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>
41 #include <sys/ioctl.h>
44 #include <sys/socket.h>
45 #include <linux/if_tun.h>
48 #include <rte_ethdev.h>
50 #include <rte_string_fns.h>
51 #include <rte_memory.h>
52 #include <rte_virtio_net.h>
54 #include "vhost_cuse/eventfd_copy.h"
55 #include "vhost-net.h"
56 #include "virtio-net.h"
59 * Device linked list structure for configuration.
61 struct virtio_net_config_ll {
62 struct virtio_net dev; /* Virtio device.*/
63 struct virtio_net_config_ll *next; /* Next dev on linked list.*/
66 /* device ops to add/remove device to/from data core. */
67 struct virtio_net_device_ops const *notify_ops;
68 /* root address of the linked list of managed virtio devices */
69 static struct virtio_net_config_ll *ll_root;
71 /* Features supported by this lib. */
72 #define VHOST_SUPPORTED_FEATURES ((1ULL << VIRTIO_NET_F_MRG_RXBUF) | \
73 (1ULL << VIRTIO_NET_F_CTRL_VQ) | \
74 (1ULL << VIRTIO_NET_F_CTRL_RX))
75 static uint64_t VHOST_FEATURES = VHOST_SUPPORTED_FEATURES;
79 * Converts QEMU virtual address to Vhost virtual address. This function is
80 * used to convert the ring addresses to our address space.
83 qva_to_vva(struct virtio_net *dev, uint64_t qemu_va)
85 struct virtio_memory_regions *region;
86 uint64_t vhost_va = 0;
87 uint32_t regionidx = 0;
89 /* Find the region where the address lives. */
90 for (regionidx = 0; regionidx < dev->mem->nregions; regionidx++) {
91 region = &dev->mem->regions[regionidx];
92 if ((qemu_va >= region->userspace_address) &&
93 (qemu_va <= region->userspace_address +
94 region->memory_size)) {
95 vhost_va = qemu_va + region->guest_phys_address +
96 region->address_offset -
97 region->userspace_address;
106 * Retrieves an entry from the devices configuration linked list.
108 static struct virtio_net_config_ll *
109 get_config_ll_entry(struct vhost_device_ctx ctx)
111 struct virtio_net_config_ll *ll_dev = ll_root;
113 /* Loop through linked list until the device_fh is found. */
114 while (ll_dev != NULL) {
115 if (ll_dev->dev.device_fh == ctx.fh)
117 ll_dev = ll_dev->next;
124 * Searches the configuration core linked list and
125 * retrieves the device if it exists.
128 get_device(struct vhost_device_ctx ctx)
130 struct virtio_net_config_ll *ll_dev;
132 ll_dev = get_config_ll_entry(ctx);
137 RTE_LOG(ERR, VHOST_CONFIG,
138 "(%"PRIu64") Device not found in linked list.\n", ctx.fh);
143 * Add entry containing a device to the device configuration linked list.
146 add_config_ll_entry(struct virtio_net_config_ll *new_ll_dev)
148 struct virtio_net_config_ll *ll_dev = ll_root;
150 /* If ll_dev == NULL then this is the first device so go to else */
152 /* If the 1st device_fh != 0 then we insert our device here. */
153 if (ll_dev->dev.device_fh != 0) {
154 new_ll_dev->dev.device_fh = 0;
155 new_ll_dev->next = ll_dev;
156 ll_root = new_ll_dev;
159 * Increment through the ll until we find un unused
160 * device_fh. Insert the device at that entry.
162 while ((ll_dev->next != NULL) &&
163 (ll_dev->dev.device_fh ==
164 (ll_dev->next->dev.device_fh - 1)))
165 ll_dev = ll_dev->next;
167 new_ll_dev->dev.device_fh = ll_dev->dev.device_fh + 1;
168 new_ll_dev->next = ll_dev->next;
169 ll_dev->next = new_ll_dev;
172 ll_root = new_ll_dev;
173 ll_root->dev.device_fh = 0;
179 * Unmap any memory, close any file descriptors and
180 * free any memory owned by a device.
