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 <linux/vhost.h>
44 #include <linux/virtio_ring.h>
45 #include <sys/eventfd.h>
47 #include <rte_memory.h>
48 #include <rte_mempool.h>
50 #define RTE_VHOST_USER_CLIENT (1ULL << 0)
51 #define RTE_VHOST_USER_NO_RECONNECT (1ULL << 1)
52 #define RTE_VHOST_USER_DEQUEUE_ZERO_COPY (1ULL << 2)
55 * Information relating to memory regions including offsets to
56 * addresses in QEMUs memory file.
58 struct rte_vhost_mem_region {
59 uint64_t guest_phys_addr;
60 uint64_t guest_user_addr;
61 uint64_t host_user_addr;
69 * Memory structure includes region and mapping information.
71 struct rte_vhost_memory {
73 struct rte_vhost_mem_region regions[0];
76 struct rte_vhost_vring {
77 struct vring_desc *desc;
78 struct vring_avail *avail;
79 struct vring_used *used;
80 uint64_t log_guest_addr;
88 * Device and vring operations.
90 struct vhost_device_ops {
91 int (*new_device)(int vid); /**< Add device. */
92 void (*destroy_device)(int vid); /**< Remove device. */
94 int (*vring_state_changed)(int vid, uint16_t queue_id, int enable); /**< triggered when a vring is enabled or disabled */
97 * Features could be changed after the feature negotiation.
98 * For example, VHOST_F_LOG_ALL will be set/cleared at the
99 * start/end of live migration, respectively. This callback
100 * is used to inform the application on such change.
102 int (*features_changed)(int vid, uint64_t features);
104 void *reserved[4]; /**< Reserved for future extension */
108 * Convert guest physical address to host virtual address
111 * the guest memory regions
113 * the guest physical address for querying
115 * the host virtual address on success, 0 on failure
117 static inline uint64_t __attribute__((always_inline))
118 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
120 struct rte_vhost_mem_region *reg;
123 for (i = 0; i < mem->nregions; i++) {
124 reg = &mem->regions[i];
125 if (gpa >= reg->guest_phys_addr &&
126 gpa < reg->guest_phys_addr + reg->size) {
127 return gpa - reg->guest_phys_addr +
135 #define RTE_VHOST_NEED_LOG(features) ((features) & (1ULL << VHOST_F_LOG_ALL))
138 * Log the memory write start with given address.
140 * This function only need be invoked when the live migration starts.
141 * Therefore, we won't need call it at all in the most of time. For
142 * making the performance impact be minimum, it's suggested to do a
143 * check before calling it:
145 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
146 * rte_vhost_log_write(vid, addr, len);
151 * the starting address for write
153 * the length to write
155 void rte_vhost_log_write(int vid, uint64_t addr, uint64_t len);
158 * Log the used ring update start at given offset.
160 * Same as rte_vhost_log_write, it's suggested to do a check before
163 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
164 * rte_vhost_log_used_vring(vid, vring_idx, offset, len);
171 * the offset inside the used ring
173 * the length to write
175 void rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
176 uint64_t offset, uint64_t len);
178 int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
181 * Register vhost driver. path could be different for multiple
184 int rte_vhost_driver_register(const char *path, uint64_t flags);
186 /* Unregister vhost driver. This is only meaningful to vhost user. */
187 int rte_vhost_driver_unregister(const char *path);
190 * Set the feature bits the vhost-user driver supports.
193 * The vhost-user socket file path
197 * 0 on success, -1 on failure
199 int rte_vhost_driver_set_features(const char *path, uint64_t features);
202 * Enable vhost-user driver features.
205 * - the param features should be a subset of the feature bits provided
206 * by rte_vhost_driver_set_features().
207 * - it must be invoked before vhost-user negotiation starts.
210 * The vhost-user socket file path
214 * 0 on success, -1 on failure
216 int rte_vhost_driver_enable_features(const char *path, uint64_t features);
219 * Disable vhost-user driver features.
221 * The two notes at rte_vhost_driver_enable_features() also apply here.
224 * The vhost-user socket file path
226 * Features to disable
228 * 0 on success, -1 on failure
230 int rte_vhost_driver_disable_features(const char *path, uint64_t features);
233 * Get the feature bits before feature negotiation.
236 * The vhost-user socket file path
238 * A pointer to store the queried feature bits
240 * 0 on success, -1 on failure
242 int rte_vhost_driver_get_features(const char *path, uint64_t *features);
245 * Get the feature bits after negotiation
250 * A pointer to store the queried feature bits
252 * 0 on success, -1 on failure
254 int rte_vhost_get_negotiated_features(int vid, uint64_t *features);
256 /* Register callbacks. */
257 int rte_vhost_driver_callback_register(const char *path,
258 struct vhost_device_ops const * const ops);
262 * Start the vhost-user driver.
264 * This function triggers the vhost-user negotiation.
267 * The vhost-user socket file path
269 * 0 on success, -1 on failure
271 int rte_vhost_driver_start(const char *path);
274 * Get the MTU value of the device if set in QEMU.
277 * virtio-net device ID
279 * The variable to store the MTU value
283 * -EAGAIN: device not yet started
284 * -ENOTSUP: device does not support MTU feature
286 int rte_vhost_get_mtu(int vid, uint16_t *mtu);
289 * Get the numa node from which the virtio net device's memory
296 * The numa node, -1 on failure
298 int rte_vhost_get_numa_node(int vid);
302 * Get the number of queues the device supports.
304 * Note this function is deprecated, as it returns a queue pair number,
305 * which is vhost specific. Instead, rte_vhost_get_vring_num should
312 * The number of queues, 0 on failure
315 uint32_t rte_vhost_get_queue_num(int vid);
318 * Get the number of vrings the device supports.
324 * The number of vrings, 0 on failure
326 uint16_t rte_vhost_get_vring_num(int vid);
329 * Get the virtio net device's ifname, which is the vhost-user socket
335 * The buffer to stored the queried ifname
340 * 0 on success, -1 on failure
342 int rte_vhost_get_ifname(int vid, char *buf, size_t len);
345 * Get how many avail entries are left in the queue
353 * num of avail entires left
355 uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
360 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
361 * be received from the physical port or from another virtual device. A packet
362 * count is returned to indicate the number of packets that were succesfully
363 * added to the RX queue.
367 * virtio queue index in mq case
369 * array to contain packets to be enqueued
371 * packets num to be enqueued
373 * num of packets enqueued
375 uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
376 struct rte_mbuf **pkts, uint16_t count);
379 * This function gets guest buffers from the virtio device TX virtqueue,
380 * construct host mbufs, copies guest buffer content to host mbufs and
381 * store them in pkts to be processed.
385 * virtio queue index in mq case
387 * mbuf_pool where host mbuf is allocated.
389 * array to contain packets to be dequeued
391 * packets num to be dequeued
393 * num of packets dequeued
395 uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
396 struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
399 * Get guest mem table: a list of memory regions.
401 * An rte_vhost_vhost_memory object will be allocated internaly, to hold the
402 * guest memory regions. Application should free it at destroy_device()
408 * To store the returned mem regions
410 * 0 on success, -1 on failure
412 int rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem);
415 * Get guest vring info, including the vring address, vring size, etc.
422 * the structure to hold the requested vring info
424 * 0 on success, -1 on failure
426 int rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
427 struct rte_vhost_vring *vring);
429 #endif /* _RTE_VHOST_H_ */