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.
40 #include <sys/types.h>
43 #ifdef RTE_LIBRTE_VHOST_NUMA
47 #include <rte_common.h>
48 #include <rte_malloc.h>
52 #include "vhost_user.h"
54 #define VIRTIO_MIN_MTU 68
55 #define VIRTIO_MAX_MTU 65535
57 static const char *vhost_message_str[VHOST_USER_MAX] = {
58 [VHOST_USER_NONE] = "VHOST_USER_NONE",
59 [VHOST_USER_GET_FEATURES] = "VHOST_USER_GET_FEATURES",
60 [VHOST_USER_SET_FEATURES] = "VHOST_USER_SET_FEATURES",
61 [VHOST_USER_SET_OWNER] = "VHOST_USER_SET_OWNER",
62 [VHOST_USER_RESET_OWNER] = "VHOST_USER_RESET_OWNER",
63 [VHOST_USER_SET_MEM_TABLE] = "VHOST_USER_SET_MEM_TABLE",
64 [VHOST_USER_SET_LOG_BASE] = "VHOST_USER_SET_LOG_BASE",
65 [VHOST_USER_SET_LOG_FD] = "VHOST_USER_SET_LOG_FD",
66 [VHOST_USER_SET_VRING_NUM] = "VHOST_USER_SET_VRING_NUM",
67 [VHOST_USER_SET_VRING_ADDR] = "VHOST_USER_SET_VRING_ADDR",
68 [VHOST_USER_SET_VRING_BASE] = "VHOST_USER_SET_VRING_BASE",
69 [VHOST_USER_GET_VRING_BASE] = "VHOST_USER_GET_VRING_BASE",
70 [VHOST_USER_SET_VRING_KICK] = "VHOST_USER_SET_VRING_KICK",
71 [VHOST_USER_SET_VRING_CALL] = "VHOST_USER_SET_VRING_CALL",
72 [VHOST_USER_SET_VRING_ERR] = "VHOST_USER_SET_VRING_ERR",
73 [VHOST_USER_GET_PROTOCOL_FEATURES] = "VHOST_USER_GET_PROTOCOL_FEATURES",
74 [VHOST_USER_SET_PROTOCOL_FEATURES] = "VHOST_USER_SET_PROTOCOL_FEATURES",
75 [VHOST_USER_GET_QUEUE_NUM] = "VHOST_USER_GET_QUEUE_NUM",
76 [VHOST_USER_SET_VRING_ENABLE] = "VHOST_USER_SET_VRING_ENABLE",
77 [VHOST_USER_SEND_RARP] = "VHOST_USER_SEND_RARP",
78 [VHOST_USER_NET_SET_MTU] = "VHOST_USER_NET_SET_MTU",
87 ret = fstat(fd, &stat);
88 return ret == -1 ? (uint64_t)-1 : (uint64_t)stat.st_blksize;
92 free_mem_region(struct virtio_net *dev)
95 struct rte_vhost_mem_region *reg;
97 if (!dev || !dev->mem)
100 for (i = 0; i < dev->mem->nregions; i++) {
101 reg = &dev->mem->regions[i];
102 if (reg->host_user_addr) {
103 munmap(reg->mmap_addr, reg->mmap_size);
110 vhost_backend_cleanup(struct virtio_net *dev)
113 free_mem_region(dev);
118 munmap((void *)(uintptr_t)dev->log_addr, dev->log_size);
124 * This function just returns success at the moment unless
125 * the device hasn't been initialised.
128 vhost_user_set_owner(void)
134 vhost_user_reset_owner(struct virtio_net *dev)
136 if (dev->flags & VIRTIO_DEV_RUNNING) {
137 dev->flags &= ~VIRTIO_DEV_RUNNING;
138 dev->notify_ops->destroy_device(dev->vid);
141 cleanup_device(dev, 0);
147 * The features that we support are requested.
