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.
34 #include <linux/vhost.h>
35 #include <linux/virtio_net.h>
39 #ifdef RTE_LIBRTE_VHOST_NUMA
43 #include <rte_errno.h>
44 #include <rte_ethdev.h>
46 #include <rte_string_fns.h>
47 #include <rte_memory.h>
48 #include <rte_malloc.h>
49 #include <rte_vhost.h>
50 #include <rte_rwlock.h>
54 #include "vhost_user.h"
56 struct virtio_net *vhost_devices[MAX_VHOST_DEVICE];
59 __vhost_iova_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
60 uint64_t iova, uint64_t size, uint8_t perm)
62 uint64_t vva, tmp_size;
69 vva = vhost_user_iotlb_cache_find(vq, iova, &tmp_size, perm);
73 if (!vhost_user_iotlb_pending_miss(vq, iova + tmp_size, perm)) {
74 vhost_user_iotlb_pending_insert(vq, iova + tmp_size, perm);
75 vhost_user_iotlb_miss(dev, iova + tmp_size, perm);
84 struct virtio_net *dev = vhost_devices[vid];
87 RTE_LOG(ERR, VHOST_CONFIG,
88 "(%d) device not found.\n", vid);
95 cleanup_vq(struct vhost_virtqueue *vq, int destroy)
97 if ((vq->callfd >= 0) && (destroy != 0))
104 * Unmap any memory, close any file descriptors and
105 * free any memory owned by a device.
108 cleanup_device(struct virtio_net *dev, int destroy)
112 vhost_backend_cleanup(dev);
114 for (i = 0; i < dev->nr_vring; i++)
115 cleanup_vq(dev->virtqueue[i], destroy);
119 * Release virtqueues and device memory.
122 free_device(struct virtio_net *dev)
125 struct vhost_virtqueue *vq;
127 for (i = 0; i < dev->nr_vring; i++) {
128 vq = dev->virtqueue[i];
130 rte_free(vq->shadow_used_ring);
131 rte_free(vq->batch_copy_elems);
132 rte_mempool_free(vq->iotlb_pool);
140 vring_translate(struct virtio_net *dev, struct vhost_virtqueue *vq)
144 if (!(dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)))
147 size = sizeof(struct vring_desc) * vq->size;
148 vq->desc = (struct vring_desc *)(uintptr_t)vhost_iova_to_vva(dev, vq,
149 vq->ring_addrs.desc_user_addr,
150 size, VHOST_ACCESS_RW);
154 size = sizeof(struct vring_avail);
155 size += sizeof(uint16_t) * vq->size;
156 vq->avail = (struct vring_avail *)(uintptr_t)vhost_iova_to_vva(dev, vq,
157 vq->ring_addrs.avail_user_addr,
158 size, VHOST_ACCESS_RW);
162 size = sizeof(struct vring_used);
163 size += sizeof(struct vring_used_elem) * vq->size;
164 vq->used = (struct vring_used *)(uintptr_t)vhost_iova_to_vva(dev, vq,
165 vq->ring_addrs.used_user_addr,
166 size, VHOST_ACCESS_RW);
177 vring_invalidate(struct virtio_net *dev, struct vhost_virtqueue *vq)
179 if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
180 vhost_user_iotlb_wr_lock(vq);
187 if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
188 vhost_user_iotlb_wr_unlock(vq);
192 init_vring_queue(struct virtio_net *dev, uint32_t vring_idx)
194 struct vhost_virtqueue *vq;
196 if (vring_idx >= VHOST_MAX_VRING) {
197 RTE_LOG(ERR, VHOST_CONFIG,
198 "Failed not init vring, out of bound (%d)\n",
203 vq = dev->virtqueue[vring_idx];
205 memset(vq, 0, sizeof(struct vhost_virtqueue));
207 vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
208 vq->callfd = VIRTIO_UNINITIALIZED_EVENTFD;
210 vhost_user_iotlb_init(dev, vring_idx);
211 /* Backends are set to -1 indicating an inactive device. */
214 TAILQ_INIT(&vq->zmbuf_list);
218 reset_vring_queue(struct virtio_net *dev, uint32_t vring_idx)
220 struct vhost_virtqueue *vq;
223 if (vring_idx >= VHOST_MAX_VRING) {
224 RTE_LOG(ERR, VHOST_CONFIG,
225 "Failed not init vring, out of bound (%d)\n",
230 vq = dev->virtqueue[vring_idx];
232 init_vring_queue(dev, vring_idx);
237 alloc_vring_queue(struct virtio_net *dev, uint32_t vring_idx)
239 struct vhost_virtqueue *vq;
241 vq = rte_malloc(NULL, sizeof(struct vhost_virtqueue), 0);
243 RTE_LOG(ERR, VHOST_CONFIG,
244 "Failed to allocate memory for vring:%u.\n", vring_idx);
248 dev->virtqueue[vring_idx] = vq;
249 init_vring_queue(dev, vring_idx);
257 * Reset some variables in device structure, while keeping few
258 * others untouched, such as vid, ifname, nr_vring: they
259 * should be same unless the device is removed.
262 reset_device(struct virtio_net *dev)
267 dev->protocol_features = 0;
270 for (i = 0; i < dev->nr_vring; i++)
271 reset_vring_queue(dev, i);
275 * Invoked when there is a new vhost-user connection established (when
276 * there is a new virtio device being attached).
