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 dev->features = features;
172 ((1 << VIRTIO_NET_F_MRG_RXBUF) | (1ULL << VIRTIO_F_VERSION_1))) {
173 dev->vhost_hlen = sizeof(struct virtio_net_hdr_mrg_rxbuf);
175 dev->vhost_hlen = sizeof(struct virtio_net_hdr);
177 LOG_DEBUG(VHOST_CONFIG,
178 "(%d) mergeable RX buffers %s, virtio 1 %s\n",
180 (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF)) ? "on" : "off",
181 (dev->features & (1ULL << VIRTIO_F_VERSION_1)) ? "on" : "off");
187 * The virtio device sends us the size of the descriptor ring.
190 vhost_user_set_vring_num(struct virtio_net *dev,
191 struct vhost_vring_state *state)
193 struct vhost_virtqueue *vq = dev->virtqueue[state->index];
195 vq->size = state->num;
197 if (dev->dequeue_zero_copy) {
199 vq->last_zmbuf_idx = 0;
200 vq->zmbuf_size = vq->size;
201 vq->zmbufs = rte_zmalloc(NULL, vq->zmbuf_size *
202 sizeof(struct zcopy_mbuf), 0);
203 if (vq->zmbufs == NULL) {
204 RTE_LOG(WARNING, VHOST_CONFIG,
205 "failed to allocate mem for zero copy; "
206 "zero copy is force disabled\n");
207 dev->dequeue_zero_copy = 0;
211 vq->shadow_used_ring = rte_malloc(NULL,
212 vq->size * sizeof(struct vring_used_elem),
213 RTE_CACHE_LINE_SIZE);
214 if (!vq->shadow_used_ring) {
215 RTE_LOG(ERR, VHOST_CONFIG,
216 "failed to allocate memory for shadow used ring.\n");
224 * Reallocate virtio_dev and vhost_virtqueue data structure to make them on the
225 * same numa node as the memory of vring descriptor.
227 #ifdef RTE_LIBRTE_VHOST_NUMA
228 static struct virtio_net*
229 numa_realloc(struct virtio_net *dev, int index)
231 int oldnode, newnode;
232 struct virtio_net *old_dev;
233 struct vhost_virtqueue *old_vq, *vq;
237 vq = old_vq = dev->virtqueue[index];
239 ret = get_mempolicy(&newnode, NULL, 0, old_vq->desc,
240 MPOL_F_NODE | MPOL_F_ADDR);
242 /* check if we need to reallocate vq */
243 ret |= get_mempolicy(&oldnode, NULL, 0, old_vq,
244 MPOL_F_NODE | MPOL_F_ADDR);
246 RTE_LOG(ERR, VHOST_CONFIG,
247 "Unable to get vq numa information.\n");
250 if (oldnode != newnode) {
251 RTE_LOG(INFO, VHOST_CONFIG,
252 "reallocate vq from %d to %d node\n", oldnode, newnode);
253 vq = rte_malloc_socket(NULL, sizeof(*vq), 0, newnode);
257 memcpy(vq, old_vq, sizeof(*vq) * VIRTIO_QNUM);
261 /* check if we need to reallocate dev */
262 ret = get_mempolicy(&oldnode, NULL, 0, old_dev,
263 MPOL_F_NODE | MPOL_F_ADDR);
265 RTE_LOG(ERR, VHOST_CONFIG,
266 "Unable to get dev numa information.\n");
269 if (oldnode != newnode) {
270 RTE_LOG(INFO, VHOST_CONFIG,
271 "reallocate dev from %d to %d node\n",
273 dev = rte_malloc_socket(NULL, sizeof(*dev), 0, newnode);
279 memcpy(dev, old_dev, sizeof(*dev));
284 dev->virtqueue[index] = vq;
285 vhost_devices[dev->vid] = dev;
290 static struct virtio_net*
291 numa_realloc(struct virtio_net *dev, int index __rte_unused)
298 * Converts QEMU virtual address to Vhost virtual address. This function is
299 * used to convert the ring addresses to our address space.