150 vhost_user_get_features(struct virtio_net *dev)
152 uint64_t features = 0;
154 rte_vhost_driver_get_features(dev->ifname, &features);
159 * We receive the negotiated features supported by us and the virtio device.
162 vhost_user_set_features(struct virtio_net *dev, uint64_t features)
164 uint64_t vhost_features = 0;
166 rte_vhost_driver_get_features(dev->ifname, &vhost_features);
167 if (features & ~vhost_features)
170 if ((dev->flags & VIRTIO_DEV_RUNNING) && dev->features != features) {
171 if (dev->notify_ops->features_changed)
172 dev->notify_ops->features_changed(dev->vid, features);
175 dev->features = features;
177 ((1 << VIRTIO_NET_F_MRG_RXBUF) | (1ULL << VIRTIO_F_VERSION_1))) {
178 dev->vhost_hlen = sizeof(struct virtio_net_hdr_mrg_rxbuf);
180 dev->vhost_hlen = sizeof(struct virtio_net_hdr);
182 LOG_DEBUG(VHOST_CONFIG,
183 "(%d) mergeable RX buffers %s, virtio 1 %s\n",
185 (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) ? "on" : "off",
186 (dev->features & (1ULL << VIRTIO_F_VERSION_1)) ? "on" : "off");
192 * The virtio device sends us the size of the descriptor ring.
195 vhost_user_set_vring_num(struct virtio_net *dev,
196 struct vhost_vring_state *state)
198 struct vhost_virtqueue *vq = dev->virtqueue[state->index];
200 vq->size = state->num;
202 if (dev->dequeue_zero_copy) {
204 vq->last_zmbuf_idx = 0;
205 vq->zmbuf_size = vq->size;
206 vq->zmbufs = rte_zmalloc(NULL, vq->zmbuf_size *
207 sizeof(struct zcopy_mbuf), 0);
208 if (vq->zmbufs == NULL) {
209 RTE_LOG(WARNING, VHOST_CONFIG,
210 "failed to allocate mem for zero copy; "
211 "zero copy is force disabled\n");
212 dev->dequeue_zero_copy = 0;
216 vq->shadow_used_ring = rte_malloc(NULL,
217 vq->size * sizeof(struct vring_used_elem),
218 RTE_CACHE_LINE_SIZE);
219 if (!vq->shadow_used_ring) {
220 RTE_LOG(ERR, VHOST_CONFIG,
221 "failed to allocate memory for shadow used ring.\n");
229 * Reallocate virtio_dev and vhost_virtqueue data structure to make them on the
230 * same numa node as the memory of vring descriptor.
232 #ifdef RTE_LIBRTE_VHOST_NUMA
233 static struct virtio_net*
234 numa_realloc(struct virtio_net *dev, int index)
236 int oldnode, newnode;
237 struct virtio_net *old_dev;
238 struct vhost_virtqueue *old_vq, *vq;
241 enum {VIRTIO_RXQ, VIRTIO_TXQ, VIRTIO_QNUM};
244 vq = old_vq = dev->virtqueue[index];
246 ret = get_mempolicy(&newnode, NULL, 0, old_vq->desc,
247 MPOL_F_NODE | MPOL_F_ADDR);
249 /* check if we need to reallocate vq */
250 ret |= get_mempolicy(&oldnode, NULL, 0, old_vq,
251 MPOL_F_NODE | MPOL_F_ADDR);
253 RTE_LOG(ERR, VHOST_CONFIG,
254 "Unable to get vq numa information.\n");
257 if (oldnode != newnode) {
258 RTE_LOG(INFO, VHOST_CONFIG,
259 "reallocate vq from %d to %d node\n", oldnode, newnode);
260 vq = rte_malloc_socket(NULL, sizeof(*vq), 0, newnode);
264 memcpy(vq, old_vq, sizeof(*vq) * VIRTIO_QNUM);
268 /* check if we need to reallocate dev */
269 ret = get_mempolicy(&oldnode, NULL, 0, old_dev,
270 MPOL_F_NODE | MPOL_F_ADDR);
272 RTE_LOG(ERR, VHOST_CONFIG,
273 "Unable to get dev numa information.\n");
276 if (oldnode != newnode) {
277 RTE_LOG(INFO, VHOST_CONFIG,
278 "reallocate dev from %d to %d node\n",
280 dev = rte_malloc_socket(NULL, sizeof(*dev), 0, newnode);
286 memcpy(dev, old_dev, sizeof(*dev));
291 dev->virtqueue[index] = vq;
292 vhost_devices[dev->vid] = dev;
297 static struct virtio_net*
298 numa_realloc(struct virtio_net *dev, int index __rte_unused)
305 * Converts QEMU virtual address to Vhost virtual address. This function is
306 * used to convert the ring addresses to our address space.
