-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2016 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2016 Intel Corporation
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
+#include <errno.h>
#include <rte_memory.h>
-#include <rte_eal_memconfig.h>
#include "vhost.h"
#include "virtio_user_dev.h"
#include "vhost_kernel_tap.h"
+struct vhost_kernel_data {
+ int *vhostfds;
+ int *tapfds;
+};
+
struct vhost_memory_kernel {
uint32_t nregions;
uint32_t padding;
#define VHOST_SET_VRING_ERR _IOW(VHOST_VIRTIO, 0x22, struct vhost_vring_file)
#define VHOST_NET_SET_BACKEND _IOW(VHOST_VIRTIO, 0x30, struct vhost_vring_file)
+/* with below features, vhost kernel does not need to do the checksum and TSO,
+ * these info will be passed to virtio_user through virtio net header.
+ */
+#define VHOST_KERNEL_GUEST_OFFLOADS_MASK \
+ ((1ULL << VIRTIO_NET_F_GUEST_CSUM) | \
+ (1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
+ (1ULL << VIRTIO_NET_F_GUEST_TSO6) | \
+ (1ULL << VIRTIO_NET_F_GUEST_ECN) | \
+ (1ULL << VIRTIO_NET_F_GUEST_UFO))
+
+/* with below features, when flows from virtio_user to vhost kernel
+ * (1) if flows goes up through the kernel networking stack, it does not need
+ * to verify checksum, which can save CPU cycles;
+ * (2) if flows goes through a Linux bridge and outside from an interface
+ * (kernel driver), checksum and TSO will be done by GSO in kernel or even
+ * offloaded into real physical device.
+ */
+#define VHOST_KERNEL_HOST_OFFLOADS_MASK \
+ ((1ULL << VIRTIO_NET_F_HOST_TSO4) | \
+ (1ULL << VIRTIO_NET_F_HOST_TSO6) | \
+ (1ULL << VIRTIO_NET_F_CSUM))
+
static uint64_t max_regions = 64;
static void
close(fd);
}
-static uint64_t vhost_req_user_to_kernel[] = {
- [VHOST_USER_SET_OWNER] = VHOST_SET_OWNER,
- [VHOST_USER_RESET_OWNER] = VHOST_RESET_OWNER,
- [VHOST_USER_SET_FEATURES] = VHOST_SET_FEATURES,
- [VHOST_USER_GET_FEATURES] = VHOST_GET_FEATURES,
- [VHOST_USER_SET_VRING_CALL] = VHOST_SET_VRING_CALL,
- [VHOST_USER_SET_VRING_NUM] = VHOST_SET_VRING_NUM,
- [VHOST_USER_SET_VRING_BASE] = VHOST_SET_VRING_BASE,
- [VHOST_USER_GET_VRING_BASE] = VHOST_GET_VRING_BASE,
- [VHOST_USER_SET_VRING_ADDR] = VHOST_SET_VRING_ADDR,
- [VHOST_USER_SET_VRING_KICK] = VHOST_SET_VRING_KICK,
- [VHOST_USER_SET_MEM_TABLE] = VHOST_SET_MEM_TABLE,
-};
+static int
+vhost_kernel_ioctl(int fd, uint64_t request, void *arg)
+{
+ int ret;
-/* By default, vhost kernel module allows 64 regions, but DPDK allows
- * 256 segments. As a relief, below function merges those virtually
- * adjacent memsegs into one region.
- */
-static struct vhost_memory_kernel *
-prepare_vhost_memory_kernel(void)
+ ret = ioctl(fd, request, arg);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "Vhost-kernel ioctl %"PRIu64" failed (%s)",
+ request, strerror(errno));
+ return -1;
+ }
+
+ return 0;
+}
+
+static int
+vhost_kernel_set_owner(struct virtio_user_dev *dev)
+{
+ struct vhost_kernel_data *data = dev->backend_data;
+
+ return vhost_kernel_ioctl(data->vhostfds[0], VHOST_SET_OWNER, NULL);
+}
+
+static int
+vhost_kernel_get_features(struct virtio_user_dev *dev, uint64_t *features)
+{
+ int ret;
+ unsigned int tap_features;
+ struct vhost_kernel_data *data = dev->backend_data;
+
+ ret = vhost_kernel_ioctl(data->vhostfds[0], VHOST_GET_FEATURES, features);
+ if (ret < 0) {
+ PMD_DRV_LOG(ERR, "Failed to get features");
+ return -1;
+ }
+
+ ret = tap_support_features(&tap_features);
+ if (ret < 0) {
+ PMD_DRV_LOG(ERR, "Failed to get TAP features");
+ return -1;
+ }
+
+ /* with tap as the backend, all these features are supported
+ * but not claimed by vhost-net, so we add them back when
+ * reporting to upper layer.
