4 * Copyright(c) 2010-2014 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>
42 #include <sys/socket.h>
44 #include <rte_ethdev.h>
46 #include <rte_string_fns.h>
47 #include <rte_memory.h>
48 #include <rte_virtio_net.h>
50 #include "vhost-net.h"
51 #include "virtio-net.h"
54 * Device linked list structure for configuration.
56 struct virtio_net_config_ll {
57 struct virtio_net dev; /* Virtio device.*/
58 struct virtio_net_config_ll *next; /* Next dev on linked list.*/
61 /* device ops to add/remove device to/from data core. */
62 struct virtio_net_device_ops const *notify_ops;
63 /* root address of the linked list of managed virtio devices */
64 static struct virtio_net_config_ll *ll_root;
66 /* Features supported by this lib. */
67 #define VHOST_SUPPORTED_FEATURES ((1ULL << VIRTIO_NET_F_MRG_RXBUF) | \
68 (1ULL << VIRTIO_NET_F_CTRL_VQ) | \
69 (1ULL << VIRTIO_NET_F_CTRL_RX) | \
70 (1ULL << VHOST_F_LOG_ALL))
71 static uint64_t VHOST_FEATURES = VHOST_SUPPORTED_FEATURES;
75 * Converts QEMU virtual address to Vhost virtual address. This function is
76 * used to convert the ring addresses to our address space.
79 qva_to_vva(struct virtio_net *dev, uint64_t qemu_va)
81 struct virtio_memory_regions *region;
82 uint64_t vhost_va = 0;
83 uint32_t regionidx = 0;
85 /* Find the region where the address lives. */
86 for (regionidx = 0; regionidx < dev->mem->nregions; regionidx++) {
87 region = &dev->mem->regions[regionidx];
88 if ((qemu_va >= region->userspace_address) &&
89 (qemu_va <= region->userspace_address +
90 region->memory_size)) {
91 vhost_va = qemu_va + region->guest_phys_address +
92 region->address_offset -
93 region->userspace_address;
102 * Retrieves an entry from the devices configuration linked list.
104 static struct virtio_net_config_ll *
105 get_config_ll_entry(struct vhost_device_ctx ctx)
107 struct virtio_net_config_ll *ll_dev = ll_root;
109 /* Loop through linked list until the device_fh is found. */
110 while (ll_dev != NULL) {
111 if (ll_dev->dev.device_fh == ctx.fh)
113 ll_dev = ll_dev->next;
120 * Searches the configuration core linked list and
121 * retrieves the device if it exists.
124 get_device(struct vhost_device_ctx ctx)
126 struct virtio_net_config_ll *ll_dev;
128 ll_dev = get_config_ll_entry(ctx);
133 RTE_LOG(ERR, VHOST_CONFIG,
134 "(%"PRIu64") Device not found in linked list.\n", ctx.fh);
139 * Add entry containing a device to the device configuration linked list.
142 add_config_ll_entry(struct virtio_net_config_ll *new_ll_dev)
144 struct virtio_net_config_ll *ll_dev = ll_root;
146 /* If ll_dev == NULL then this is the first device so go to else */
148 /* If the 1st device_fh != 0 then we insert our device here. */
149 if (ll_dev->dev.device_fh != 0) {
150 new_ll_dev->dev.device_fh = 0;
151 new_ll_dev->next = ll_dev;
152 ll_root = new_ll_dev;
155 * Increment through the ll until we find un unused
156 * device_fh. Insert the device at that entry.
158 while ((ll_dev->next != NULL) &&
159 (ll_dev->dev.device_fh ==
160 (ll_dev->next->dev.device_fh - 1)))
161 ll_dev = ll_dev->next;
163 new_ll_dev->dev.device_fh = ll_dev->dev.device_fh + 1;
164 new_ll_dev->next = ll_dev->next;
165 ll_dev->next = new_ll_dev;
168 ll_root = new_ll_dev;
169 ll_root->dev.device_fh = 0;
175 * Unmap any memory, close any file descriptors and
176 * free any memory owned by a device.