309 qva_to_vva(struct virtio_net *dev, uint64_t qva)
311 struct rte_vhost_mem_region *reg;
314 /* Find the region where the address lives. */
315 for (i = 0; i < dev->mem->nregions; i++) {
316 reg = &dev->mem->regions[i];
318 if (qva >= reg->guest_user_addr &&
319 qva < reg->guest_user_addr + reg->size) {
320 return qva - reg->guest_user_addr +
329 * The virtio device sends us the desc, used and avail ring addresses.
330 * This function then converts these to our address space.
333 vhost_user_set_vring_addr(struct virtio_net *dev, struct vhost_vring_addr *addr)
335 struct vhost_virtqueue *vq;
337 if (dev->mem == NULL)
340 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
341 vq = dev->virtqueue[addr->index];
343 /* The addresses are converted from QEMU virtual to Vhost virtual. */
344 vq->desc = (struct vring_desc *)(uintptr_t)qva_to_vva(dev,
345 addr->desc_user_addr);
347 RTE_LOG(ERR, VHOST_CONFIG,
348 "(%d) failed to find desc ring address.\n",
353 dev = numa_realloc(dev, addr->index);
354 vq = dev->virtqueue[addr->index];
356 vq->avail = (struct vring_avail *)(uintptr_t)qva_to_vva(dev,
357 addr->avail_user_addr);
358 if (vq->avail == 0) {
359 RTE_LOG(ERR, VHOST_CONFIG,
360 "(%d) failed to find avail ring address.\n",
365 vq->used = (struct vring_used *)(uintptr_t)qva_to_vva(dev,
366 addr->used_user_addr);
368 RTE_LOG(ERR, VHOST_CONFIG,
369 "(%d) failed to find used ring address.\n",
374 if (vq->last_used_idx != vq->used->idx) {
375 RTE_LOG(WARNING, VHOST_CONFIG,
376 "last_used_idx (%u) and vq->used->idx (%u) mismatches; "
377 "some packets maybe resent for Tx and dropped for Rx\n",
378 vq->last_used_idx, vq->used->idx);
379 vq->last_used_idx = vq->used->idx;
380 vq->last_avail_idx = vq->used->idx;
383 vq->log_guest_addr = addr->log_guest_addr;
385 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address desc: %p\n",
387 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address avail: %p\n",
388 dev->vid, vq->avail);
389 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address used: %p\n",
391 LOG_DEBUG(VHOST_CONFIG, "(%d) log_guest_addr: %" PRIx64 "\n",
392 dev->vid, vq->log_guest_addr);
398 * The virtio device sends us the available ring last used index.