+ */
+ if (tap_features & IFF_VNET_HDR) {
+ *features |= VHOST_KERNEL_GUEST_OFFLOADS_MASK;
+ *features |= VHOST_KERNEL_HOST_OFFLOADS_MASK;
+ }
+
+ /* vhost_kernel will not declare this feature, but it does
+ * support multi-queue.
+ */
+ if (tap_features & IFF_MULTI_QUEUE)
+ *features |= (1ull << VIRTIO_NET_F_MQ);
+
+ return 0;
+}
+
+static int
+vhost_kernel_set_features(struct virtio_user_dev *dev, uint64_t features)
+{
+ struct vhost_kernel_data *data = dev->backend_data;
+
+ /* We don't need memory protection here */
+ features &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+ /* VHOST kernel does not know about below flags */
+ features &= ~VHOST_KERNEL_GUEST_OFFLOADS_MASK;
+ features &= ~VHOST_KERNEL_HOST_OFFLOADS_MASK;
+ features &= ~(1ULL << VIRTIO_NET_F_MQ);
+
+ return vhost_kernel_ioctl(data->vhostfds[0], VHOST_SET_FEATURES, &features);
+}
+
+static int
+add_memseg_list(const struct rte_memseg_list *msl, void *arg)
{
- uint32_t i, j, k = 0;
- struct rte_memseg *seg;
+ struct vhost_memory_kernel *vm = arg;
struct vhost_memory_region *mr;
+ void *start_addr;
+ uint64_t len;
+
+ if (msl->external)
+ return 0;
+
+ if (vm->nregions >= max_regions)
+ return -1;
+
+ start_addr = msl->base_va;
+ len = msl->page_sz * msl->memseg_arr.len;
+
+ mr = &vm->regions[vm->nregions++];
+
+ mr->guest_phys_addr = (uint64_t)(uintptr_t)start_addr;
+ mr->userspace_addr = (uint64_t)(uintptr_t)start_addr;
+ mr->memory_size = len;
+ mr->mmap_offset = 0; /* flags_padding */
+
+ PMD_DRV_LOG(DEBUG, "index=%u addr=%p len=%" PRIu64,
+ vm->nregions - 1, start_addr, len);
+
+ return 0;
+}
+
+/* By default, vhost kernel module allows 64 regions, but DPDK may
+ * have much more memory regions. Below function will treat each
+ * contiguous memory space reserved by DPDK as one region.
+ */
+static int
+vhost_kernel_set_memory_table(struct virtio_user_dev *dev)
+{
+ struct vhost_kernel_data *data = dev->backend_data;
struct vhost_memory_kernel *vm;
+ int ret;
vm = malloc(sizeof(struct vhost_memory_kernel) +
- max_regions *
- sizeof(struct vhost_memory_region));
-
- for (i = 0; i < RTE_MAX_MEMSEG; ++i) {
- seg = &rte_eal_get_configuration()->mem_config->memseg[i];
- if (!seg->addr)
- break;
-
- int new_region = 1;
-
- for (j = 0; j < k; ++j) {
- mr = &vm->regions[j];
-
- if (mr->userspace_addr + mr->memory_size ==
- (uint64_t)(uintptr_t)seg->addr) {
- mr->memory_size += seg->len;
- new_region = 0;
- break;
- }
-
- if ((uint64_t)(uintptr_t)seg->addr + seg->len ==
- mr->userspace_addr) {
- mr->guest_phys_addr =
- (uint64_t)(uintptr_t)seg->addr;
- mr->userspace_addr =
- (uint64_t)(uintptr_t)seg->addr;
- mr->memory_size += seg->len;
- new_region = 0;
- break;
- }
- }
+ max_regions *
+ sizeof(struct vhost_memory_region));
+ if (!vm)
+ goto err;
- if (new_region == 0)
- continue;
+ vm->nregions = 0;
+ vm->padding = 0;
- mr = &vm->regions[k++];
- /* use vaddr here! */
- mr->guest_phys_addr = (uint64_t)(uintptr_t)seg->addr;
- mr->userspace_addr = (uint64_t)(uintptr_t)seg->addr;
- mr->memory_size = seg->len;
- mr->mmap_offset = 0;
+ /*
+ * The memory lock has already been taken by memory subsystem
+ * or virtio_user_start_device().