302 qva_to_vva(struct virtio_net *dev, uint64_t qva)
304 struct rte_vhost_mem_region *reg;
307 /* Find the region where the address lives. */
308 for (i = 0; i < dev->mem->nregions; i++) {
309 reg = &dev->mem->regions[i];
311 if (qva >= reg->guest_user_addr &&
312 qva < reg->guest_user_addr + reg->size) {
313 return qva - reg->guest_user_addr +
322 * The virtio device sends us the desc, used and avail ring addresses.
323 * This function then converts these to our address space.
326 vhost_user_set_vring_addr(struct virtio_net *dev, struct vhost_vring_addr *addr)
328 struct vhost_virtqueue *vq;
330 if (dev->mem == NULL)
333 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
334 vq = dev->virtqueue[addr->index];
336 /* The addresses are converted from QEMU virtual to Vhost virtual. */
337 vq->desc = (struct vring_desc *)(uintptr_t)qva_to_vva(dev,
338 addr->desc_user_addr);
340 RTE_LOG(ERR, VHOST_CONFIG,
341 "(%d) failed to find desc ring address.\n",
346 dev = numa_realloc(dev, addr->index);
347 vq = dev->virtqueue[addr->index];
349 vq->avail = (struct vring_avail *)(uintptr_t)qva_to_vva(dev,
350 addr->avail_user_addr);
351 if (vq->avail == 0) {
352 RTE_LOG(ERR, VHOST_CONFIG,
353 "(%d) failed to find avail ring address.\n",
358 vq->used = (struct vring_used *)(uintptr_t)qva_to_vva(dev,
359 addr->used_user_addr);
361 RTE_LOG(ERR, VHOST_CONFIG,
362 "(%d) failed to find used ring address.\n",
367 if (vq->last_used_idx != vq->used->idx) {
368 RTE_LOG(WARNING, VHOST_CONFIG,
369 "last_used_idx (%u) and vq->used->idx (%u) mismatches; "
370 "some packets maybe resent for Tx and dropped for Rx\n",
371 vq->last_used_idx, vq->used->idx);
372 vq->last_used_idx = vq->used->idx;
373 vq->last_avail_idx = vq->used->idx;
376 vq->log_guest_addr = addr->log_guest_addr;
378 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address desc: %p\n",
380 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address avail: %p\n",
381 dev->vid, vq->avail);
382 LOG_DEBUG(VHOST_CONFIG, "(%d) mapped address used: %p\n",
384 LOG_DEBUG(VHOST_CONFIG, "(%d) log_guest_addr: %" PRIx64 "\n",
385 dev->vid, vq->log_guest_addr);
391 * The virtio device sends us the available ring last used index.
