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 = dev->mem->mapped_address + qemu_va -
96 dev->mem->base_address;
105 * Retrieves an entry from the devices configuration linked list.
107 static struct virtio_net_config_ll *
108 get_config_ll_entry(struct vhost_device_ctx ctx)
110 struct virtio_net_config_ll *ll_dev = ll_root;
112 /* Loop through linked list until the device_fh is found. */
113 while (ll_dev != NULL) {
114 if (ll_dev->dev.device_fh == ctx.fh)
116 ll_dev = ll_dev->next;
123 * Searches the configuration core linked list and
124 * retrieves the device if it exists.
127 get_device(struct vhost_device_ctx ctx)
129 struct virtio_net_config_ll *ll_dev;
131 ll_dev = get_config_ll_entry(ctx);
136 RTE_LOG(ERR, VHOST_CONFIG,
137 "(%"PRIu64") Device not found in linked list.\n", ctx.fh);
142 * Add entry containing a device to the device configuration linked list.
145 add_config_ll_entry(struct virtio_net_config_ll *new_ll_dev)
147 struct virtio_net_config_ll *ll_dev = ll_root;
149 /* If ll_dev == NULL then this is the first device so go to else */
151 /* If the 1st device_fh != 0 then we insert our device here. */
152 if (ll_dev->dev.device_fh != 0) {
153 new_ll_dev->dev.device_fh = 0;
154 new_ll_dev->next = ll_dev;
155 ll_root = new_ll_dev;
158 * Increment through the ll until we find un unused
159 * device_fh. Insert the device at that entry.
161 while ((ll_dev->next != NULL) &&
162 (ll_dev->dev.device_fh ==
163 (ll_dev->next->dev.device_fh - 1)))
164 ll_dev = ll_dev->next;
166 new_ll_dev->dev.device_fh = ll_dev->dev.device_fh + 1;
167 new_ll_dev->next = ll_dev->next;
168 ll_dev->next = new_ll_dev;
171 ll_root = new_ll_dev;
172 ll_root->dev.device_fh = 0;
178 * Unmap any memory, close any file descriptors and
179 * free any memory owned by a device.
182 cleanup_device(struct virtio_net *dev)
184 /* Unmap QEMU memory file if mapped. */
186 munmap((void *)(uintptr_t)dev->mem->mapped_address,
187 (size_t)dev->mem->mapped_size);
191 /* Close any event notifiers opened by device. */
192 if (dev->virtqueue[VIRTIO_RXQ]->callfd)
193 close((int)dev->virtqueue[VIRTIO_RXQ]->callfd);
194 if (dev->virtqueue[VIRTIO_RXQ]->kickfd)
195 close((int)dev->virtqueue[VIRTIO_RXQ]->kickfd);
196 if (dev->virtqueue[VIRTIO_TXQ]->callfd)
197 close((int)dev->virtqueue[VIRTIO_TXQ]->callfd);
198 if (dev->virtqueue[VIRTIO_TXQ]->kickfd)
199 close((int)dev->virtqueue[VIRTIO_TXQ]->kickfd);
203 * Release virtqueues and device memory.
206 free_device(struct virtio_net_config_ll *ll_dev)
208 /* Free any malloc'd memory */
209 free(ll_dev->dev.virtqueue[VIRTIO_RXQ]);
210 free(ll_dev->dev.virtqueue[VIRTIO_TXQ]);
215 * Remove an entry from the device configuration linked list.
217 static struct virtio_net_config_ll *
218 rm_config_ll_entry(struct virtio_net_config_ll *ll_dev,
219 struct virtio_net_config_ll *ll_dev_last)
221 /* First remove the device and then clean it up. */
222 if (ll_dev == ll_root) {
223 ll_root = ll_dev->next;
224 cleanup_device(&ll_dev->dev);
228 if (likely(ll_dev_last != NULL)) {
229 ll_dev_last->next = ll_dev->next;
230 cleanup_device(&ll_dev->dev);
232 return ll_dev_last->next;
234 cleanup_device(&ll_dev->dev);
236 RTE_LOG(ERR, VHOST_CONFIG,
237 "Remove entry from config_ll failed\n");
244 * Initialise all variables in device structure.
247 init_device(struct virtio_net *dev)
252 * Virtqueues have already been malloced so
253 * we don't want to set them to NULL.
