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 int (*new_connection)(int vid);
111 void (*destroy_connection)(int vid);
113 void *reserved[2]; /**< Reserved for future extension */
117 * Convert guest physical address to host virtual address
120 * the guest memory regions
122 * the guest physical address for querying
124 * the host virtual address on success, 0 on failure
126 static __rte_always_inline uint64_t
127 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
129 struct rte_vhost_mem_region *reg;
132 for (i = 0; i < mem->nregions; i++) {
133 reg = &mem->regions[i];
134 if (gpa >= reg->guest_phys_addr &&
135 gpa < reg->guest_phys_addr + reg->size) {
136 return gpa - reg->guest_phys_addr +
144 #define RTE_VHOST_NEED_LOG(features) ((features) & (1ULL << VHOST_F_LOG_ALL))
147 * Log the memory write start with given address.
149 * This function only need be invoked when the live migration starts.
150 * Therefore, we won't need call it at all in the most of time. For
151 * making the performance impact be minimum, it's suggested to do a
152 * check before calling it:
154 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
155 * rte_vhost_log_write(vid, addr, len);
160 * the starting address for write
162 * the length to write
164 void rte_vhost_log_write(int vid, uint64_t addr, uint64_t len);
167 * Log the used ring update start at given offset.
169 * Same as rte_vhost_log_write, it's suggested to do a check before
172 * if (unlikely(RTE_VHOST_NEED_LOG(features)))
173 * rte_vhost_log_used_vring(vid, vring_idx, offset, len);
180 * the offset inside the used ring
182 * the length to write
184 void rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
185 uint64_t offset, uint64_t len);
187 int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
190 * Register vhost driver. path could be different for multiple
193 int rte_vhost_driver_register(const char *path, uint64_t flags);
195 /* Unregister vhost driver. This is only meaningful to vhost user. */
196 int rte_vhost_driver_unregister(const char *path);
199 * Set the feature bits the vhost-user driver supports.
202 * The vhost-user socket file path
206 * 0 on success, -1 on failure
208 int rte_vhost_driver_set_features(const char *path, uint64_t features);
211 * Enable vhost-user driver features.
214 * - the param features should be a subset of the feature bits provided
215 * by rte_vhost_driver_set_features().
216 * - it must be invoked before vhost-user negotiation starts.
219 * The vhost-user socket file path
223 * 0 on success, -1 on failure
225 int rte_vhost_driver_enable_features(const char *path, uint64_t features);
228 * Disable vhost-user driver features.
230 * The two notes at rte_vhost_driver_enable_features() also apply here.
233 * The vhost-user socket file path
235 * Features to disable
237 * 0 on success, -1 on failure
239 int rte_vhost_driver_disable_features(const char *path, uint64_t features);
242 * Get the feature bits before feature negotiation.
245 * The vhost-user socket file path
247 * A pointer to store the queried feature bits
249 * 0 on success, -1 on failure
251 int rte_vhost_driver_get_features(const char *path, uint64_t *features);
254 * Get the feature bits after negotiation
259 * A pointer to store the queried feature bits
261 * 0 on success, -1 on failure
263 int rte_vhost_get_negotiated_features(int vid, uint64_t *features);
265 /* Register callbacks. */
266 int rte_vhost_driver_callback_register(const char *path,
267 struct vhost_device_ops const * const ops);
271 * Start the vhost-user driver.
273 * This function triggers the vhost-user negotiation.
276 * The vhost-user socket file path
278 * 0 on success, -1 on failure
280 int rte_vhost_driver_start(const char *path);
283 * Get the MTU value of the device if set in QEMU.
286 * virtio-net device ID
288 * The variable to store the MTU value
292 * -EAGAIN: device not yet started
293 * -ENOTSUP: device does not support MTU feature
295 int rte_vhost_get_mtu(int vid, uint16_t *mtu);
298 * Get the numa node from which the virtio net device's memory
305 * The numa node, -1 on failure
307 int rte_vhost_get_numa_node(int vid);
311 * Get the number of queues the device supports.
313 * Note this function is deprecated, as it returns a queue pair number,
314 * which is vhost specific. Instead, rte_vhost_get_vring_num should
321 * The number of queues, 0 on failure
324 uint32_t rte_vhost_get_queue_num(int vid);
327 * Get the number of vrings the device supports.
333 * The number of vrings, 0 on failure
335 uint16_t rte_vhost_get_vring_num(int vid);
338 * Get the virtio net device's ifname, which is the vhost-user socket
344 * The buffer to stored the queried ifname
349 * 0 on success, -1 on failure
351 int rte_vhost_get_ifname(int vid, char *buf, size_t len);
354 * Get how many avail entries are left in the queue
362 * num of avail entires left
364 uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
369 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
370 * be received from the physical port or from another virtual device. A packet
371 * count is returned to indicate the number of packets that were successfully
372 * added to the RX queue.
376 * virtio queue index in mq case
378 * array to contain packets to be enqueued
380 * packets num to be enqueued
382 * num of packets enqueued
384 uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
385 struct rte_mbuf **pkts, uint16_t count);
388 * This function gets guest buffers from the virtio device TX virtqueue,
389 * construct host mbufs, copies guest buffer content to host mbufs and
390 * store them in pkts to be processed.
394 * virtio queue index in mq case
396 * mbuf_pool where host mbuf is allocated.
398 * array to contain packets to be dequeued
400 * packets num to be dequeued
402 * num of packets dequeued
404 uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
405 struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
408 * Get guest mem table: a list of memory regions.
410 * An rte_vhost_vhost_memory object will be allocated internaly, to hold the
411 * guest memory regions. Application should free it at destroy_device()
417 * To store the returned mem regions
419 * 0 on success, -1 on failure
421 int rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem);
424 * Get guest vring info, including the vring address, vring size, etc.
431 * the structure to hold the requested vring info
433 * 0 on success, -1 on failure
435 int rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
436 struct rte_vhost_vring *vring);
439 * Get vhost RX queue avail count.
444 * virtio queue index in mq case
446 * num of desc available
448 uint32_t rte_vhost_rx_queue_count(int vid, uint16_t qid);
454 #endif /* _RTE_VHOST_H_ */