183 cleanup_device(struct virtio_net *dev)
185 /* Unmap QEMU memory file if mapped. */
187 munmap((void *)(uintptr_t)dev->mem->mapped_address,
188 (size_t)dev->mem->mapped_size);
192 /* Close any event notifiers opened by device. */
193 if (dev->virtqueue[VIRTIO_RXQ]->callfd)
194 close((int)dev->virtqueue[VIRTIO_RXQ]->callfd);
195 if (dev->virtqueue[VIRTIO_RXQ]->kickfd)
196 close((int)dev->virtqueue[VIRTIO_RXQ]->kickfd);
197 if (dev->virtqueue[VIRTIO_TXQ]->callfd)
198 close((int)dev->virtqueue[VIRTIO_TXQ]->callfd);
199 if (dev->virtqueue[VIRTIO_TXQ]->kickfd)
200 close((int)dev->virtqueue[VIRTIO_TXQ]->kickfd);
204 * Release virtqueues and device memory.
207 free_device(struct virtio_net_config_ll *ll_dev)
209 /* Free any malloc'd memory */
210 free(ll_dev->dev.virtqueue[VIRTIO_RXQ]);
211 free(ll_dev->dev.virtqueue[VIRTIO_TXQ]);
216 * Remove an entry from the device configuration linked list.
218 static struct virtio_net_config_ll *
219 rm_config_ll_entry(struct virtio_net_config_ll *ll_dev,
220 struct virtio_net_config_ll *ll_dev_last)
222 /* First remove the device and then clean it up. */
223 if (ll_dev == ll_root) {
224 ll_root = ll_dev->next;
225 cleanup_device(&ll_dev->dev);
229 if (likely(ll_dev_last != NULL)) {
230 ll_dev_last->next = ll_dev->next;
231 cleanup_device(&ll_dev->dev);
233 return ll_dev_last->next;
235 cleanup_device(&ll_dev->dev);
237 RTE_LOG(ERR, VHOST_CONFIG,
238 "Remove entry from config_ll failed\n");
245 * Initialise all variables in device structure.
248 init_device(struct virtio_net *dev)
253 * Virtqueues have already been malloced so
254 * we don't want to set them to NULL.
256 vq_offset = offsetof(struct virtio_net, mem);
258 /* Set everything to 0. */
259 memset((void *)(uintptr_t)((uint64_t)(uintptr_t)dev + vq_offset), 0,
260 (sizeof(struct virtio_net) - (size_t)vq_offset));
261 memset(dev->virtqueue[VIRTIO_RXQ], 0, sizeof(struct vhost_virtqueue));
262 memset(dev->virtqueue[VIRTIO_TXQ], 0, sizeof(struct vhost_virtqueue));
264 dev->virtqueue[VIRTIO_RXQ]->kickfd = (eventfd_t)-1;
265 dev->virtqueue[VIRTIO_RXQ]->callfd = (eventfd_t)-1;
266 dev->virtqueue[VIRTIO_TXQ]->kickfd = (eventfd_t)-1;
267 dev->virtqueue[VIRTIO_TXQ]->callfd = (eventfd_t)-1;
269 /* Backends are set to -1 indicating an inactive device. */
270 dev->virtqueue[VIRTIO_RXQ]->backend = VIRTIO_DEV_STOPPED;
271 dev->virtqueue[VIRTIO_TXQ]->backend = VIRTIO_DEV_STOPPED;
275 * Function is called from the CUSE open function. The device structure is
276 * initialised and a new entry is added to the device configuration linked
280 new_device(struct vhost_device_ctx ctx)
282 struct virtio_net_config_ll *new_ll_dev;
283 struct vhost_virtqueue *virtqueue_rx, *virtqueue_tx;
285 /* Setup device and virtqueues. */
286 new_ll_dev = malloc(sizeof(struct virtio_net_config_ll));
287 if (new_ll_dev == NULL) {
288 RTE_LOG(ERR, VHOST_CONFIG,
289 "(%"PRIu64") Failed to allocate memory for dev.\n",
294 virtqueue_rx = malloc(sizeof(struct vhost_virtqueue));
295 if (virtqueue_rx == NULL) {
297 RTE_LOG(ERR, VHOST_CONFIG,
298 "(%"PRIu64") Failed to allocate memory for rxq.\n",
303 virtqueue_tx = malloc(sizeof(struct vhost_virtqueue));
304 if (virtqueue_tx == NULL) {
307 RTE_LOG(ERR, VHOST_CONFIG,
