4 * Copyright(c) 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.
34 #include <sys/types.h>
39 #include <rte_memory.h>
40 #include <rte_eal_memconfig.h>
43 #include "virtio_user_dev.h"
44 #include "vhost_kernel_tap.h"
46 struct vhost_memory_kernel {
49 struct vhost_memory_region regions[0];
52 /* vhost kernel ioctls */
53 #define VHOST_VIRTIO 0xAF
54 #define VHOST_GET_FEATURES _IOR(VHOST_VIRTIO, 0x00, __u64)
55 #define VHOST_SET_FEATURES _IOW(VHOST_VIRTIO, 0x00, __u64)
56 #define VHOST_SET_OWNER _IO(VHOST_VIRTIO, 0x01)
57 #define VHOST_RESET_OWNER _IO(VHOST_VIRTIO, 0x02)
58 #define VHOST_SET_MEM_TABLE _IOW(VHOST_VIRTIO, 0x03, struct vhost_memory_kernel)
59 #define VHOST_SET_LOG_BASE _IOW(VHOST_VIRTIO, 0x04, __u64)
60 #define VHOST_SET_LOG_FD _IOW(VHOST_VIRTIO, 0x07, int)
61 #define VHOST_SET_VRING_NUM _IOW(VHOST_VIRTIO, 0x10, struct vhost_vring_state)
62 #define VHOST_SET_VRING_ADDR _IOW(VHOST_VIRTIO, 0x11, struct vhost_vring_addr)
63 #define VHOST_SET_VRING_BASE _IOW(VHOST_VIRTIO, 0x12, struct vhost_vring_state)
64 #define VHOST_GET_VRING_BASE _IOWR(VHOST_VIRTIO, 0x12, struct vhost_vring_state)
65 #define VHOST_SET_VRING_KICK _IOW(VHOST_VIRTIO, 0x20, struct vhost_vring_file)
66 #define VHOST_SET_VRING_CALL _IOW(VHOST_VIRTIO, 0x21, struct vhost_vring_file)
67 #define VHOST_SET_VRING_ERR _IOW(VHOST_VIRTIO, 0x22, struct vhost_vring_file)
68 #define VHOST_NET_SET_BACKEND _IOW(VHOST_VIRTIO, 0x30, struct vhost_vring_file)
70 static uint64_t max_regions = 64;
73 get_vhost_kernel_max_regions(void)
76 char buf[20] = {'\0'};
78 fd = open("/sys/module/vhost/parameters/max_mem_regions", O_RDONLY);
82 if (read(fd, buf, sizeof(buf) - 1) > 0)
83 max_regions = strtoull(buf, NULL, 10);
88 static uint64_t vhost_req_user_to_kernel[] = {
89 [VHOST_USER_SET_OWNER] = VHOST_SET_OWNER,
90 [VHOST_USER_RESET_OWNER] = VHOST_RESET_OWNER,
91 [VHOST_USER_SET_FEATURES] = VHOST_SET_FEATURES,
92 [VHOST_USER_GET_FEATURES] = VHOST_GET_FEATURES,
93 [VHOST_USER_SET_VRING_CALL] = VHOST_SET_VRING_CALL,
94 [VHOST_USER_SET_VRING_NUM] = VHOST_SET_VRING_NUM,
95 [VHOST_USER_SET_VRING_BASE] = VHOST_SET_VRING_BASE,
96 [VHOST_USER_GET_VRING_BASE] = VHOST_GET_VRING_BASE,
97 [VHOST_USER_SET_VRING_ADDR] = VHOST_SET_VRING_ADDR,
98 [VHOST_USER_SET_VRING_KICK] = VHOST_SET_VRING_KICK,
99 [VHOST_USER_SET_MEM_TABLE] = VHOST_SET_MEM_TABLE,
102 /* By default, vhost kernel module allows 64 regions, but DPDK allows
103 * 256 segments. As a relief, below function merges those virtually
104 * adjacent memsegs into one region.