179 cleanup_device(struct virtio_net *dev)
181 /* Unmap QEMU memory file if mapped. */
183 munmap((void *)(uintptr_t)dev->mem->mapped_address,
184 (size_t)dev->mem->mapped_size);
188 /* Close any event notifiers opened by device. */
189 if ((int)dev->virtqueue[VIRTIO_RXQ]->callfd >= 0)
190 close((int)dev->virtqueue[VIRTIO_RXQ]->callfd);
191 if ((int)dev->virtqueue[VIRTIO_RXQ]->kickfd >= 0)
192 close((int)dev->virtqueue[VIRTIO_RXQ]->kickfd);
193 if ((int)dev->virtqueue[VIRTIO_TXQ]->callfd >= 0)
194 close((int)dev->virtqueue[VIRTIO_TXQ]->callfd);
195 if ((int)dev->virtqueue[VIRTIO_TXQ]->kickfd >= 0)
196 close((int)dev->virtqueue[VIRTIO_TXQ]->kickfd);
200 * Release virtqueues and device memory.
203 free_device(struct virtio_net_config_ll *ll_dev)
205 /* Free any malloc'd memory */
206 free(ll_dev->dev.virtqueue[VIRTIO_RXQ]);
207 free(ll_dev->dev.virtqueue[VIRTIO_TXQ]);
212 * Remove an entry from the device configuration linked list.
214 static struct virtio_net_config_ll *
215 rm_config_ll_entry(struct virtio_net_config_ll *ll_dev,
216 struct virtio_net_config_ll *ll_dev_last)
218 /* First remove the device and then clean it up. */
219 if (ll_dev == ll_root) {
220 ll_root = ll_dev->next;
221 cleanup_device(&ll_dev->dev);
225 if (likely(ll_dev_last != NULL)) {
226 ll_dev_last->next = ll_dev->next;
227 cleanup_device(&ll_dev->dev);
229 return ll_dev_last->next;
231 cleanup_device(&ll_dev->dev);
233 RTE_LOG(ERR, VHOST_CONFIG,
234 "Remove entry from config_ll failed\n");
241 * Initialise all variables in device structure.
244 init_device(struct virtio_net *dev)
249 * Virtqueues have already been malloced so
250 * we don't want to set them to NULL.
252 vq_offset = offsetof(struct virtio_net, mem);
254 /* Set everything to 0. */
255 memset((void *)(uintptr_t)((uint64_t)(uintptr_t)dev + vq_offset), 0,
256 (sizeof(struct virtio_net) - (size_t)vq_offset));
257 memset(dev->virtqueue[VIRTIO_RXQ], 0, sizeof(struct vhost_virtqueue));
258 memset(dev->virtqueue[VIRTIO_TXQ], 0, sizeof(struct vhost_virtqueue));
260 dev->virtqueue[VIRTIO_RXQ]->kickfd = (eventfd_t)-1;
261 dev->virtqueue[VIRTIO_RXQ]->callfd = (eventfd_t)-1;
262 dev->virtqueue[VIRTIO_TXQ]->kickfd = (eventfd_t)-1;
263 dev->virtqueue[VIRTIO_TXQ]->callfd = (eventfd_t)-1;
265 /* Backends are set to -1 indicating an inactive device. */
266 dev->virtqueue[VIRTIO_RXQ]->backend = VIRTIO_DEV_STOPPED;
267 dev->virtqueue[VIRTIO_TXQ]->backend = VIRTIO_DEV_STOPPED;
271 * Function is called from the CUSE open function. The device structure is
272 * initialised and a new entry is added to the device configuration linked
276 new_device(struct vhost_device_ctx ctx)
278 struct virtio_net_config_ll *new_ll_dev;
279 struct vhost_virtqueue *virtqueue_rx, *virtqueue_tx;
281 /* Setup device and virtqueues. */
282 new_ll_dev = malloc(sizeof(struct virtio_net_config_ll));
283 if (new_ll_dev == NULL) {
