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)
59 #define RTE_VHOST_USER_IOMMU_SUPPORT (1ULL << 3)
62 * Information relating to memory regions including offsets to
63 * addresses in QEMUs memory file.
65 struct rte_vhost_mem_region {
66 uint64_t guest_phys_addr;
67 uint64_t guest_user_addr;
68 uint64_t host_user_addr;
76 * Memory structure includes region and mapping information.
78 struct rte_vhost_memory {
80 struct rte_vhost_mem_region regions[];
83 struct rte_vhost_vring {
84 struct vring_desc *desc;
85 struct vring_avail *avail;
86 struct vring_used *used;
87 uint64_t log_guest_addr;
95 * Device and vring operations.
97 struct vhost_device_ops {
98 int (*new_device)(int vid); /**< Add device. */
99 void (*destroy_device)(int vid); /**< Remove device. */
101 int (*vring_state_changed)(int vid, uint16_t queue_id, int enable); /**< triggered when a vring is enabled or disabled */
104 * Features could be changed after the feature negotiation.
105 * For example, VHOST_F_LOG_ALL will be set/cleared at the
106 * start/end of live migration, respectively. This callback
107 * is used to inform the application on such change.
109 int (*features_changed)(int vid, uint64_t features);
111 int (*new_connection)(int vid);
112 void (*destroy_connection)(int vid);
114 void *reserved[2]; /**< Reserved for future extension */
118 * Convert guest physical address to host virtual address
121 * the guest memory regions
123 * the guest physical address for querying
125 * the host virtual address on success, 0 on failure
127 static __rte_always_inline uint64_t
128 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
130 struct rte_vhost_mem_region *reg;
133 for (i = 0; i < mem->nregions; i++) {
134 reg = &mem->regions[i];
135 if (gpa >= reg->guest_phys_addr &&
136 gpa < reg->guest_phys_addr + reg->size) {
137 return gpa - reg->guest_phys_addr +
145 #define RTE_VHOST_NEED_LOG(features) ((features) & (1ULL << VHOST_F_LOG_ALL))
148 * Log the memory write start with given address.
150 * This function only need be invoked when the live migration starts.
151 * Therefore, we won't need call it at all in the most of time. For
152 * making the performance impact be minimum, it's suggested to do a
153 * check before calling it:
155 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
156 * rte_vhost_log_write(vid, addr, len);
161 * the starting address for write
163 * the length to write
165 void rte_vhost_log_write(int vid, uint64_t addr, uint64_t len);
168 * Log the used ring update start at given offset.
170 * Same as rte_vhost_log_write, it's suggested to do a check before
173 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
174 * rte_vhost_log_used_vring(vid, vring_idx, offset, len);
181 * the offset inside the used ring
183 * the length to write
185 void rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
186 uint64_t offset, uint64_t len);
188 int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
191 * Register vhost driver. path could be different for multiple
194 int rte_vhost_driver_register(const char *path, uint64_t flags);
196 /* Unregister vhost driver. This is only meaningful to vhost user. */
197 int rte_vhost_driver_unregister(const char *path);
200 * Set the feature bits the vhost-user driver supports.
203 * The vhost-user socket file path
207 * 0 on success, -1 on failure
209 int rte_vhost_driver_set_features(const char *path, uint64_t features);
212 * Enable vhost-user driver features.
215 * - the param features should be a subset of the feature bits provided
216 * by rte_vhost_driver_set_features().
217 * - it must be invoked before vhost-user negotiation starts.
220 * The vhost-user socket file path
224 * 0 on success, -1 on failure
226 int rte_vhost_driver_enable_features(const char *path, uint64_t features);
229 * Disable vhost-user driver features.
231 * The two notes at rte_vhost_driver_enable_features() also apply here.
234 * The vhost-user socket file path
236 * Features to disable
238 * 0 on success, -1 on failure
240 int rte_vhost_driver_disable_features(const char *path, uint64_t features);
243 * Get the feature bits before feature negotiation.
246 * The vhost-user socket file path
248 * A pointer to store the queried feature bits
250 * 0 on success, -1 on failure
252 int rte_vhost_driver_get_features(const char *path, uint64_t *features);
255 * Get the feature bits after negotiation
260 * A pointer to store the queried feature bits
262 * 0 on success, -1 on failure
264 int rte_vhost_get_negotiated_features(int vid, uint64_t *features);
266 /* Register callbacks. */
267 int rte_vhost_driver_callback_register(const char *path,
268 struct vhost_device_ops const * const ops);
272 * Start the vhost-user driver.
274 * This function triggers the vhost-user negotiation.
277 * The vhost-user socket file path
279 * 0 on success, -1 on failure
281 int rte_vhost_driver_start(const char *path);
284 * Get the MTU value of the device if set in QEMU.
287 * virtio-net device ID
289 * The variable to store the MTU value
293 * -EAGAIN: device not yet started
294 * -ENOTSUP: device does not support MTU feature
296 int rte_vhost_get_mtu(int vid, uint16_t *mtu);
299 * Get the numa node from which the virtio net device's memory
306 * The numa node, -1 on failure
308 int rte_vhost_get_numa_node(int vid);
312 * Get the number of queues the device supports.
314 * Note this function is deprecated, as it returns a queue pair number,
315 * which is vhost specific. Instead, rte_vhost_get_vring_num should
322 * The number of queues, 0 on failure
325 uint32_t rte_vhost_get_queue_num(int vid);
328 * Get the number of vrings the device supports.
334 * The number of vrings, 0 on failure
336 uint16_t rte_vhost_get_vring_num(int vid);
339 * Get the virtio net device's ifname, which is the vhost-user socket
345 * The buffer to stored the queried ifname
350 * 0 on success, -1 on failure
352 int rte_vhost_get_ifname(int vid, char *buf, size_t len);
355 * Get how many avail entries are left in the queue
363 * num of avail entires left
365 uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
370 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
371 * be received from the physical port or from another virtual device. A packet
372 * count is returned to indicate the number of packets that were successfully
373 * added to the RX queue.
377 * virtio queue index in mq case
379 * array to contain packets to be enqueued
381 * packets num to be enqueued
383 * num of packets enqueued
385 uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
386 struct rte_mbuf **pkts, uint16_t count);
389 * This function gets guest buffers from the virtio device TX virtqueue,
390 * construct host mbufs, copies guest buffer content to host mbufs and
391 * store them in pkts to be processed.
395 * virtio queue index in mq case
397 * mbuf_pool where host mbuf is allocated.
399 * array to contain packets to be dequeued
401 * packets num to be dequeued
403 * num of packets dequeued
405 uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
406 struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
409 * Get guest mem table: a list of memory regions.
411 * An rte_vhost_vhost_memory object will be allocated internaly, to hold the
412 * guest memory regions. Application should free it at destroy_device()
418 * To store the returned mem regions
420 * 0 on success, -1 on failure
422 int rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem);
425 * Get guest vring info, including the vring address, vring size, etc.
432 * the structure to hold the requested vring info
434 * 0 on success, -1 on failure
436 int rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
437 struct rte_vhost_vring *vring);
440 * Get vhost RX queue avail count.
445 * virtio queue index in mq case
447 * num of desc available
449 uint32_t rte_vhost_rx_queue_count(int vid, uint16_t qid);
455 #endif /* _RTE_VHOST_H_ */