255 vq_offset = offsetof(struct virtio_net, mem);
257 /* Set everything to 0. */
258 memset((void *)(uintptr_t)((uint64_t)(uintptr_t)dev + vq_offset), 0,
259 (sizeof(struct virtio_net) - (size_t)vq_offset));
260 memset(dev->virtqueue[VIRTIO_RXQ], 0, sizeof(struct vhost_virtqueue));
261 memset(dev->virtqueue[VIRTIO_TXQ], 0, sizeof(struct vhost_virtqueue));
263 /* Backends are set to -1 indicating an inactive device. */
264 dev->virtqueue[VIRTIO_RXQ]->backend = VIRTIO_DEV_STOPPED;
265 dev->virtqueue[VIRTIO_TXQ]->backend = VIRTIO_DEV_STOPPED;
269 * Function is called from the CUSE open function. The device structure is
270 * initialised and a new entry is added to the device configuration linked
274 new_device(struct vhost_device_ctx ctx)
276 struct virtio_net_config_ll *new_ll_dev;
277 struct vhost_virtqueue *virtqueue_rx, *virtqueue_tx;
279 /* Setup device and virtqueues. */
280 new_ll_dev = malloc(sizeof(struct virtio_net_config_ll));
281 if (new_ll_dev == NULL) {
282 RTE_LOG(ERR, VHOST_CONFIG,
283 "(%"PRIu64") Failed to allocate memory for dev.\n",
288 virtqueue_rx = malloc(sizeof(struct vhost_virtqueue));
289 if (virtqueue_rx == NULL) {
291 RTE_LOG(ERR, VHOST_CONFIG,
292 "(%"PRIu64") Failed to allocate memory for rxq.\n",
297 virtqueue_tx = malloc(sizeof(struct vhost_virtqueue));
298 if (virtqueue_tx == NULL) {
301 RTE_LOG(ERR, VHOST_CONFIG,
302 "(%"PRIu64") Failed to allocate memory for txq.\n",
307 new_ll_dev->dev.virtqueue[VIRTIO_RXQ] = virtqueue_rx;
308 new_ll_dev->dev.virtqueue[VIRTIO_TXQ] = virtqueue_tx;
310 /* Initialise device and virtqueues. */
311 init_device(&new_ll_dev->dev);
313 new_ll_dev->next = NULL;
315 /* Add entry to device configuration linked list. */
316 add_config_ll_entry(new_ll_dev);
318 return new_ll_dev->dev.device_fh;
322 * Function is called from the CUSE release function. This function will
323 * cleanup the device and remove it from device configuration linked list.
326 destroy_device(struct vhost_device_ctx ctx)
328 struct virtio_net_config_ll *ll_dev_cur_ctx, *ll_dev_last = NULL;
329 struct virtio_net_config_ll *ll_dev_cur = ll_root;
331 /* Find the linked list entry for the device to be removed. */
332 ll_dev_cur_ctx = get_config_ll_entry(ctx);
333 while (ll_dev_cur != NULL) {
335 * If the device is found or
336 * a device that doesn't exist is found then it is removed.
338 if (ll_dev_cur == ll_dev_cur_ctx) {
340 * If the device is running on a data core then call
341 * the function to remove it from the data core.
343 if ((ll_dev_cur->dev.flags & VIRTIO_DEV_RUNNING))
344 notify_ops->destroy_device(&(ll_dev_cur->dev));
345 ll_dev_cur = rm_config_ll_entry(ll_dev_cur,
348 ll_dev_last = ll_dev_cur;
349 ll_dev_cur = ll_dev_cur->next;
355 * Called from CUSE IOCTL: VHOST_SET_OWNER
356 * This function just returns success at the moment unless
357 * the device hasn't been initialised.
360 set_owner(struct vhost_device_ctx ctx)
362 struct virtio_net *dev;
364 dev = get_device(ctx);
372 * Called from CUSE IOCTL: VHOST_RESET_OWNER
375 reset_owner(struct vhost_device_ctx ctx)
377 struct virtio_net_config_ll *ll_dev;
379 ll_dev = get_config_ll_entry(ctx);
381 cleanup_device(&ll_dev->dev);
382 init_device(&ll_dev->dev);
388 * Called from CUSE IOCTL: VHOST_GET_FEATURES
389 * The features that we support are requested.
392 get_features(struct vhost_device_ctx ctx, uint64_t *pu)
394 struct virtio_net *dev;
396 dev = get_device(ctx);
400 /* Send our supported features. */
401 *pu = VHOST_FEATURES;
406 * Called from CUSE IOCTL: VHOST_SET_FEATURES
407 * We receive the negotiated features supported by us and the virtio device.