+ */
+ ret = rte_memseg_list_walk_thread_unsafe(add_memseg_list, vm);
+ if (ret < 0)
+ goto err_free;
- if (k >= max_regions) {
- free(vm);
- return NULL;
- }
+ ret = vhost_kernel_ioctl(data->vhostfds[0], VHOST_SET_MEM_TABLE, vm);
+ if (ret < 0)
+ goto err_free;
+
+ free(vm);
+
+ return 0;
+err_free:
+ free(vm);
+err:
+ PMD_DRV_LOG(ERR, "Failed to set memory table");
+ return -1;
+}
+
+static int
+vhost_kernel_set_vring(struct virtio_user_dev *dev, uint64_t req, struct vhost_vring_state *state)
+{
+ int ret, fd;
+ unsigned int index = state->index;
+ struct vhost_kernel_data *data = dev->backend_data;
+
+ /* Convert from queue index to queue-pair & offset */
+ fd = data->vhostfds[state->index / 2];
+ state->index %= 2;
+
+ ret = vhost_kernel_ioctl(fd, req, state);
+ if (ret < 0) {
+ PMD_DRV_LOG(ERR, "Failed to set vring (request %" PRIu64 ")", req);
+ return -1;
}
- vm->nregions = k;
- vm->padding = 0;
- return vm;
+ /* restore index back to queue index */
+ state->index = index;
+
+ return 0;
+}
+
+static int
+vhost_kernel_set_vring_num(struct virtio_user_dev *dev, struct vhost_vring_state *state)
+{
+ return vhost_kernel_set_vring(dev, VHOST_SET_VRING_NUM, state);
+}
+
+static int
+vhost_kernel_set_vring_base(struct virtio_user_dev *dev, struct vhost_vring_state *state)
+{
+ return vhost_kernel_set_vring(dev, VHOST_SET_VRING_BASE, state);
}
static int
-vhost_kernel_ioctl(struct virtio_user_dev *dev,
- enum vhost_user_request req,
- void *arg)
+vhost_kernel_get_vring_base(struct virtio_user_dev *dev, struct vhost_vring_state *state)
{
- int ret = -1;
- unsigned int i;
- uint64_t req_kernel;
- struct vhost_memory_kernel *vm = NULL;
+ return vhost_kernel_set_vring(dev, VHOST_GET_VRING_BASE, state);
+}
- PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
+static int
+vhost_kernel_set_vring_file(struct virtio_user_dev *dev, uint64_t req,
+ struct vhost_vring_file *file)
+{
+ int ret, fd;
+ unsigned int index = file->index;
+ struct vhost_kernel_data *data = dev->backend_data;
- req_kernel = vhost_req_user_to_kernel[req];
+ /* Convert from queue index to queue-pair & offset */
+ fd = data->vhostfds[file->index / 2];
+ file->index %= 2;
- if (req_kernel == VHOST_SET_MEM_TABLE) {
- vm = prepare_vhost_memory_kernel();
- if (!vm)
- return -1;
- arg = (void *)vm;
+ ret = vhost_kernel_ioctl(fd, req, file);
+ if (ret < 0) {
+ PMD_DRV_LOG(ERR, "Failed to set vring file (request %" PRIu64 ")", req);
+ return -1;
}
- /* We don't need memory protection here */
- if (req_kernel == VHOST_SET_FEATURES)
- *(uint64_t *)arg &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+ /* restore index back to queue index */
+ file->index = index;
- for (i = 0; i < dev->max_queue_pairs; ++i) {
- if (dev->vhostfds[i] < 0)
- continue;
+ return 0;
+}
+
+static int
+vhost_kernel_set_vring_kick(struct virtio_user_dev *dev, struct vhost_vring_file *file)
+{
+ return vhost_kernel_set_vring_file(dev, VHOST_SET_VRING_KICK, file);
+}
+
+static int
+vhost_kernel_set_vring_call(struct virtio_user_dev *dev, struct vhost_vring_file *file)
+{