401 vhost_user_set_vring_base(struct virtio_net *dev,
402 struct vhost_vring_state *state)
404 dev->virtqueue[state->index]->last_used_idx = state->num;
405 dev->virtqueue[state->index]->last_avail_idx = state->num;
411 add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
412 uint64_t host_phys_addr, uint64_t size)
414 struct guest_page *page, *last_page;
416 if (dev->nr_guest_pages == dev->max_guest_pages) {
417 dev->max_guest_pages *= 2;
418 dev->guest_pages = realloc(dev->guest_pages,
419 dev->max_guest_pages * sizeof(*page));
422 if (dev->nr_guest_pages > 0) {
423 last_page = &dev->guest_pages[dev->nr_guest_pages - 1];
424 /* merge if the two pages are continuous */
425 if (host_phys_addr == last_page->host_phys_addr +
427 last_page->size += size;
432 page = &dev->guest_pages[dev->nr_guest_pages++];
433 page->guest_phys_addr = guest_phys_addr;
434 page->host_phys_addr = host_phys_addr;
439 add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,
442 uint64_t reg_size = reg->size;
443 uint64_t host_user_addr = reg->host_user_addr;
444 uint64_t guest_phys_addr = reg->guest_phys_addr;
445 uint64_t host_phys_addr;
448 host_phys_addr = rte_mem_virt2phy((void *)(uintptr_t)host_user_addr);
449 size = page_size - (guest_phys_addr & (page_size - 1));
450 size = RTE_MIN(size, reg_size);
452 add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size);
453 host_user_addr += size;
454 guest_phys_addr += size;
457 while (reg_size > 0) {
458 size = RTE_MIN(reg_size, page_size);
459 host_phys_addr = rte_mem_virt2phy((void *)(uintptr_t)
461 add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size);
463 host_user_addr += size;
464 guest_phys_addr += size;
469 #ifdef RTE_LIBRTE_VHOST_DEBUG
470 /* TODO: enable it only in debug mode? */
472 dump_guest_pages(struct virtio_net *dev)
475 struct guest_page *page;
477 for (i = 0; i < dev->nr_guest_pages; i++) {
478 page = &dev->guest_pages[i];
480 RTE_LOG(INFO, VHOST_CONFIG,
481 "guest physical page region %u\n"
482 "\t guest_phys_addr: %" PRIx64 "\n"
483 "\t host_phys_addr : %" PRIx64 "\n"
484 "\t size : %" PRIx64 "\n",
486 page->guest_phys_addr,
487 page->host_phys_addr,
492 #define dump_guest_pages(dev)
496 vhost_user_set_mem_table(struct virtio_net *dev, struct VhostUserMsg *pmsg)
498 struct VhostUserMemory memory = pmsg->payload.memory;
499 struct rte_vhost_mem_region *reg;
502 uint64_t mmap_offset;
508 free_mem_region(dev);
513 dev->nr_guest_pages = 0;
514 if (!dev->guest_pages) {
515 dev->max_guest_pages = 8;
516 dev->guest_pages = malloc(dev->max_guest_pages *
517 sizeof(struct guest_page));
520 dev->mem = rte_zmalloc("vhost-mem-table", sizeof(struct rte_vhost_memory) +
521 sizeof(struct rte_vhost_mem_region) * memory.nregions, 0);
522 if (dev->mem == NULL) {
523 RTE_LOG(ERR, VHOST_CONFIG,
524 "(%d) failed to allocate memory for dev->mem\n",
528 dev->mem->nregions = memory.nregions;
530 for (i = 0; i < memory.nregions; i++) {
532 reg = &dev->mem->regions[i];
534 reg->guest_phys_addr = memory.regions[i].guest_phys_addr;
535 reg->guest_user_addr = memory.regions[i].userspace_addr;
536 reg->size = memory.regions[i].memory_size;
539 mmap_offset = memory.regions[i].mmap_offset;
540 mmap_size = reg->size + mmap_offset;
542 /* mmap() without flag of MAP_ANONYMOUS, should be called
543 * with length argument aligned with hugepagesz at older
544 * longterm version Linux, like 2.6.32 and 3.2.72, or
545 * mmap() will fail with EINVAL.