410 set_features(struct vhost_device_ctx ctx, uint64_t *pu)
412 struct virtio_net *dev;
414 dev = get_device(ctx);
417 if (*pu & ~VHOST_FEATURES)
420 /* Store the negotiated feature list for the device. */
423 /* Set the vhost_hlen depending on if VIRTIO_NET_F_MRG_RXBUF is set. */
424 if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
425 LOG_DEBUG(VHOST_CONFIG,
426 "(%"PRIu64") Mergeable RX buffers enabled\n",
428 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
429 sizeof(struct virtio_net_hdr_mrg_rxbuf);
430 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
431 sizeof(struct virtio_net_hdr_mrg_rxbuf);
433 LOG_DEBUG(VHOST_CONFIG,
434 "(%"PRIu64") Mergeable RX buffers disabled\n",
436 dev->virtqueue[VIRTIO_RXQ]->vhost_hlen =
437 sizeof(struct virtio_net_hdr);
438 dev->virtqueue[VIRTIO_TXQ]->vhost_hlen =
439 sizeof(struct virtio_net_hdr);
445 * Called from CUSE IOCTL: VHOST_SET_VRING_NUM
446 * The virtio device sends us the size of the descriptor ring.
449 set_vring_num(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
451 struct virtio_net *dev;
453 dev = get_device(ctx);
457 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
458 dev->virtqueue[state->index]->size = state->num;
464 * Called from CUSE IOCTL: VHOST_SET_VRING_ADDR
465 * The virtio device sends us the desc, used and avail ring addresses.
466 * This function then converts these to our address space.
469 set_vring_addr(struct vhost_device_ctx ctx, struct vhost_vring_addr *addr)
471 struct virtio_net *dev;
472 struct vhost_virtqueue *vq;
474 dev = get_device(ctx);
478 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
479 vq = dev->virtqueue[addr->index];
481 /* The addresses are converted from QEMU virtual to Vhost virtual. */
482 vq->desc = (struct vring_desc *)(uintptr_t)qva_to_vva(dev,
483 addr->desc_user_addr);
485 RTE_LOG(ERR, VHOST_CONFIG,
486 "(%"PRIu64") Failed to find desc ring address.\n",
491 vq->avail = (struct vring_avail *)(uintptr_t)qva_to_vva(dev,
492 addr->avail_user_addr);
493 if (vq->avail == 0) {
494 RTE_LOG(ERR, VHOST_CONFIG,
495 "(%"PRIu64") Failed to find avail ring address.\n",
500 vq->used = (struct vring_used *)(uintptr_t)qva_to_vva(dev,
501 addr->used_user_addr);
503 RTE_LOG(ERR, VHOST_CONFIG,
504 "(%"PRIu64") Failed to find used ring address.\n",
509 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address desc: %p\n",
510 dev->device_fh, vq->desc);
511 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address avail: %p\n",
512 dev->device_fh, vq->avail);
513 LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") mapped address used: %p\n",
514 dev->device_fh, vq->used);
520 * Called from CUSE IOCTL: VHOST_SET_VRING_BASE
521 * The virtio device sends us the available ring last used index.
524 set_vring_base(struct vhost_device_ctx ctx, struct vhost_vring_state *state)
526 struct virtio_net *dev;
528 dev = get_device(ctx);
532 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
533 dev->virtqueue[state->index]->last_used_idx = state->num;
534 dev->virtqueue[state->index]->last_used_idx_res = state->num;
540 * Called from CUSE IOCTL: VHOST_GET_VRING_BASE
541 * We send the virtio device our available ring last used index.
544 get_vring_base(struct vhost_device_ctx ctx, uint32_t index,
545 struct vhost_vring_state *state)
547 struct virtio_net *dev;
549 dev = get_device(ctx);
553 state->index = index;
554 /* State->index refers to the queue index. The txq is 1, rxq is 0. */
555 state->num = dev->virtqueue[state->index]->last_used_idx;
562 * Called from CUSE IOCTL: VHOST_SET_VRING_CALL
563 * The virtio device sends an eventfd to interrupt the guest. This fd gets
564 * copied into our process space.
567 set_vring_call(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
569 struct virtio_net *dev;
570 struct vhost_virtqueue *vq;
572 dev = get_device(ctx);
576 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
577 vq = dev->virtqueue[file->index];
580 close((int)vq->kickfd);
582 vq->kickfd = file->fd;
588 * Called from CUSE IOCTL: VHOST_SET_VRING_KICK
589 * The virtio device sends an eventfd that it can use to notify us.