284 RTE_LOG(ERR, VHOST_CONFIG,
285 "(%"PRIu64") Failed to allocate memory for dev.\n",
290 virtqueue_rx = malloc(sizeof(struct vhost_virtqueue));
291 if (virtqueue_rx == NULL) {
293 RTE_LOG(ERR, VHOST_CONFIG,
294 "(%"PRIu64") Failed to allocate memory for rxq.\n",
299 virtqueue_tx = malloc(sizeof(struct vhost_virtqueue));
300 if (virtqueue_tx == NULL) {
303 RTE_LOG(ERR, VHOST_CONFIG,
304 "(%"PRIu64") Failed to allocate memory for txq.\n",
309 new_ll_dev->dev.virtqueue[VIRTIO_RXQ] = virtqueue_rx;
310 new_ll_dev->dev.virtqueue[VIRTIO_TXQ] = virtqueue_tx;
312 /* Initialise device and virtqueues. */
313 init_device(&new_ll_dev->dev);
315 new_ll_dev->next = NULL;
317 /* Add entry to device configuration linked list. */
318 add_config_ll_entry(new_ll_dev);
320 return new_ll_dev->dev.device_fh;
324 * Function is called from the CUSE release function. This function will
325 * cleanup the device and remove it from device configuration linked list.
328 destroy_device(struct vhost_device_ctx ctx)
330 struct virtio_net_config_ll *ll_dev_cur_ctx, *ll_dev_last = NULL;
331 struct virtio_net_config_ll *ll_dev_cur = ll_root;
333 /* Find the linked list entry for the device to be removed. */
334 ll_dev_cur_ctx = get_config_ll_entry(ctx);
335 while (ll_dev_cur != NULL) {
337 * If the device is found or
338 * a device that doesn't exist is found then it is removed.
340 if (ll_dev_cur == ll_dev_cur_ctx) {
342 * If the device is running on a data core then call
343 * the function to remove it from the data core.
345 if ((ll_dev_cur->dev.flags & VIRTIO_DEV_RUNNING))
346 notify_ops->destroy_device(&(ll_dev_cur->dev));
347 ll_dev_cur = rm_config_ll_entry(ll_dev_cur,
350 ll_dev_last = ll_dev_cur;
351 ll_dev_cur = ll_dev_cur->next;
357 set_ifname(struct vhost_device_ctx ctx,
358 const char *if_name, unsigned int if_len)
360 struct virtio_net *dev;
363 dev = get_device(ctx);
367 len = if_len > sizeof(dev->ifname) ?
368 sizeof(dev->ifname) : if_len;
370 strncpy(dev->ifname, if_name, len);
375 * Called from CUSE IOCTL: VHOST_SET_OWNER
376 * This function just returns success at the moment unless
377 * the device hasn't been initialised.
380 set_owner(struct vhost_device_ctx ctx)
382 struct virtio_net *dev;
384 dev = get_device(ctx);
392 * Called from CUSE IOCTL: VHOST_RESET_OWNER
395 reset_owner(struct vhost_device_ctx ctx)
397 struct virtio_net_config_ll *ll_dev;
399 ll_dev = get_config_ll_entry(ctx);
401 cleanup_device(&ll_dev->dev);
402 init_device(&ll_dev->dev);
408 * Called from CUSE IOCTL: VHOST_GET_FEATURES
409 * The features that we support are requested.
412 get_features(struct vhost_device_ctx ctx, uint64_t *pu)
414 struct virtio_net *dev;
416 dev = get_device(ctx);
420 /* Send our supported features. */
421 *pu = VHOST_FEATURES;
426 * Called from CUSE IOCTL: VHOST_SET_FEATURES
427 * We receive the negotiated features supported by us and the virtio device.