547 * to avoid failure, make sure in caller to keep length
550 alignment = get_blk_size(fd);
551 if (alignment == (uint64_t)-1) {
552 RTE_LOG(ERR, VHOST_CONFIG,
553 "couldn't get hugepage size through fstat\n");
556 mmap_size = RTE_ALIGN_CEIL(mmap_size, alignment);
558 mmap_addr = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE,
559 MAP_SHARED | MAP_POPULATE, fd, 0);
561 if (mmap_addr == MAP_FAILED) {
562 RTE_LOG(ERR, VHOST_CONFIG,
563 "mmap region %u failed.\n", i);
567 reg->mmap_addr = mmap_addr;
568 reg->mmap_size = mmap_size;
569 reg->host_user_addr = (uint64_t)(uintptr_t)mmap_addr +
572 if (dev->dequeue_zero_copy)
573 add_guest_pages(dev, reg, alignment);
575 RTE_LOG(INFO, VHOST_CONFIG,
576 "guest memory region %u, size: 0x%" PRIx64 "\n"
577 "\t guest physical addr: 0x%" PRIx64 "\n"
578 "\t guest virtual addr: 0x%" PRIx64 "\n"
579 "\t host virtual addr: 0x%" PRIx64 "\n"
580 "\t mmap addr : 0x%" PRIx64 "\n"
581 "\t mmap size : 0x%" PRIx64 "\n"
582 "\t mmap align: 0x%" PRIx64 "\n"
583 "\t mmap off : 0x%" PRIx64 "\n",
585 reg->guest_phys_addr,
586 reg->guest_user_addr,
588 (uint64_t)(uintptr_t)mmap_addr,
594 dump_guest_pages(dev);
599 free_mem_region(dev);
606 vq_is_ready(struct vhost_virtqueue *vq)
608 return vq && vq->desc &&
609 vq->kickfd != VIRTIO_UNINITIALIZED_EVENTFD &&
610 vq->callfd != VIRTIO_UNINITIALIZED_EVENTFD;
614 virtio_is_ready(struct virtio_net *dev)
616 struct vhost_virtqueue *vq;
619 if (dev->nr_vring == 0)
622 for (i = 0; i < dev->nr_vring; i++) {
623 vq = dev->virtqueue[i];
625 if (!vq_is_ready(vq))
629 RTE_LOG(INFO, VHOST_CONFIG,
630 "virtio is now ready for processing.\n");
635 vhost_user_set_vring_call(struct virtio_net *dev, struct VhostUserMsg *pmsg)
637 struct vhost_vring_file file;
638 struct vhost_virtqueue *vq;
640 file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
641 if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
642 file.fd = VIRTIO_INVALID_EVENTFD;
644 file.fd = pmsg->fds[0];
645 RTE_LOG(INFO, VHOST_CONFIG,
646 "vring call idx:%d file:%d\n", file.index, file.fd);
648 vq = dev->virtqueue[file.index];
652 vq->callfd = file.fd;
656 vhost_user_set_vring_kick(struct virtio_net *dev, struct VhostUserMsg *pmsg)
658 struct vhost_vring_file file;
659 struct vhost_virtqueue *vq;
661 file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
662 if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
663 file.fd = VIRTIO_INVALID_EVENTFD;
665 file.fd = pmsg->fds[0];
666 RTE_LOG(INFO, VHOST_CONFIG,
667 "vring kick idx:%d file:%d\n", file.index, file.fd);
669 vq = dev->virtqueue[file.index];
672 vq->kickfd = file.fd;
676 free_zmbufs(struct vhost_virtqueue *vq)
678 struct zcopy_mbuf *zmbuf, *next;
680 for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list);
681 zmbuf != NULL; zmbuf = next) {
682 next = TAILQ_NEXT(zmbuf, next);
684 rte_pktmbuf_free(zmbuf->mbuf);
685 TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next);
688 rte_free(vq->zmbufs);
692 * when virtio is stopped, qemu will send us the GET_VRING_BASE message.