394 vhost_user_set_vring_base(struct virtio_net *dev,
395 struct vhost_vring_state *state)
397 dev->virtqueue[state->index]->last_used_idx = state->num;
398 dev->virtqueue[state->index]->last_avail_idx = state->num;
404 add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
405 uint64_t host_phys_addr, uint64_t size)
407 struct guest_page *page, *last_page;
409 if (dev->nr_guest_pages == dev->max_guest_pages) {
410 dev->max_guest_pages *= 2;
411 dev->guest_pages = realloc(dev->guest_pages,
412 dev->max_guest_pages * sizeof(*page));
415 if (dev->nr_guest_pages > 0) {
416 last_page = &dev->guest_pages[dev->nr_guest_pages - 1];
417 /* merge if the two pages are continuous */
418 if (host_phys_addr == last_page->host_phys_addr +
420 last_page->size += size;
425 page = &dev->guest_pages[dev->nr_guest_pages++];
426 page->guest_phys_addr = guest_phys_addr;
427 page->host_phys_addr = host_phys_addr;
432 add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,
435 uint64_t reg_size = reg->size;
436 uint64_t host_user_addr = reg->host_user_addr;
437 uint64_t guest_phys_addr = reg->guest_phys_addr;
438 uint64_t host_phys_addr;
441 host_phys_addr = rte_mem_virt2phy((void *)(uintptr_t)host_user_addr);
442 size = page_size - (guest_phys_addr & (page_size - 1));
443 size = RTE_MIN(size, reg_size);
445 add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size);
446 host_user_addr += size;
447 guest_phys_addr += size;
450 while (reg_size > 0) {
451 size = RTE_MIN(reg_size, page_size);
452 host_phys_addr = rte_mem_virt2phy((void *)(uintptr_t)
454 add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size);
456 host_user_addr += size;
457 guest_phys_addr += size;
462 #ifdef RTE_LIBRTE_VHOST_DEBUG
463 /* TODO: enable it only in debug mode? */
465 dump_guest_pages(struct virtio_net *dev)
468 struct guest_page *page;
470 for (i = 0; i < dev->nr_guest_pages; i++) {
471 page = &dev->guest_pages[i];
473 RTE_LOG(INFO, VHOST_CONFIG,
474 "guest physical page region %u\n"
475 "\t guest_phys_addr: %" PRIx64 "\n"
476 "\t host_phys_addr : %" PRIx64 "\n"
477 "\t size : %" PRIx64 "\n",
479 page->guest_phys_addr,
480 page->host_phys_addr,
485 #define dump_guest_pages(dev)
489 vhost_user_set_mem_table(struct virtio_net *dev, struct VhostUserMsg *pmsg)
491 struct VhostUserMemory memory = pmsg->payload.memory;
492 struct rte_vhost_mem_region *reg;
495 uint64_t mmap_offset;
500 /* Remove from the data plane. */
501 if (dev->flags & VIRTIO_DEV_RUNNING) {
502 dev->flags &= ~VIRTIO_DEV_RUNNING;
503 dev->notify_ops->destroy_device(dev->vid);
507 free_mem_region(dev);
512 dev->nr_guest_pages = 0;
513 if (!dev->guest_pages) {
514 dev->max_guest_pages = 8;
515 dev->guest_pages = malloc(dev->max_guest_pages *
516 sizeof(struct guest_page));
519 dev->mem = rte_zmalloc("vhost-mem-table", sizeof(struct rte_vhost_memory) +
520 sizeof(struct rte_vhost_mem_region) * memory.nregions, 0);
521 if (dev->mem == NULL) {
522 RTE_LOG(ERR, VHOST_CONFIG,
523 "(%d) failed to allocate memory for dev->mem\n",
527 dev->mem->nregions = memory.nregions;
529 for (i = 0; i < memory.nregions; i++) {
531 reg = &dev->mem->regions[i];
533 reg->guest_phys_addr = memory.regions[i].guest_phys_addr;
534 reg->guest_user_addr = memory.regions[i].userspace_addr;
535 reg->size = memory.regions[i].memory_size;
538 mmap_offset = memory.regions[i].mmap_offset;
539 mmap_size = reg->size + mmap_offset;
541 /* mmap() without flag of MAP_ANONYMOUS, should be called
542 * with length argument aligned with hugepagesz at older
543 * longterm version Linux, like 2.6.32 and 3.2.72, or
544 * mmap() will fail with EINVAL.