430 set_features(struct vhost_device_ctx ctx, uint64_t *pu)
432 struct virtio_net *dev;
434 dev = get_device(ctx);
437 if (*pu & ~VHOST_FEATURES)
440 /* Store the negotiated feature list for the device. */
443 /* Set the vhost_hlen depending on if VIRTIO_NET_F_MRG_RXBUF is set. */
444 if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
445 LOG_DEBUG(VHOST_CONFIG,
446 "(%"PRIu64") Mergeable RX buffers enabled\n",
448 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
449 sizeof(struct virtio_net_hdr_mrg_rxbuf);
450 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
451 sizeof(struct virtio_net_hdr_mrg_rxbuf);
453 LOG_DEBUG(VHOST_CONFIG,
454 "(%"PRIu64") Mergeable RX buffers disabled\n",
456 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
457 sizeof(struct virtio_net_hdr);
458 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
459 sizeof(struct virtio_net_hdr);
465 * Called from CUSE IOCTL: VHOST_SET_VRING_NUM
466 * The virtio device sends us the size of the descriptor ring.
469 set_vring_num(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
471 struct virtio_net *dev;
473 dev = get_device(ctx);
477 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
478 dev->virtqueue[state->index]->size = state->num;
484 * Called from CUSE IOCTL: VHOST_SET_VRING_ADDR
485 * The virtio device sends us the desc, used and avail ring addresses.
486 * This function then converts these to our address space.
489 set_vring_addr(struct vhost_device_ctx ctx, struct vhost_vring_addr *addr)
491 struct virtio_net *dev;
492 struct vhost_virtqueue *vq;
494 dev = get_device(ctx);
498 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
499 vq = dev->virtqueue[addr->index];
501 /* The addresses are converted from QEMU virtual to Vhost virtual. */
502 vq->desc = (struct vring_desc *)(uintptr_t)qva_to_vva(dev,
503 addr->desc_user_addr);
505 RTE_LOG(ERR, VHOST_CONFIG,
506 "(%"PRIu64") Failed to find desc ring address.\n",
511 vq->avail = (struct vring_avail *)(uintptr_t)qva_to_vva(dev,
512 addr->avail_user_addr);
513 if (vq->avail == 0) {
514 RTE_LOG(ERR, VHOST_CONFIG,
515 "(%"PRIu64") Failed to find avail ring address.\n",
520 vq->used = (struct vring_used *)(uintptr_t)qva_to_vva(dev,
521 addr->used_user_addr);
523 RTE_LOG(ERR, VHOST_CONFIG,
524 "(%"PRIu64") Failed to find used ring address.\n",
529 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address desc: %p\n",
530 dev->device_fh, vq->desc);
531 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address avail: %p\n",
532 dev->device_fh, vq->avail);
533 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address used: %p\n",
534 dev->device_fh, vq->used);
540 * Called from CUSE IOCTL: VHOST_SET_VRING_BASE
541 * The virtio device sends us the available ring last used index.
544 set_vring_base(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
546 struct virtio_net *dev;
548 dev = get_device(ctx);
552 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
553 dev->virtqueue[state->index]->last_used_idx = state->num;
554 dev->virtqueue[state->index]->last_used_idx_res = state->num;
560 * Called from CUSE IOCTL: VHOST_GET_VRING_BASE
561 * We send the virtio device our available ring last used index.
564 get_vring_base(struct vhost_device_ctx ctx, uint32_t index,
565 struct vhost_vring_state *state)
567 struct virtio_net *dev;
569 dev = get_device(ctx);
573 state->index = index;
574 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
575 state->num = dev->virtqueue[state->index]->last_used_idx;
582 * Called from CUSE IOCTL: VHOST_SET_VRING_CALL
583 * The virtio device sends an eventfd to interrupt the guest. This fd gets
584 * copied into our process space.