590 * This fd gets copied into our process space.
593 set_vring_kick(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
595 struct virtio_net *dev;
596 struct vhost_virtqueue *vq;
598 dev = get_device(ctx);
602 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
603 vq = dev->virtqueue[file->index];
606 close((int)vq->callfd);
607 vq->callfd = file->fd;
613 * Function to get the tap device name from the provided file descriptor and
614 * save it in the device structure.
617 get_ifname(struct virtio_net *dev, int tap_fd, int pid)
621 uint32_t size, ifr_size;
624 fd_tap = eventfd_copy(tap_fd, pid);
628 ret = ioctl(fd_tap, TUNGETIFF, &ifr);
630 if (close(fd_tap) < 0)
631 RTE_LOG(ERR, VHOST_CONFIG,
632 "(%"PRIu64") fd close failed\n",
636 ifr_size = strnlen(ifr.ifr_name, sizeof(ifr.ifr_name));
637 size = ifr_size > sizeof(dev->ifname) ?
638 sizeof(dev->ifname) : ifr_size;
640 strncpy(dev->ifname, ifr.ifr_name, size);
642 RTE_LOG(ERR, VHOST_CONFIG,
643 "(%"PRIu64") TUNGETIFF ioctl failed\n",
650 * Called from CUSE IOCTL: VHOST_NET_SET_BACKEND
651 * To complete device initialisation when the virtio driver is loaded,
652 * we are provided with a valid fd for a tap device (not used by us).
653 * If this happens then we can add the device to a data core.
654 * When the virtio driver is removed we get fd=-1.
655 * At that point we remove the device from the data core.
656 * The device will still exist in the device configuration linked list.
659 set_backend(struct vhost_device_ctx ctx, struct vhost_vring_file *file)
661 struct virtio_net *dev;
663 dev = get_device(ctx);
667 /* file->index refers to the queue index. The txq is 1, rxq is 0. */
668 dev->virtqueue[file->index]->backend = file->fd;
671 * If the device isn't already running and both backend fds are set,
674 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
675 if (((int)dev->virtqueue[VIRTIO_TXQ]->backend != VIRTIO_DEV_STOPPED) &&
676 ((int)dev->virtqueue[VIRTIO_RXQ]->backend != VIRTIO_DEV_STOPPED)) {
677 get_ifname(dev, file->fd, ctx.pid);
678 return notify_ops->new_device(dev);
680 /* Otherwise we remove it. */
682 if (file->fd == VIRTIO_DEV_STOPPED)
683 notify_ops->destroy_device(dev);
688 * Function pointers are set for the device operations to allow CUSE to call
689 * functions when an IOCTL, device_add or device_release is received.
691 static const struct vhost_net_device_ops vhost_device_ops = {
692 .new_device = new_device,
693 .destroy_device = destroy_device,
695 .get_features = get_features,
696 .set_features = set_features,
698 .set_vring_num = set_vring_num,
699 .set_vring_addr = set_vring_addr,
700 .set_vring_base = set_vring_base,
701 .get_vring_base = get_vring_base,
703 .set_vring_kick = set_vring_kick,
704 .set_vring_call = set_vring_call,
706 .set_backend = set_backend,
708 .set_owner = set_owner,
709 .reset_owner = reset_owner,
713 * Called by main to setup callbacks when registering CUSE device.
715 struct vhost_net_device_ops const *
716 get_virtio_net_callbacks(void)
718 return &vhost_device_ops;
721 int rte_vhost_enable_guest_notification(struct virtio_net *dev,
722 uint16_t queue_id, int enable)
725 RTE_LOG(ERR, VHOST_CONFIG,
726 "guest notification isn't supported.\n");
730 dev->virtqueue[queue_id]->used->flags =
731 enable ? 0 : VRING_USED_F_NO_NOTIFY;
735 uint64_t rte_vhost_feature_get(void)
737 return VHOST_FEATURES;
740 int rte_vhost_feature_disable(uint64_t feature_mask)
742 VHOST_FEATURES = VHOST_FEATURES & ~feature_mask;
746 int rte_vhost_feature_enable(uint64_t feature_mask)
748 if ((feature_mask & VHOST_SUPPORTED_FEATURES) == feature_mask) {
749 VHOST_FEATURES = VHOST_FEATURES | feature_mask;
756 * Register ops so that we can add/remove device to data core.
759 rte_vhost_driver_callback_register(struct virtio_net_device_ops const * const ops)