106 static struct vhost_memory_kernel *
107 prepare_vhost_memory_kernel(void)
109 uint32_t i, j, k = 0;
110 struct rte_memseg *seg;
111 struct vhost_memory_region *mr;
112 struct vhost_memory_kernel *vm;
114 vm = malloc(sizeof(struct vhost_memory_kernel) +
116 sizeof(struct vhost_memory_region));
120 for (i = 0; i < RTE_MAX_MEMSEG; ++i) {
121 seg = &rte_eal_get_configuration()->mem_config->memseg[i];
127 for (j = 0; j < k; ++j) {
128 mr = &vm->regions[j];
130 if (mr->userspace_addr + mr->memory_size ==
131 (uint64_t)(uintptr_t)seg->addr) {
132 mr->memory_size += seg->len;
137 if ((uint64_t)(uintptr_t)seg->addr + seg->len ==
138 mr->userspace_addr) {
139 mr->guest_phys_addr =
140 (uint64_t)(uintptr_t)seg->addr;
142 (uint64_t)(uintptr_t)seg->addr;
143 mr->memory_size += seg->len;
152 mr = &vm->regions[k++];
153 /* use vaddr here! */
154 mr->guest_phys_addr = (uint64_t)(uintptr_t)seg->addr;
155 mr->userspace_addr = (uint64_t)(uintptr_t)seg->addr;
156 mr->memory_size = seg->len;
159 if (k >= max_regions) {
170 /* with below features, vhost kernel does not need to do the checksum and TSO,
171 * these info will be passed to virtio_user through virtio net header.
173 #define VHOST_KERNEL_GUEST_OFFLOADS_MASK \
174 ((1ULL << VIRTIO_NET_F_GUEST_CSUM) | \
175 (1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
176 (1ULL << VIRTIO_NET_F_GUEST_TSO6) | \
177 (1ULL << VIRTIO_NET_F_GUEST_ECN) | \
178 (1ULL << VIRTIO_NET_F_GUEST_UFO))
180 /* with below features, when flows from virtio_user to vhost kernel
181 * (1) if flows goes up through the kernel networking stack, it does not need
182 * to verify checksum, which can save CPU cycles;
183 * (2) if flows goes through a Linux bridge and outside from an interface
184 * (kernel driver), checksum and TSO will be done by GSO in kernel or even
185 * offloaded into real physical device.
187 #define VHOST_KERNEL_HOST_OFFLOADS_MASK \
188 ((1ULL << VIRTIO_NET_F_HOST_TSO4) | \
189 (1ULL << VIRTIO_NET_F_HOST_TSO6) | \
190 (1ULL << VIRTIO_NET_F_CSUM))
193 tap_supporte_mq(void)
196 unsigned int tap_features;
198 tapfd = open(PATH_NET_TUN, O_RDWR);
200 PMD_DRV_LOG(ERR, "fail to open %s: %s",
201 PATH_NET_TUN, strerror(errno));
205 if (ioctl(tapfd, TUNGETFEATURES, &tap_features) == -1) {
206 PMD_DRV_LOG(ERR, "TUNGETFEATURES failed: %s", strerror(errno));
212 return tap_features & IFF_MULTI_QUEUE;
216 vhost_kernel_ioctl(struct virtio_user_dev *dev,
217 enum vhost_user_request req,
223 struct vhost_memory_kernel *vm = NULL;
225 unsigned int queue_sel;
227 PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
229 req_kernel = vhost_req_user_to_kernel[req];
231 if (req_kernel == VHOST_SET_MEM_TABLE) {
232 vm = prepare_vhost_memory_kernel();
238 if (req_kernel == VHOST_SET_FEATURES) {
239 /* We don't need memory protection here */
240 *(uint64_t *)arg &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
242 /* VHOST kernel does not know about below flags */
243 *(uint64_t *)arg &= ~VHOST_KERNEL_GUEST_OFFLOADS_MASK;
244 *(uint64_t *)arg &= ~VHOST_KERNEL_HOST_OFFLOADS_MASK;
246 *(uint64_t *)arg &= ~(1ULL << VIRTIO_NET_F_MQ);
249 switch (req_kernel) {
250 case VHOST_SET_VRING_NUM:
251 case VHOST_SET_VRING_ADDR:
252 case VHOST_SET_VRING_BASE:
253 case VHOST_GET_VRING_BASE:
254 case VHOST_SET_VRING_KICK:
255 case VHOST_SET_VRING_CALL:
256 queue_sel = *(unsigned int *)arg;
257 vhostfd = dev->vhostfds[queue_sel / 2];
258 *(unsigned int *)arg = queue_sel % 2;
259 PMD_DRV_LOG(DEBUG, "vhostfd=%d, index=%u",
260 vhostfd, *(unsigned int *)arg);
266 for (i = 0; i < dev->max_queue_pairs; ++i) {
267 if (dev->vhostfds[i] < 0)
270 ret = ioctl(dev->vhostfds[i], req_kernel, arg);
275 ret = ioctl(vhostfd, req_kernel, arg);
278 if (!ret && req_kernel == VHOST_GET_FEATURES) {
279 /* with tap as the backend, all these features are supported
280 * but not claimed by vhost-net, so we add them back when
281 * reporting to upper layer.