587 set_vring_call(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
589 struct virtio_net *dev;
590 struct vhost_virtqueue *vq;
592 dev = get_device(ctx);
596 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
597 vq = dev->virtqueue[file->index];
599 if ((int)vq->callfd >= 0)
600 close((int)vq->callfd);
602 vq->callfd = file->fd;
608 * Called from CUSE IOCTL: VHOST_SET_VRING_KICK
609 * The virtio device sends an eventfd that it can use to notify us.
610 * This fd gets copied into our process space.
613 set_vring_kick(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
615 struct virtio_net *dev;
616 struct vhost_virtqueue *vq;
618 dev = get_device(ctx);
622 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
623 vq = dev->virtqueue[file->index];
625 if ((int)vq->kickfd >= 0)
626 close((int)vq->kickfd);
628 vq->kickfd = file->fd;
634 * Called from CUSE IOCTL: VHOST_NET_SET_BACKEND
635 * To complete device initialisation when the virtio driver is loaded,
636 * we are provided with a valid fd for a tap device (not used by us).
637 * If this happens then we can add the device to a data core.
638 * When the virtio driver is removed we get fd=-1.
639 * At that point we remove the device from the data core.
640 * The device will still exist in the device configuration linked list.
643 set_backend(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
645 struct virtio_net *dev;
647 dev = get_device(ctx);
651 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
652 dev->virtqueue[file->index]->backend = file->fd;
655 * If the device isn't already running and both backend fds are set,
658 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
659 if (((int)dev->virtqueue[VIRTIO_TXQ]->backend != VIRTIO_DEV_STOPPED) &&
660 ((int)dev->virtqueue[VIRTIO_RXQ]->backend != VIRTIO_DEV_STOPPED)) {
661 return notify_ops->new_device(dev);
663 /* Otherwise we remove it. */
665 if (file->fd == VIRTIO_DEV_STOPPED)
666 notify_ops->destroy_device(dev);
671 * Function pointers are set for the device operations to allow CUSE to call
672 * functions when an IOCTL, device_add or device_release is received.
674 static const struct vhost_net_device_ops vhost_device_ops = {
675 .new_device = new_device,
676 .destroy_device = destroy_device,
678 .set_ifname = set_ifname,
680 .get_features = get_features,
681 .set_features = set_features,
683 .set_vring_num = set_vring_num,
684 .set_vring_addr = set_vring_addr,
685 .set_vring_base = set_vring_base,
686 .get_vring_base = get_vring_base,
688 .set_vring_kick = set_vring_kick,
689 .set_vring_call = set_vring_call,
691 .set_backend = set_backend,
693 .set_owner = set_owner,
694 .reset_owner = reset_owner,
698 * Called by main to setup callbacks when registering CUSE device.
700 struct vhost_net_device_ops const *
701 get_virtio_net_callbacks(void)
703 return &vhost_device_ops;
706 int rte_vhost_enable_guest_notification(struct virtio_net *dev,
707 uint16_t queue_id, int enable)
710 RTE_LOG(ERR, VHOST_CONFIG,
711 "guest notification isn't supported.\n");
715 dev->virtqueue[queue_id]->used->flags =
716 enable ? 0 : VRING_USED_F_NO_NOTIFY;
720 uint64_t rte_vhost_feature_get(void)
722 return VHOST_FEATURES;
725 int rte_vhost_feature_disable(uint64_t feature_mask)
727 VHOST_FEATURES = VHOST_FEATURES & ~feature_mask;
731 int rte_vhost_feature_enable(uint64_t feature_mask)
733 if ((feature_mask & VHOST_SUPPORTED_FEATURES) == feature_mask) {
734 VHOST_FEATURES = VHOST_FEATURES | feature_mask;
741 * Register ops so that we can add/remove device to data core.
744 rte_vhost_driver_callback_register(struct virtio_net_device_ops const * const ops)