695 vhost_user_get_vring_base(struct virtio_net *dev,
696 struct vhost_vring_state *state)
698 struct vhost_virtqueue *vq = dev->virtqueue[state->index];
700 /* We have to stop the queue (virtio) if it is running. */
701 if (dev->flags & VIRTIO_DEV_RUNNING) {
702 dev->flags &= ~VIRTIO_DEV_RUNNING;
703 dev->notify_ops->destroy_device(dev->vid);
706 dev->flags &= ~VIRTIO_DEV_READY;
708 /* Here we are safe to get the last used index */
709 state->num = vq->last_used_idx;
711 RTE_LOG(INFO, VHOST_CONFIG,
712 "vring base idx:%d file:%d\n", state->index, state->num);
714 * Based on current qemu vhost-user implementation, this message is
715 * sent and only sent in vhost_vring_stop.
716 * TODO: cleanup the vring, it isn't usable since here.
721 vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
723 if (dev->dequeue_zero_copy)
725 rte_free(vq->shadow_used_ring);
726 vq->shadow_used_ring = NULL;
732 * when virtio queues are ready to work, qemu will send us to
733 * enable the virtio queue pair.
736 vhost_user_set_vring_enable(struct virtio_net *dev,
737 struct vhost_vring_state *state)
739 int enable = (int)state->num;
741 RTE_LOG(INFO, VHOST_CONFIG,
742 "set queue enable: %d to qp idx: %d\n",
743 enable, state->index);
745 if (dev->notify_ops->vring_state_changed)
746 dev->notify_ops->vring_state_changed(dev->vid, state->index, enable);
748 dev->virtqueue[state->index]->enabled = enable;
754 vhost_user_set_protocol_features(struct virtio_net *dev,
755 uint64_t protocol_features)
757 if (protocol_features & ~VHOST_USER_PROTOCOL_FEATURES)
760 dev->protocol_features = protocol_features;
764 vhost_user_set_log_base(struct virtio_net *dev, struct VhostUserMsg *msg)
766 int fd = msg->fds[0];
771 RTE_LOG(ERR, VHOST_CONFIG, "invalid log fd: %d\n", fd);
775 if (msg->size != sizeof(VhostUserLog)) {
776 RTE_LOG(ERR, VHOST_CONFIG,
777 "invalid log base msg size: %"PRId32" != %d\n",
778 msg->size, (int)sizeof(VhostUserLog));
782 size = msg->payload.log.mmap_size;
783 off = msg->payload.log.mmap_offset;
784 RTE_LOG(INFO, VHOST_CONFIG,
785 "log mmap size: %"PRId64", offset: %"PRId64"\n",
789 * mmap from 0 to workaround a hugepage mmap bug: mmap will
790 * fail when offset is not page size aligned.
792 addr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
794 if (addr == MAP_FAILED) {
795 RTE_LOG(ERR, VHOST_CONFIG, "mmap log base failed!\n");
800 * Free previously mapped log memory on occasionally
801 * multiple VHOST_USER_SET_LOG_BASE.
804 munmap((void *)(uintptr_t)dev->log_addr, dev->log_size);
806 dev->log_addr = (uint64_t)(uintptr_t)addr;
807 dev->log_base = dev->log_addr + off;
808 dev->log_size = size;
814 * An rarp packet is constructed and broadcasted to notify switches about
815 * the new location of the migrated VM, so that packets from outside will
816 * not be lost after migration.
818 * However, we don't actually "send" a rarp packet here, instead, we set
819 * a flag 'broadcast_rarp' to let rte_vhost_dequeue_burst() inject it.
822 vhost_user_send_rarp(struct virtio_net *dev, struct VhostUserMsg *msg)
824 uint8_t *mac = (uint8_t *)&msg->payload.u64;
826 RTE_LOG(DEBUG, VHOST_CONFIG,
827 ":: mac: %02x:%02x:%02x:%02x:%02x:%02x\n",
828 mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]);
829 memcpy(dev->mac.addr_bytes, mac, 6);
832 * Set the flag to inject a RARP broadcast packet at
833 * rte_vhost_dequeue_burst().
835 * rte_smp_wmb() is for making sure the mac is copied
836 * before the flag is set.