308 "(%"PRIu64") Failed to allocate memory for txq.\n",
313 new_ll_dev->dev.virtqueue[VIRTIO_RXQ] = virtqueue_rx;
314 new_ll_dev->dev.virtqueue[VIRTIO_TXQ] = virtqueue_tx;
316 /* Initialise device and virtqueues. */
317 init_device(&new_ll_dev->dev);
319 new_ll_dev->next = NULL;
321 /* Add entry to device configuration linked list. */
322 add_config_ll_entry(new_ll_dev);
324 return new_ll_dev->dev.device_fh;
328 * Function is called from the CUSE release function. This function will
329 * cleanup the device and remove it from device configuration linked list.
332 destroy_device(struct vhost_device_ctx ctx)
334 struct virtio_net_config_ll *ll_dev_cur_ctx, *ll_dev_last = NULL;
335 struct virtio_net_config_ll *ll_dev_cur = ll_root;
337 /* Find the linked list entry for the device to be removed. */
338 ll_dev_cur_ctx = get_config_ll_entry(ctx);
339 while (ll_dev_cur != NULL) {
341 * If the device is found or
342 * a device that doesn't exist is found then it is removed.
344 if (ll_dev_cur == ll_dev_cur_ctx) {
346 * If the device is running on a data core then call
347 * the function to remove it from the data core.
349 if ((ll_dev_cur->dev.flags & VIRTIO_DEV_RUNNING))
350 notify_ops->destroy_device(&(ll_dev_cur->dev));
351 ll_dev_cur = rm_config_ll_entry(ll_dev_cur,
354 ll_dev_last = ll_dev_cur;
355 ll_dev_cur = ll_dev_cur->next;
361 * Called from CUSE IOCTL: VHOST_SET_OWNER
362 * This function just returns success at the moment unless
363 * the device hasn't been initialised.
366 set_owner(struct vhost_device_ctx ctx)
368 struct virtio_net *dev;
370 dev = get_device(ctx);
378 * Called from CUSE IOCTL: VHOST_RESET_OWNER
381 reset_owner(struct vhost_device_ctx ctx)
383 struct virtio_net_config_ll *ll_dev;
385 ll_dev = get_config_ll_entry(ctx);
387 cleanup_device(&ll_dev->dev);
388 init_device(&ll_dev->dev);
394 * Called from CUSE IOCTL: VHOST_GET_FEATURES
395 * The features that we support are requested.
398 get_features(struct vhost_device_ctx ctx, uint64_t *pu)
400 struct virtio_net *dev;
402 dev = get_device(ctx);
406 /* Send our supported features. */
407 *pu = VHOST_FEATURES;
412 * Called from CUSE IOCTL: VHOST_SET_FEATURES
413 * We receive the negotiated features supported by us and the virtio device.
416 set_features(struct vhost_device_ctx ctx, uint64_t *pu)
418 struct virtio_net *dev;
420 dev = get_device(ctx);
423 if (*pu & ~VHOST_FEATURES)
426 /* Store the negotiated feature list for the device. */
429 /* Set the vhost_hlen depending on if VIRTIO_NET_F_MRG_RXBUF is set. */
430 if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
431 LOG_DEBUG(VHOST_CONFIG,
432 "(%"PRIu64") Mergeable RX buffers enabled\n",
434 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
435 sizeof(struct virtio_net_hdr_mrg_rxbuf);
436 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
437 sizeof(struct virtio_net_hdr_mrg_rxbuf);
439 LOG_DEBUG(VHOST_CONFIG,
440 "(%"PRIu64") Mergeable RX buffers disabled\n",
442 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
443 sizeof(struct virtio_net_hdr);
444 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
445 sizeof(struct virtio_net_hdr);
451 * Called from CUSE IOCTL: VHOST_SET_VRING_NUM
452 * The virtio device sends us the size of the descriptor ring.
