4 * Copyright(c) 2010-2017 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.
39 * Interface to vhost-user
43 #include <sys/eventfd.h>
45 #include <rte_memory.h>
46 #include <rte_mempool.h>
52 /* These are not C++-aware. */
53 #include <linux/vhost.h>
54 #include <linux/virtio_ring.h>
56 #define RTE_VHOST_USER_CLIENT (1ULL << 0)
57 #define RTE_VHOST_USER_NO_RECONNECT (1ULL << 1)
58 #define RTE_VHOST_USER_DEQUEUE_ZERO_COPY (1ULL << 2)
61 * Information relating to memory regions including offsets to
62 * addresses in QEMUs memory file.
64 struct rte_vhost_mem_region {
65 uint64_t guest_phys_addr;
66 uint64_t guest_user_addr;
67 uint64_t host_user_addr;
75 * Memory structure includes region and mapping information.
77 struct rte_vhost_memory {
79 struct rte_vhost_mem_region regions[];
82 struct rte_vhost_vring {
83 struct vring_desc *desc;
84 struct vring_avail *avail;
85 struct vring_used *used;
86 uint64_t log_guest_addr;
94 * Device and vring operations.
96 struct vhost_device_ops {
97 int (*new_device)(int vid); /**< Add device. */
98 void (*destroy_device)(int vid); /**< Remove device. */
100 int (*vring_state_changed)(int vid, uint16_t queue_id, int enable); /**< triggered when a vring is enabled or disabled */
103 * Features could be changed after the feature negotiation.
104 * For example, VHOST_F_LOG_ALL will be set/cleared at the
105 * start/end of live migration, respectively. This callback
106 * is used to inform the application on such change.
108 int (*features_changed)(int vid, uint64_t features);
110 void *reserved[4]; /**< Reserved for future extension */
114 * Convert guest physical address to host virtual address
117 * the guest memory regions
119 * the guest physical address for querying
121 * the host virtual address on success, 0 on failure
123 static __rte_always_inline uint64_t
124 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
126 struct rte_vhost_mem_region *reg;
129 for (i = 0; i < mem->nregions; i++) {
130 reg = &mem->regions[i];
131 if (gpa >= reg->guest_phys_addr &&
132 gpa < reg->guest_phys_addr + reg->size) {
133 return gpa - reg->guest_phys_addr +
141 #define RTE_VHOST_NEED_LOG(features) ((features) & (1ULL << VHOST_F_LOG_ALL))
144 * Log the memory write start with given address.
146 * This function only need be invoked when the live migration starts.
147 * Therefore, we won't need call it at all in the most of time. For
148 * making the performance impact be minimum, it's suggested to do a
149 * check before calling it:
151 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
152 * rte_vhost_log_write(vid, addr, len);
157 * the starting address for write
159 * the length to write
161 void rte_vhost_log_write(int vid, uint64_t addr, uint64_t len);
164 * Log the used ring update start at given offset.
166 * Same as rte_vhost_log_write, it's suggested to do a check before
169 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
170 * rte_vhost_log_used_vring(vid, vring_idx, offset, len);
177 * the offset inside the used ring
179 * the length to write
181 void rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
182 uint64_t offset, uint64_t len);
184 int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
187 * Register vhost driver. path could be different for multiple
190 int rte_vhost_driver_register(const char *path, uint64_t flags);
192 /* Unregister vhost driver. This is only meaningful to vhost user. */
193 int rte_vhost_driver_unregister(const char *path);
196 * Set the feature bits the vhost-user driver supports.
199 * The vhost-user socket file path
203 * 0 on success, -1 on failure
205 int rte_vhost_driver_set_features(const char *path, uint64_t features);
208 * Enable vhost-user driver features.
211 * - the param features should be a subset of the feature bits provided
212 * by rte_vhost_driver_set_features().
213 * - it must be invoked before vhost-user negotiation starts.