546 * to avoid failure, make sure in caller to keep length
549 alignment = get_blk_size(fd);
550 if (alignment == (uint64_t)-1) {
551 RTE_LOG(ERR, VHOST_CONFIG,
552 "couldn't get hugepage size through fstat\n");
555 mmap_size = RTE_ALIGN_CEIL(mmap_size, alignment);
557 mmap_addr = mmap(NULL, mmap_size, PROT_READ | PROT_WRITE,
558 MAP_SHARED | MAP_POPULATE, fd, 0);
560 if (mmap_addr == MAP_FAILED) {
561 RTE_LOG(ERR, VHOST_CONFIG,
562 "mmap region %u failed.\n", i);
566 reg->mmap_addr = mmap_addr;
567 reg->mmap_size = mmap_size;
568 reg->host_user_addr = (uint64_t)(uintptr_t)mmap_addr +
571 if (dev->dequeue_zero_copy)
572 add_guest_pages(dev, reg, alignment);
574 RTE_LOG(INFO, VHOST_CONFIG,
575 "guest memory region %u, size: 0x%" PRIx64 "\n"
576 "\t guest physical addr: 0x%" PRIx64 "\n"
577 "\t guest virtual addr: 0x%" PRIx64 "\n"
578 "\t host virtual addr: 0x%" PRIx64 "\n"
579 "\t mmap addr : 0x%" PRIx64 "\n"
580 "\t mmap size : 0x%" PRIx64 "\n"
581 "\t mmap align: 0x%" PRIx64 "\n"
582 "\t mmap off : 0x%" PRIx64 "\n",
584 reg->guest_phys_addr,
585 reg->guest_user_addr,
587 (uint64_t)(uintptr_t)mmap_addr,
593 dump_guest_pages(dev);
598 free_mem_region(dev);
605 vq_is_ready(struct vhost_virtqueue *vq)
607 return vq && vq->desc &&
608 vq->kickfd != VIRTIO_UNINITIALIZED_EVENTFD &&
609 vq->callfd != VIRTIO_UNINITIALIZED_EVENTFD;
613 virtio_is_ready(struct virtio_net *dev)
615 struct vhost_virtqueue *vq;
618 for (i = 0; i < dev->nr_vring; i++) {
619 vq = dev->virtqueue[i];
621 if (!vq_is_ready(vq)) {
622 RTE_LOG(INFO, VHOST_CONFIG,
623 "virtio is not ready for processing.\n");
628 RTE_LOG(INFO, VHOST_CONFIG,
629 "virtio is now ready for processing.\n");
634 vhost_user_set_vring_call(struct virtio_net *dev, struct VhostUserMsg *pmsg)
636 struct vhost_vring_file file;
637 struct vhost_virtqueue *vq;
639 file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
640 if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
641 file.fd = VIRTIO_INVALID_EVENTFD;
643 file.fd = pmsg->fds[0];
644 RTE_LOG(INFO, VHOST_CONFIG,
645 "vring call idx:%d file:%d\n", file.index, file.fd);
647 vq = dev->virtqueue[file.index];
651 vq->callfd = file.fd;
655 * In vhost-user, when we receive kick message, will test whether virtio
656 * device is ready for packet processing.
659 vhost_user_set_vring_kick(struct virtio_net *dev, struct VhostUserMsg *pmsg)
661 struct vhost_vring_file file;
662 struct vhost_virtqueue *vq;
664 file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
665 if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
666 file.fd = VIRTIO_INVALID_EVENTFD;
668 file.fd = pmsg->fds[0];
669 RTE_LOG(INFO, VHOST_CONFIG,
670 "vring kick idx:%d file:%d\n", file.index, file.fd);
672 vq = dev->virtqueue[file.index];
675 vq->kickfd = file.fd;
677 if (virtio_is_ready(dev)) {
678 dev->flags |= VIRTIO_DEV_READY;
680 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
681 if (dev->dequeue_zero_copy) {
682 RTE_LOG(INFO, VHOST_CONFIG,
683 "dequeue zero copy is enabled\n");
686 if (dev->notify_ops->new_device(dev->vid) == 0)
687 dev->flags |= VIRTIO_DEV_RUNNING;
693 free_zmbufs(struct vhost_virtqueue *vq)
695 struct zcopy_mbuf *zmbuf, *next;
697 for (zmbuf = TAILQ_FIRST(&vq->zmbuf_list);
698 zmbuf != NULL; zmbuf = next) {
699 next = TAILQ_NEXT(zmbuf, next);
701 rte_pktmbuf_free(zmbuf->mbuf);
702 TAILQ_REMOVE(&vq->zmbuf_list, zmbuf, next);
705 rte_free(vq->zmbufs);
709 * when virtio is stopped, qemu will send us the GET_VRING_BASE message.