455 set_vring_num(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
457 struct virtio_net *dev;
459 dev = get_device(ctx);
463 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
464 dev->virtqueue[state->index]->size = state->num;
470 * Called from CUSE IOCTL: VHOST_SET_VRING_ADDR
471 * The virtio device sends us the desc, used and avail ring addresses.
472 * This function then converts these to our address space.
475 set_vring_addr(struct vhost_device_ctx ctx, struct vhost_vring_addr *addr)
477 struct virtio_net *dev;
478 struct vhost_virtqueue *vq;
480 dev = get_device(ctx);
484 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
485 vq = dev->virtqueue[addr->index];
487 /* The addresses are converted from QEMU virtual to Vhost virtual. */
488 vq->desc = (struct vring_desc *)(uintptr_t)qva_to_vva(dev,
489 addr->desc_user_addr);
491 RTE_LOG(ERR, VHOST_CONFIG,
492 "(%"PRIu64") Failed to find desc ring address.\n",
497 vq->avail = (struct vring_avail *)(uintptr_t)qva_to_vva(dev,
498 addr->avail_user_addr);
499 if (vq->avail == 0) {
500 RTE_LOG(ERR, VHOST_CONFIG,
501 "(%"PRIu64") Failed to find avail ring address.\n",
506 vq->used = (struct vring_used *)(uintptr_t)qva_to_vva(dev,
507 addr->used_user_addr);
509 RTE_LOG(ERR, VHOST_CONFIG,
510 "(%"PRIu64") Failed to find used ring address.\n",
515 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address desc: %p\n",
516 dev->device_fh, vq->desc);
517 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address avail: %p\n",
518 dev->device_fh, vq->avail);
519 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address used: %p\n",
520 dev->device_fh, vq->used);
526 * Called from CUSE IOCTL: VHOST_SET_VRING_BASE
527 * The virtio device sends us the available ring last used index.
530 set_vring_base(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
532 struct virtio_net *dev;
534 dev = get_device(ctx);
538 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
539 dev->virtqueue[state->index]->last_used_idx = state->num;
540 dev->virtqueue[state->index]->last_used_idx_res = state->num;
546 * Called from CUSE IOCTL: VHOST_GET_VRING_BASE
547 * We send the virtio device our available ring last used index.
550 get_vring_base(struct vhost_device_ctx ctx, uint32_t index,
551 struct vhost_vring_state *state)
553 struct virtio_net *dev;
555 dev = get_device(ctx);
559 state->index = index;
560 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
561 state->num = dev->virtqueue[state->index]->last_used_idx;
568 * Called from CUSE IOCTL: VHOST_SET_VRING_CALL
569 * The virtio device sends an eventfd to interrupt the guest. This fd gets
570 * copied into our process space.
573 set_vring_call(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
575 struct virtio_net *dev;
576 struct vhost_virtqueue *vq;
578 dev = get_device(ctx);
582 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
583 vq = dev->virtqueue[file->index];
585 if ((int)vq->kickfd >= 0)
586 close((int)vq->kickfd);
588 vq->kickfd = file->fd;
594 * Called from CUSE IOCTL: VHOST_SET_VRING_KICK
595 * The virtio device sends an eventfd that it can use to notify us.
596 * This fd gets copied into our process space.
599 set_vring_kick(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
601 struct virtio_net *dev;
602 struct vhost_virtqueue *vq;
604 dev = get_device(ctx);
608 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
609 vq = dev->virtqueue[file->index];
611 if ((int)vq->callfd >= 0)
612 close((int)vq->callfd);
614 vq->callfd = file->fd;
620 * Function to get the tap device name from the provided file descriptor and
621 * save it in the device structure.