283 *((uint64_t *)arg) |= VHOST_KERNEL_GUEST_OFFLOADS_MASK;
284 *((uint64_t *)arg) |= VHOST_KERNEL_HOST_OFFLOADS_MASK;
286 /* vhost_kernel will not declare this feature, but it does
287 * support multi-queue.
289 if (tap_supporte_mq())
290 *(uint64_t *)arg |= (1ull << VIRTIO_NET_F_MQ);
297 PMD_DRV_LOG(ERR, "%s failed: %s",
298 vhost_msg_strings[req], strerror(errno));
304 * Set up environment to talk with a vhost kernel backend.
307 * - (-1) if fail to set up;
308 * - (>=0) if successful.
311 vhost_kernel_setup(struct virtio_user_dev *dev)
316 get_vhost_kernel_max_regions();
318 for (i = 0; i < dev->max_queue_pairs; ++i) {
319 vhostfd = open(dev->path, O_RDWR);
321 PMD_DRV_LOG(ERR, "fail to open %s, %s",
322 dev->path, strerror(errno));
326 dev->vhostfds[i] = vhostfd;
333 vhost_kernel_set_backend(int vhostfd, int tapfd)
335 struct vhost_vring_file f;
339 if (ioctl(vhostfd, VHOST_NET_SET_BACKEND, &f) < 0) {
340 PMD_DRV_LOG(ERR, "VHOST_NET_SET_BACKEND fails, %s",
346 if (ioctl(vhostfd, VHOST_NET_SET_BACKEND, &f) < 0) {
347 PMD_DRV_LOG(ERR, "VHOST_NET_SET_BACKEND fails, %s",
356 vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
363 int req_mq = (dev->max_queue_pairs > 1);
365 vhostfd = dev->vhostfds[pair_idx];
368 if (dev->tapfds[pair_idx] >= 0) {
369 close(dev->tapfds[pair_idx]);
370 dev->tapfds[pair_idx] = -1;
372 return vhost_kernel_set_backend(vhostfd, -1);
373 } else if (dev->tapfds[pair_idx] >= 0) {
377 if ((dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF)) ||
378 (dev->features & (1ULL << VIRTIO_F_VERSION_1)))
379 hdr_size = sizeof(struct virtio_net_hdr_mrg_rxbuf);
381 hdr_size = sizeof(struct virtio_net_hdr);
383 tapfd = vhost_kernel_open_tap(&dev->ifname, hdr_size, req_mq);
385 PMD_DRV_LOG(ERR, "fail to open tap for vhost kernel");
389 if (vhost_kernel_set_backend(vhostfd, tapfd) < 0) {
390 PMD_DRV_LOG(ERR, "fail to set backend for vhost kernel");
395 dev->tapfds[pair_idx] = tapfd;
399 struct virtio_user_backend_ops ops_kernel = {
400 .setup = vhost_kernel_setup,
401 .send_request = vhost_kernel_ioctl,
402 .enable_qp = vhost_kernel_enable_queue_pair