712 vhost_user_get_vring_base(struct virtio_net *dev,
713 struct vhost_vring_state *state)
715 struct vhost_virtqueue *vq = dev->virtqueue[state->index];
717 /* We have to stop the queue (virtio) if it is running. */
718 if (dev->flags & VIRTIO_DEV_RUNNING) {
719 dev->flags &= ~VIRTIO_DEV_RUNNING;
720 dev->notify_ops->destroy_device(dev->vid);
723 dev->flags &= ~VIRTIO_DEV_READY;
725 /* Here we are safe to get the last used index */
726 state->num = vq->last_used_idx;
728 RTE_LOG(INFO, VHOST_CONFIG,
729 "vring base idx:%d file:%d\n", state->index, state->num);
731 * Based on current qemu vhost-user implementation, this message is
732 * sent and only sent in vhost_vring_stop.
733 * TODO: cleanup the vring, it isn't usable since here.
738 vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
740 if (dev->dequeue_zero_copy)
742 rte_free(vq->shadow_used_ring);
743 vq->shadow_used_ring = NULL;
749 * when virtio queues are ready to work, qemu will send us to
750 * enable the virtio queue pair.
753 vhost_user_set_vring_enable(struct virtio_net *dev,
754 struct vhost_vring_state *state)
756 int enable = (int)state->num;
758 RTE_LOG(INFO, VHOST_CONFIG,
759 "set queue enable: %d to qp idx: %d\n",
760 enable, state->index);
762 if (dev->notify_ops->vring_state_changed)
763 dev->notify_ops->vring_state_changed(dev->vid, state->index, enable);
765 dev->virtqueue[state->index]->enabled = enable;
771 vhost_user_set_protocol_features(struct virtio_net *dev,
772 uint64_t protocol_features)
774 if (protocol_features & ~VHOST_USER_PROTOCOL_FEATURES)
777 dev->protocol_features = protocol_features;
781 vhost_user_set_log_base(struct virtio_net *dev, struct VhostUserMsg *msg)
783 int fd = msg->fds[0];
788 RTE_LOG(ERR, VHOST_CONFIG, "invalid log fd: %d\n", fd);
792 if (msg->size != sizeof(VhostUserLog)) {
793 RTE_LOG(ERR, VHOST_CONFIG,
794 "invalid log base msg size: %"PRId32" != %d\n",
795 msg->size, (int)sizeof(VhostUserLog));
799 size = msg->payload.log.mmap_size;
800 off = msg->payload.log.mmap_offset;
801 RTE_LOG(INFO, VHOST_CONFIG,
802 "log mmap size: %"PRId64", offset: %"PRId64"\n",
806 * mmap from 0 to workaround a hugepage mmap bug: mmap will
807 * fail when offset is not page size aligned.
809 addr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
811 if (addr == MAP_FAILED) {
812 RTE_LOG(ERR, VHOST_CONFIG, "mmap log base failed!\n");
817 * Free previously mapped log memory on occasionally
818 * multiple VHOST_USER_SET_LOG_BASE.
821 munmap((void *)(uintptr_t)dev->log_addr, dev->log_size);
823 dev->log_addr = (uint64_t)(uintptr_t)addr;
824 dev->log_base = dev->log_addr + off;
825 dev->log_size = size;
831 * An rarp packet is constructed and broadcasted to notify switches about
832 * the new location of the migrated VM, so that packets from outside will
833 * not be lost after migration.
835 * However, we don't actually "send" a rarp packet here, instead, we set
836 * a flag 'broadcast_rarp' to let rte_vhost_dequeue_burst() inject it.
