 /* Consume avail ring, using used ring idx as first one */
487 while (vring->used->idx != vring->avail->idx) {
488 avail_idx = (vring->used->idx) & (vring->num - 1);
489 desc_idx = vring->avail->ring[avail_idx];
491 n_descs = virtio_user_handle_ctrl_msg(dev, vring, desc_idx);
493 /* Update used ring */
494 uep = &vring->used->ring[avail_idx];