+ return vhost_kernel_set_vring_file(dev, VHOST_SET_VRING_CALL, file);
+}
+
+static int
+vhost_kernel_set_vring_addr(struct virtio_user_dev *dev, struct vhost_vring_addr *addr)
+{
+ int ret, fd;
+ unsigned int index = addr->index;
+ struct vhost_kernel_data *data = dev->backend_data;
- ret = ioctl(dev->vhostfds[i], req_kernel, arg);
- if (ret < 0)
- break;
+ /* Convert from queue index to queue-pair & offset */
+ fd = data->vhostfds[addr->index / 2];
+ addr->index %= 2;
+
+ ret = vhost_kernel_ioctl(fd, VHOST_SET_VRING_ADDR, addr);
+ if (ret < 0) {
+ PMD_DRV_LOG(ERR, "Failed to set vring address");
+ return -1;
}
- if (vm)
- free(vm);
+ /* restore index back to queue index */
+ addr->index = index;
- if (ret < 0)
- PMD_DRV_LOG(ERR, "%s failed: %s",
- vhost_msg_strings[req], strerror(errno));
+ return 0;
+}
- return ret;
+static int
+vhost_kernel_get_status(struct virtio_user_dev *dev __rte_unused, uint8_t *status __rte_unused)
+{
+ return -ENOTSUP;
+}
+
+static int
+vhost_kernel_set_status(struct virtio_user_dev *dev __rte_unused, uint8_t status __rte_unused)
+{
+ return -ENOTSUP;
}
/**
vhost_kernel_setup(struct virtio_user_dev *dev)
{
int vhostfd;
- uint32_t i;
+ uint32_t q, i;
+ struct vhost_kernel_data *data;
+
+ data = malloc(sizeof(*data));
+ if (!data) {
+ PMD_INIT_LOG(ERR, "(%s) Failed to allocate Vhost-kernel data", dev->path);
+ return -1;
+ }
+
+ data->vhostfds = malloc(dev->max_queue_pairs * sizeof(int));
+ if (!data->vhostfds) {
+ PMD_INIT_LOG(ERR, "(%s) Failed to allocate Vhost FDs", dev->path);
+ goto err_data;
+ }
+ data->tapfds = malloc(dev->max_queue_pairs * sizeof(int));
+ if (!data->tapfds) {
+ PMD_INIT_LOG(ERR, "(%s) Failed to allocate TAP FDs", dev->path);
+ goto err_vhostfds;
+ }
+
+ for (q = 0; q < dev->max_queue_pairs; ++q) {
+ data->vhostfds[q] = -1;
+ data->tapfds[q] = -1;
+ }
get_vhost_kernel_max_regions();
for (i = 0; i < dev->max_queue_pairs; ++i) {
vhostfd = open(dev->path, O_RDWR);
if (vhostfd < 0) {
- PMD_DRV_LOG(ERR, "fail to open %s, %s",
- dev->path, strerror(errno));
- return -1;
+ PMD_DRV_LOG(ERR, "fail to open %s, %s", dev->path, strerror(errno));
+ goto err_tapfds;
}
- dev->vhostfds[i] = vhostfd;
+ data->vhostfds[i] = vhostfd;
}
+ dev->backend_data = data;
+
+ return 0;
+
+err_tapfds:
+ for (i = 0; i < dev->max_queue_pairs; i++)
+ if (data->vhostfds[i] >= 0)
+ close(data->vhostfds[i]);
+
+ free(data->tapfds);
+err_vhostfds:
+ free(data->vhostfds);
+err_data:
+ free(data);
+
+ return -1;
+}
+
+static int
+vhost_kernel_destroy(struct virtio_user_dev *dev)
+{
+ struct vhost_kernel_data *data = dev->backend_data;
+ uint32_t i;
+
+ if (!data)
+ return 0;
+
+ for (i = 0; i < dev->max_queue_pairs; ++i) {
+ if (data->vhostfds[i] >= 0)
+ close(data->vhostfds[i]);
+ if (data->tapfds[i] >= 0)
+ close(data->tapfds[i]);
+ }
+
+ free(data->vhostfds);
+ free(data->tapfds);
+ free(data);
+ dev->backend_data = NULL;
+
return 0;
}
int hdr_size;
int vhostfd;
int tapfd;
+ int req_mq = (dev->max_queue_pairs > 1);
+ struct vhost_kernel_data *data = dev->backend_data;
- vhostfd = dev->vhostfds[pair_idx];
+ vhostfd = data->vhostfds[pair_idx];
+