839 vhost_user_send_rarp(struct virtio_net *dev, struct VhostUserMsg *msg)
841 uint8_t *mac = (uint8_t *)&msg->payload.u64;
843 RTE_LOG(DEBUG, VHOST_CONFIG,
844 ":: mac: %02x:%02x:%02x:%02x:%02x:%02x\n",
845 mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]);
846 memcpy(dev->mac.addr_bytes, mac, 6);
849 * Set the flag to inject a RARP broadcast packet at
850 * rte_vhost_dequeue_burst().
852 * rte_smp_wmb() is for making sure the mac is copied
853 * before the flag is set.
856 rte_atomic16_set(&dev->broadcast_rarp, 1);
862 vhost_user_net_set_mtu(struct virtio_net *dev, struct VhostUserMsg *msg)
864 if (msg->payload.u64 < VIRTIO_MIN_MTU ||
865 msg->payload.u64 > VIRTIO_MAX_MTU) {
866 RTE_LOG(ERR, VHOST_CONFIG, "Invalid MTU size (%"PRIu64")\n",
872 dev->mtu = msg->payload.u64;
877 /* return bytes# of read on success or negative val on failure. */
879 read_vhost_message(int sockfd, struct VhostUserMsg *msg)
883 ret = read_fd_message(sockfd, (char *)msg, VHOST_USER_HDR_SIZE,
884 msg->fds, VHOST_MEMORY_MAX_NREGIONS);
888 if (msg && msg->size) {
889 if (msg->size > sizeof(msg->payload)) {
890 RTE_LOG(ERR, VHOST_CONFIG,
891 "invalid msg size: %d\n", msg->size);
894 ret = read(sockfd, &msg->payload, msg->size);
897 if (ret != (int)msg->size) {
898 RTE_LOG(ERR, VHOST_CONFIG,
899 "read control message failed\n");
908 send_vhost_message(int sockfd, struct VhostUserMsg *msg)
915 msg->flags &= ~VHOST_USER_VERSION_MASK;
916 msg->flags &= ~VHOST_USER_NEED_REPLY;
917 msg->flags |= VHOST_USER_VERSION;
918 msg->flags |= VHOST_USER_REPLY_MASK;
920 ret = send_fd_message(sockfd, (char *)msg,
921 VHOST_USER_HDR_SIZE + msg->size, NULL, 0);
927 * Allocate a queue pair if it hasn't been allocated yet
930 vhost_user_check_and_alloc_queue_pair(struct virtio_net *dev, VhostUserMsg *msg)
934 switch (msg->request) {
935 case VHOST_USER_SET_VRING_KICK:
936 case VHOST_USER_SET_VRING_CALL:
937 case VHOST_USER_SET_VRING_ERR:
938 vring_idx = msg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
940 case VHOST_USER_SET_VRING_NUM:
941 case VHOST_USER_SET_VRING_BASE:
942 case VHOST_USER_SET_VRING_ENABLE:
943 vring_idx = msg->payload.state.index;
945 case VHOST_USER_SET_VRING_ADDR:
946 vring_idx = msg->payload.addr.index;
952 if (vring_idx >= VHOST_MAX_VRING) {
953 RTE_LOG(ERR, VHOST_CONFIG,
954 "invalid vring index: %u\n", vring_idx);
958 if (dev->virtqueue[vring_idx])
961 return alloc_vring_queue(dev, vring_idx);
965 vhost_user_msg_handler(int vid, int fd)
967 struct virtio_net *dev;
968 struct VhostUserMsg msg;
971 dev = get_device(vid);
975 if (!dev->notify_ops) {
976 dev->notify_ops = vhost_driver_callback_get(dev->ifname);
977 if (!dev->notify_ops) {
978 RTE_LOG(ERR, VHOST_CONFIG,
979 "failed to get callback ops for driver %s\n",
985 ret = read_vhost_message(fd, &msg);