624 get_ifname(struct virtio_net *dev, int tap_fd, int pid)
628 uint32_t size, ifr_size;
631 fd_tap = eventfd_copy(tap_fd, pid);
635 ret = ioctl(fd_tap, TUNGETIFF, &ifr);
637 if (close(fd_tap) < 0)
638 RTE_LOG(ERR, VHOST_CONFIG,
639 "(%"PRIu64") fd close failed\n",
643 ifr_size = strnlen(ifr.ifr_name, sizeof(ifr.ifr_name));
644 size = ifr_size > sizeof(dev->ifname) ?
645 sizeof(dev->ifname) : ifr_size;
647 strncpy(dev->ifname, ifr.ifr_name, size);
649 RTE_LOG(ERR, VHOST_CONFIG,
650 "(%"PRIu64") TUNGETIFF ioctl failed\n",
657 * Called from CUSE IOCTL: VHOST_NET_SET_BACKEND
658 * To complete device initialisation when the virtio driver is loaded,
659 * we are provided with a valid fd for a tap device (not used by us).
660 * If this happens then we can add the device to a data core.
661 * When the virtio driver is removed we get fd=-1.
662 * At that point we remove the device from the data core.
663 * The device will still exist in the device configuration linked list.
666 set_backend(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
668 struct virtio_net *dev;
670 dev = get_device(ctx);
674 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
675 dev->virtqueue[file->index]->backend = file->fd;
678 * If the device isn't already running and both backend fds are set,
681 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
682 if (((int)dev->virtqueue[VIRTIO_TXQ]->backend != VIRTIO_DEV_STOPPED) &&
683 ((int)dev->virtqueue[VIRTIO_RXQ]->backend != VIRTIO_DEV_STOPPED)) {
684 get_ifname(dev, file->fd, ctx.pid);
685 return notify_ops->new_device(dev);
687 /* Otherwise we remove it. */
689 if (file->fd == VIRTIO_DEV_STOPPED)
690 notify_ops->destroy_device(dev);
695 * Function pointers are set for the device operations to allow CUSE to call
696 * functions when an IOCTL, device_add or device_release is received.
698 static const struct vhost_net_device_ops vhost_device_ops = {
699 .new_device = new_device,
700 .destroy_device = destroy_device,
702 .get_features = get_features,
703 .set_features = set_features,
705 .set_vring_num = set_vring_num,
706 .set_vring_addr = set_vring_addr,
707 .set_vring_base = set_vring_base,
708 .get_vring_base = get_vring_base,
710 .set_vring_kick = set_vring_kick,
711 .set_vring_call = set_vring_call,
713 .set_backend = set_backend,
715 .set_owner = set_owner,
716 .reset_owner = reset_owner,
720 * Called by main to setup callbacks when registering CUSE device.
722 struct vhost_net_device_ops const *
723 get_virtio_net_callbacks(void)
725 return &vhost_device_ops;
728 int rte_vhost_enable_guest_notification(struct virtio_net *dev,
729 uint16_t queue_id, int enable)
732 RTE_LOG(ERR, VHOST_CONFIG,
733 "guest notification isn't supported.\n");
737 dev->virtqueue[queue_id]->used->flags =
738 enable ? 0 : VRING_USED_F_NO_NOTIFY;
742 uint64_t rte_vhost_feature_get(void)
744 return VHOST_FEATURES;
747 int rte_vhost_feature_disable(uint64_t feature_mask)
749 VHOST_FEATURES = VHOST_FEATURES & ~feature_mask;
753 int rte_vhost_feature_enable(uint64_t feature_mask)
755 if ((feature_mask & VHOST_SUPPORTED_FEATURES) == feature_mask) {
756 VHOST_FEATURES = VHOST_FEATURES | feature_mask;
763 * Register ops so that we can add/remove device to data core.
766 rte_vhost_driver_callback_register(struct virtio_net_device_ops const * const ops)