839 rte_atomic16_set(&dev->broadcast_rarp, 1);
845 vhost_user_net_set_mtu(struct virtio_net *dev, struct VhostUserMsg *msg)
847 if (msg->payload.u64 < VIRTIO_MIN_MTU ||
848 msg->payload.u64 > VIRTIO_MAX_MTU) {
849 RTE_LOG(ERR, VHOST_CONFIG, "Invalid MTU size (%"PRIu64")\n",
855 dev->mtu = msg->payload.u64;
860 /* return bytes# of read on success or negative val on failure. */
862 read_vhost_message(int sockfd, struct VhostUserMsg *msg)
866 ret = read_fd_message(sockfd, (char *)msg, VHOST_USER_HDR_SIZE,
867 msg->fds, VHOST_MEMORY_MAX_NREGIONS);
871 if (msg && msg->size) {
872 if (msg->size > sizeof(msg->payload)) {
873 RTE_LOG(ERR, VHOST_CONFIG,
874 "invalid msg size: %d\n", msg->size);
877 ret = read(sockfd, &msg->payload, msg->size);
880 if (ret != (int)msg->size) {
881 RTE_LOG(ERR, VHOST_CONFIG,
882 "read control message failed\n");
891 send_vhost_message(int sockfd, struct VhostUserMsg *msg)
898 msg->flags &= ~VHOST_USER_VERSION_MASK;
899 msg->flags &= ~VHOST_USER_NEED_REPLY;
900 msg->flags |= VHOST_USER_VERSION;
901 msg->flags |= VHOST_USER_REPLY_MASK;
903 ret = send_fd_message(sockfd, (char *)msg,
904 VHOST_USER_HDR_SIZE + msg->size, NULL, 0);
910 * Allocate a queue pair if it hasn't been allocated yet
913 vhost_user_check_and_alloc_queue_pair(struct virtio_net *dev, VhostUserMsg *msg)
917 switch (msg->request) {
918 case VHOST_USER_SET_VRING_KICK:
919 case VHOST_USER_SET_VRING_CALL:
920 case VHOST_USER_SET_VRING_ERR:
921 vring_idx = msg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
923 case VHOST_USER_SET_VRING_NUM:
924 case VHOST_USER_SET_VRING_BASE:
925 case VHOST_USER_SET_VRING_ENABLE:
926 vring_idx = msg->payload.state.index;
928 case VHOST_USER_SET_VRING_ADDR:
929 vring_idx = msg->payload.addr.index;
935 if (vring_idx >= VHOST_MAX_VRING) {
936 RTE_LOG(ERR, VHOST_CONFIG,
937 "invalid vring index: %u\n", vring_idx);
941 if (dev->virtqueue[vring_idx])
944 return alloc_vring_queue(dev, vring_idx);
948 vhost_user_msg_handler(int vid, int fd)
950 struct virtio_net *dev;
951 struct VhostUserMsg msg;
954 dev = get_device(vid);
958 if (!dev->notify_ops) {
959 dev->notify_ops = vhost_driver_callback_get(dev->ifname);
960 if (!dev->notify_ops) {
961 RTE_LOG(ERR, VHOST_CONFIG,
962 "failed to get callback ops for driver %s\n",
968 ret = read_vhost_message(fd, &msg);
969 if (ret <= 0 || msg.request >= VHOST_USER_MAX) {
971 RTE_LOG(ERR, VHOST_CONFIG,
972 "vhost read message failed\n");
974 RTE_LOG(INFO, VHOST_CONFIG,
975 "vhost peer closed\n");
977 RTE_LOG(ERR, VHOST_CONFIG,
978 "vhost read incorrect message\n");
984 RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
985 vhost_message_str[msg.request]);
987 ret = vhost_user_check_and_alloc_queue_pair(dev, &msg);
989 RTE_LOG(ERR, VHOST_CONFIG,