986 if (ret <= 0 || msg.request >= VHOST_USER_MAX) {
988 RTE_LOG(ERR, VHOST_CONFIG,
989 "vhost read message failed\n");
991 RTE_LOG(INFO, VHOST_CONFIG,
992 "vhost peer closed\n");
994 RTE_LOG(ERR, VHOST_CONFIG,
995 "vhost read incorrect message\n");
1001 RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
1002 vhost_message_str[msg.request]);
1004 ret = vhost_user_check_and_alloc_queue_pair(dev, &msg);
1006 RTE_LOG(ERR, VHOST_CONFIG,
1007 "failed to alloc queue\n");
1011 switch (msg.request) {
1012 case VHOST_USER_GET_FEATURES:
1013 msg.payload.u64 = vhost_user_get_features(dev);
1014 msg.size = sizeof(msg.payload.u64);
1015 send_vhost_message(fd, &msg);
1017 case VHOST_USER_SET_FEATURES:
1018 vhost_user_set_features(dev, msg.payload.u64);
1021 case VHOST_USER_GET_PROTOCOL_FEATURES:
1022 msg.payload.u64 = VHOST_USER_PROTOCOL_FEATURES;
1023 msg.size = sizeof(msg.payload.u64);
1024 send_vhost_message(fd, &msg);
1026 case VHOST_USER_SET_PROTOCOL_FEATURES:
1027 vhost_user_set_protocol_features(dev, msg.payload.u64);
1030 case VHOST_USER_SET_OWNER:
1031 vhost_user_set_owner();
1033 case VHOST_USER_RESET_OWNER:
1034 vhost_user_reset_owner(dev);
1037 case VHOST_USER_SET_MEM_TABLE:
1038 ret = vhost_user_set_mem_table(dev, &msg);
1041 case VHOST_USER_SET_LOG_BASE:
1042 vhost_user_set_log_base(dev, &msg);
1044 /* it needs a reply */
1045 msg.size = sizeof(msg.payload.u64);
1046 send_vhost_message(fd, &msg);
1048 case VHOST_USER_SET_LOG_FD:
1050 RTE_LOG(INFO, VHOST_CONFIG, "not implemented.\n");
1053 case VHOST_USER_SET_VRING_NUM:
1054 vhost_user_set_vring_num(dev, &msg.payload.state);
1056 case VHOST_USER_SET_VRING_ADDR:
1057 vhost_user_set_vring_addr(dev, &msg.payload.addr);
1059 case VHOST_USER_SET_VRING_BASE:
1060 vhost_user_set_vring_base(dev, &msg.payload.state);
1063 case VHOST_USER_GET_VRING_BASE:
1064 vhost_user_get_vring_base(dev, &msg.payload.state);
1065 msg.size = sizeof(msg.payload.state);
1066 send_vhost_message(fd, &msg);
1069 case VHOST_USER_SET_VRING_KICK:
1070 vhost_user_set_vring_kick(dev, &msg);
1072 case VHOST_USER_SET_VRING_CALL:
1073 vhost_user_set_vring_call(dev, &msg);
1076 case VHOST_USER_SET_VRING_ERR:
1077 if (!(msg.payload.u64 & VHOST_USER_VRING_NOFD_MASK))
1079 RTE_LOG(INFO, VHOST_CONFIG, "not implemented\n");
1082 case VHOST_USER_GET_QUEUE_NUM:
1083 msg.payload.u64 = VHOST_MAX_QUEUE_PAIRS;
1084 msg.size = sizeof(msg.payload.u64);
1085 send_vhost_message(fd, &msg);
1088 case VHOST_USER_SET_VRING_ENABLE:
1089 vhost_user_set_vring_enable(dev, &msg.payload.state);
1091 case VHOST_USER_SEND_RARP:
1092 vhost_user_send_rarp(dev, &msg);
1095 case VHOST_USER_NET_SET_MTU:
1096 ret = vhost_user_net_set_mtu(dev, &msg);
1105 if (msg.flags & VHOST_USER_NEED_REPLY) {
1106 msg.payload.u64 = !!ret;
1107 msg.size = sizeof(msg.payload.u64);
1108 send_vhost_message(fd, &msg);