+ if (dev->qp_enabled[pair_idx] == enable)
+ return 0;
if (!enable) {
- if (dev->tapfds[pair_idx]) {
- close(dev->tapfds[pair_idx]);
- dev->tapfds[pair_idx] = -1;
+ tapfd = data->tapfds[pair_idx];
+ if (vhost_kernel_set_backend(vhostfd, -1) < 0) {
+ PMD_DRV_LOG(ERR, "fail to set backend for vhost kernel");
+ return -1;
+ }
+ if (req_mq && vhost_kernel_tap_set_queue(tapfd, false) < 0) {
+ PMD_DRV_LOG(ERR, "fail to disable tap for vhost kernel");
+ return -1;
}
- return vhost_kernel_set_backend(vhostfd, -1);
- } else if (dev->tapfds[pair_idx] >= 0) {
+ dev->qp_enabled[pair_idx] = false;
return 0;
}
+ if (data->tapfds[pair_idx] >= 0) {
+ tapfd = data->tapfds[pair_idx];
+ if (vhost_kernel_tap_set_offload(tapfd, dev->features) == -1)
+ return -1;
+ if (req_mq && vhost_kernel_tap_set_queue(tapfd, true) < 0) {
+ PMD_DRV_LOG(ERR, "fail to enable tap for vhost kernel");
+ return -1;
+ }
+ goto set_backend;
+ }
+
if ((dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF)) ||
(dev->features & (1ULL << VIRTIO_F_VERSION_1)))
hdr_size = sizeof(struct virtio_net_hdr_mrg_rxbuf);
else
hdr_size = sizeof(struct virtio_net_hdr);
- tapfd = vhost_kernel_open_tap(&dev->ifname, hdr_size);
+ tapfd = vhost_kernel_open_tap(&dev->ifname, hdr_size, req_mq,
+ (char *)dev->mac_addr, dev->features);
if (tapfd < 0) {
PMD_DRV_LOG(ERR, "fail to open tap for vhost kernel");
return -1;
}
+ data->tapfds[pair_idx] = tapfd;
+
+set_backend:
if (vhost_kernel_set_backend(vhostfd, tapfd) < 0) {
PMD_DRV_LOG(ERR, "fail to set backend for vhost kernel");
- close(tapfd);
return -1;
}
- dev->tapfds[pair_idx] = tapfd;
+ dev->qp_enabled[pair_idx] = true;
return 0;
}
-struct virtio_user_backend_ops ops_kernel = {
+static int
+vhost_kernel_get_backend_features(uint64_t *features)
+{
+ *features = 0;
+
+ return 0;
+}
+
+static int
+vhost_kernel_update_link_state(struct virtio_user_dev *dev __rte_unused)
+{
+ /* Nothing to update (Maybe get TAP interface link state?) */
+ return 0;
+}
+
+static int
+vhost_kernel_get_intr_fd(struct virtio_user_dev *dev __rte_unused)
+{
+ /* No link state interrupt with Vhost-kernel */
+ return -1;
+}
+
+struct virtio_user_backend_ops virtio_ops_kernel = {
.setup = vhost_kernel_setup,
- .send_request = vhost_kernel_ioctl,
- .enable_qp = vhost_kernel_enable_queue_pair
+ .destroy = vhost_kernel_destroy,
+ .get_backend_features = vhost_kernel_get_backend_features,
+ .set_owner = vhost_kernel_set_owner,
+ .get_features = vhost_kernel_get_features,
+ .set_features = vhost_kernel_set_features,
+ .set_memory_table = vhost_kernel_set_memory_table,
+ .set_vring_num = vhost_kernel_set_vring_num,
+ .set_vring_base = vhost_kernel_set_vring_base,
+ .get_vring_base = vhost_kernel_get_vring_base,
+ .set_vring_call = vhost_kernel_set_vring_call,
+ .set_vring_kick = vhost_kernel_set_vring_kick,
+ .set_vring_addr = vhost_kernel_set_vring_addr,
+ .get_status = vhost_kernel_get_status,
+ .set_status = vhost_kernel_set_status,
+ .enable_qp = vhost_kernel_enable_queue_pair,
+ .update_link_state = vhost_kernel_update_link_state,
+ .get_intr_fd = vhost_kernel_get_intr_fd,
};