990 "failed to alloc queue\n");
994 switch (msg.request) {
995 case VHOST_USER_GET_FEATURES:
996 msg.payload.u64 = vhost_user_get_features(dev);
997 msg.size = sizeof(msg.payload.u64);
998 send_vhost_message(fd, &msg);
1000 case VHOST_USER_SET_FEATURES:
1001 vhost_user_set_features(dev, msg.payload.u64);
1004 case VHOST_USER_GET_PROTOCOL_FEATURES:
1005 msg.payload.u64 = VHOST_USER_PROTOCOL_FEATURES;
1006 msg.size = sizeof(msg.payload.u64);
1007 send_vhost_message(fd, &msg);
1009 case VHOST_USER_SET_PROTOCOL_FEATURES:
1010 vhost_user_set_protocol_features(dev, msg.payload.u64);
1013 case VHOST_USER_SET_OWNER:
1014 vhost_user_set_owner();
1016 case VHOST_USER_RESET_OWNER:
1017 vhost_user_reset_owner(dev);
1020 case VHOST_USER_SET_MEM_TABLE:
1021 ret = vhost_user_set_mem_table(dev, &msg);
1024 case VHOST_USER_SET_LOG_BASE:
1025 vhost_user_set_log_base(dev, &msg);
1027 /* it needs a reply */
1028 msg.size = sizeof(msg.payload.u64);
1029 send_vhost_message(fd, &msg);
1031 case VHOST_USER_SET_LOG_FD:
1033 RTE_LOG(INFO, VHOST_CONFIG, "not implemented.\n");
1036 case VHOST_USER_SET_VRING_NUM:
1037 vhost_user_set_vring_num(dev, &msg.payload.state);
1039 case VHOST_USER_SET_VRING_ADDR:
1040 vhost_user_set_vring_addr(dev, &msg.payload.addr);
1042 case VHOST_USER_SET_VRING_BASE:
1043 vhost_user_set_vring_base(dev, &msg.payload.state);
1046 case VHOST_USER_GET_VRING_BASE:
1047 vhost_user_get_vring_base(dev, &msg.payload.state);
1048 msg.size = sizeof(msg.payload.state);
1049 send_vhost_message(fd, &msg);
1052 case VHOST_USER_SET_VRING_KICK:
1053 vhost_user_set_vring_kick(dev, &msg);
1055 case VHOST_USER_SET_VRING_CALL:
1056 vhost_user_set_vring_call(dev, &msg);
1059 case VHOST_USER_SET_VRING_ERR:
1060 if (!(msg.payload.u64 & VHOST_USER_VRING_NOFD_MASK))
1062 RTE_LOG(INFO, VHOST_CONFIG, "not implemented\n");
1065 case VHOST_USER_GET_QUEUE_NUM:
1066 msg.payload.u64 = VHOST_MAX_QUEUE_PAIRS;
1067 msg.size = sizeof(msg.payload.u64);
1068 send_vhost_message(fd, &msg);
1071 case VHOST_USER_SET_VRING_ENABLE:
1072 vhost_user_set_vring_enable(dev, &msg.payload.state);
1074 case VHOST_USER_SEND_RARP:
1075 vhost_user_send_rarp(dev, &msg);
1078 case VHOST_USER_NET_SET_MTU:
1079 ret = vhost_user_net_set_mtu(dev, &msg);
1088 if (msg.flags & VHOST_USER_NEED_REPLY) {
1089 msg.payload.u64 = !!ret;
1090 msg.size = sizeof(msg.payload.u64);
1091 send_vhost_message(fd, &msg);
1094 if (!(dev->flags & VIRTIO_DEV_RUNNING) && virtio_is_ready(dev)) {
1095 dev->flags |= VIRTIO_DEV_READY;
1097 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
1098 if (dev->dequeue_zero_copy) {
1099 RTE_LOG(INFO, VHOST_CONFIG,
1100 "dequeue zero copy is enabled\n");
1103 if (dev->notify_ops->new_device(dev->vid) == 0)
1104 dev->flags |= VIRTIO_DEV_RUNNING;