279 vhost_new_device(void)
281 struct virtio_net *dev;
284 dev = rte_zmalloc(NULL, sizeof(struct virtio_net), 0);
286 RTE_LOG(ERR, VHOST_CONFIG,
287 "Failed to allocate memory for new dev.\n");
291 for (i = 0; i < MAX_VHOST_DEVICE; i++) {
292 if (vhost_devices[i] == NULL)
295 if (i == MAX_VHOST_DEVICE) {
296 RTE_LOG(ERR, VHOST_CONFIG,
297 "Failed to find a free slot for new device.\n");
302 vhost_devices[i] = dev;
304 dev->slave_req_fd = -1;
310 * Invoked when there is the vhost-user connection is broken (when
311 * the virtio device is being detached).
314 vhost_destroy_device(int vid)
316 struct virtio_net *dev = get_device(vid);
321 if (dev->flags & VIRTIO_DEV_RUNNING) {
322 dev->flags &= ~VIRTIO_DEV_RUNNING;
323 dev->notify_ops->destroy_device(vid);
326 cleanup_device(dev, 1);
329 vhost_devices[vid] = NULL;
333 vhost_set_ifname(int vid, const char *if_name, unsigned int if_len)
335 struct virtio_net *dev;
338 dev = get_device(vid);
342 len = if_len > sizeof(dev->ifname) ?
343 sizeof(dev->ifname) : if_len;
345 strncpy(dev->ifname, if_name, len);
346 dev->ifname[sizeof(dev->ifname) - 1] = '\0';
350 vhost_enable_dequeue_zero_copy(int vid)
352 struct virtio_net *dev = get_device(vid);
357 dev->dequeue_zero_copy = 1;
361 rte_vhost_get_mtu(int vid, uint16_t *mtu)
363 struct virtio_net *dev = get_device(vid);
368 if (!(dev->flags & VIRTIO_DEV_READY))
371 if (!(dev->features & (1ULL << VIRTIO_NET_F_MTU)))
380 rte_vhost_get_numa_node(int vid)
382 #ifdef RTE_LIBRTE_VHOST_NUMA
383 struct virtio_net *dev = get_device(vid);
390 ret = get_mempolicy(&numa_node, NULL, 0, dev,
391 MPOL_F_NODE | MPOL_F_ADDR);
393 RTE_LOG(ERR, VHOST_CONFIG,
394 "(%d) failed to query numa node: %s\n",
395 vid, rte_strerror(errno));
407 rte_vhost_get_queue_num(int vid)
409 struct virtio_net *dev = get_device(vid);
414 return dev->nr_vring / 2;
418 rte_vhost_get_vring_num(int vid)
420 struct virtio_net *dev = get_device(vid);
425 return dev->nr_vring;
429 rte_vhost_get_ifname(int vid, char *buf, size_t len)
431 struct virtio_net *dev = get_device(vid);
436 len = RTE_MIN(len, sizeof(dev->ifname));
438 strncpy(buf, dev->ifname, len);
445 rte_vhost_get_negotiated_features(int vid, uint64_t *features)
447 struct virtio_net *dev;
449 dev = get_device(vid);
453 *features = dev->features;
458 rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem)
460 struct virtio_net *dev;
461 struct rte_vhost_memory *m;
464 dev = get_device(vid);
468 size = dev->mem->nregions * sizeof(struct rte_vhost_mem_region);
469 m = malloc(sizeof(struct rte_vhost_memory) + size);
473 m->nregions = dev->mem->nregions;
474 memcpy(m->regions, dev->mem->regions, size);
481 rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
482 struct rte_vhost_vring *vring)
484 struct virtio_net *dev;
485 struct vhost_virtqueue *vq;
487 dev = get_device(vid);
491 if (vring_idx >= VHOST_MAX_VRING)
494 vq = dev->virtqueue[vring_idx];
498 vring->desc = vq->desc;
499 vring->avail = vq->avail;
500 vring->used = vq->used;
501 vring->log_guest_addr = vq->log_guest_addr;
503 vring->callfd = vq->callfd;
504 vring->kickfd = vq->kickfd;
505 vring->size = vq->size;
511 rte_vhost_avail_entries(int vid, uint16_t queue_id)
513 struct virtio_net *dev;
514 struct vhost_virtqueue *vq;
516 dev = get_device(vid);
520 vq = dev->virtqueue[queue_id];
524 return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx;
528 rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
530 struct virtio_net *dev = get_device(vid);
536 RTE_LOG(ERR, VHOST_CONFIG,
537 "guest notification isn't supported.\n");
541 dev->virtqueue[queue_id]->used->flags = VRING_USED_F_NO_NOTIFY;
546 rte_vhost_log_write(int vid, uint64_t addr, uint64_t len)
548 struct virtio_net *dev = get_device(vid);
553 vhost_log_write(dev, addr, len);
557 rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
558 uint64_t offset, uint64_t len)
560 struct virtio_net *dev;
561 struct vhost_virtqueue *vq;
563 dev = get_device(vid);
567 if (vring_idx >= VHOST_MAX_VRING)
569 vq = dev->virtqueue[vring_idx];
573 vhost_log_used_vring(dev, vq, offset, len);
577 rte_vhost_rx_queue_count(int vid, uint16_t qid)
579 struct virtio_net *dev;
580 struct vhost_virtqueue *vq;
582 dev = get_device(vid);
586 if (unlikely(qid >= dev->nr_vring || (qid & 1) == 0)) {
587 RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n",
588 dev->vid, __func__, qid);
592 vq = dev->virtqueue[qid];
596 if (unlikely(vq->enabled == 0 || vq->avail == NULL))
599 return *((volatile uint16_t *)&vq->avail->idx) - vq->last_avail_idx;