216 * The vhost-user socket file path
220 * 0 on success, -1 on failure
222 int rte_vhost_driver_enable_features(const char *path, uint64_t features);
225 * Disable vhost-user driver features.
227 * The two notes at rte_vhost_driver_enable_features() also apply here.
230 * The vhost-user socket file path
232 * Features to disable
234 * 0 on success, -1 on failure
236 int rte_vhost_driver_disable_features(const char *path, uint64_t features);
239 * Get the feature bits before feature negotiation.
242 * The vhost-user socket file path
244 * A pointer to store the queried feature bits
246 * 0 on success, -1 on failure
248 int rte_vhost_driver_get_features(const char *path, uint64_t *features);
251 * Get the feature bits after negotiation
256 * A pointer to store the queried feature bits
258 * 0 on success, -1 on failure
260 int rte_vhost_get_negotiated_features(int vid, uint64_t *features);
262 /* Register callbacks. */
263 int rte_vhost_driver_callback_register(const char *path,
264 struct vhost_device_ops const * const ops);
268 * Start the vhost-user driver.
270 * This function triggers the vhost-user negotiation.
273 * The vhost-user socket file path
275 * 0 on success, -1 on failure
277 int rte_vhost_driver_start(const char *path);
280 * Get the MTU value of the device if set in QEMU.
283 * virtio-net device ID
285 * The variable to store the MTU value
289 * -EAGAIN: device not yet started
290 * -ENOTSUP: device does not support MTU feature
292 int rte_vhost_get_mtu(int vid, uint16_t *mtu);
295 * Get the numa node from which the virtio net device's memory
302 * The numa node, -1 on failure
304 int rte_vhost_get_numa_node(int vid);
308 * Get the number of queues the device supports.
310 * Note this function is deprecated, as it returns a queue pair number,
311 * which is vhost specific. Instead, rte_vhost_get_vring_num should
318 * The number of queues, 0 on failure
321 uint32_t rte_vhost_get_queue_num(int vid);
324 * Get the number of vrings the device supports.
330 * The number of vrings, 0 on failure
332 uint16_t rte_vhost_get_vring_num(int vid);
335 * Get the virtio net device's ifname, which is the vhost-user socket
341 * The buffer to stored the queried ifname
346 * 0 on success, -1 on failure
348 int rte_vhost_get_ifname(int vid, char *buf, size_t len);
351 * Get how many avail entries are left in the queue
359 * num of avail entires left
361 uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
366 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
367 * be received from the physical port or from another virtual device. A packet
368 * count is returned to indicate the number of packets that were successfully
369 * added to the RX queue.
373 * virtio queue index in mq case
375 * array to contain packets to be enqueued
377 * packets num to be enqueued
379 * num of packets enqueued
381 uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
382 struct rte_mbuf **pkts, uint16_t count);
385 * This function gets guest buffers from the virtio device TX virtqueue,
386 * construct host mbufs, copies guest buffer content to host mbufs and
387 * store them in pkts to be processed.
391 * virtio queue index in mq case
393 * mbuf_pool where host mbuf is allocated.
395 * array to contain packets to be dequeued
397 * packets num to be dequeued
399 * num of packets dequeued
401 uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
402 struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
405 * Get guest mem table: a list of memory regions.
407 * An rte_vhost_vhost_memory object will be allocated internaly, to hold the
408 * guest memory regions. Application should free it at destroy_device()
414 * To store the returned mem regions
416 * 0 on success, -1 on failure
418 int rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem);
421 * Get guest vring info, including the vring address, vring size, etc.
428 * the structure to hold the requested vring info
430 * 0 on success, -1 on failure
432 int rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
433 struct rte_vhost_vring *vring);
436 * Get vhost RX queue avail count.
441 * virtio queue index in mq case
443 * num of desc available
445 uint32_t rte_vhost_rx_queue_count(int vid, uint16_t qid);
451 #endif /* _RTE_VHOST_H_ */