]> git.proxmox.com Git - ceph.git/blame - ceph/src/spdk/lib/vhost/rte_vhost/vhost.c
import 15.2.0 Octopus source
[ceph.git] / ceph / src / spdk / lib / vhost / rte_vhost / vhost.c
CommitLineData
7c673cae
FG
1/*-
2 * BSD LICENSE
3 *
4 * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
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
16 * distribution.
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.
20 *
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.
32 */
33
34#include <linux/vhost.h>
35#include <linux/virtio_net.h>
36#include <stddef.h>
37#include <stdint.h>
38#include <stdlib.h>
39#ifdef RTE_LIBRTE_VHOST_NUMA
40#include <numaif.h>
41#endif
42
43#include <rte_ethdev.h>
44#include <rte_log.h>
45#include <rte_string_fns.h>
46#include <rte_memory.h>
47#include <rte_malloc.h>
11fdf7f2 48#include <rte_vhost.h>
7c673cae
FG
49
50#include "vhost.h"
51
7c673cae
FG
52struct virtio_net *vhost_devices[MAX_VHOST_DEVICE];
53
7c673cae
FG
54struct virtio_net *
55get_device(int vid)
56{
57 struct virtio_net *dev = vhost_devices[vid];
58
59 if (unlikely(!dev)) {
60 RTE_LOG(ERR, VHOST_CONFIG,
61 "(%d) device not found.\n", vid);
62 }
63
64 return dev;
65}
66
67static void
68cleanup_vq(struct vhost_virtqueue *vq, int destroy)
69{
70 if ((vq->callfd >= 0) && (destroy != 0))
71 close(vq->callfd);
72 if (vq->kickfd >= 0)
73 close(vq->kickfd);
74}
75
76/*
77 * Unmap any memory, close any file descriptors and
78 * free any memory owned by a device.
79 */
80void
81cleanup_device(struct virtio_net *dev, int destroy)
82{
83 uint32_t i;
84
85 vhost_backend_cleanup(dev);
86
11fdf7f2
TL
87 for (i = 0; i < dev->nr_vring; i++)
88 cleanup_vq(dev->virtqueue[i], destroy);
7c673cae
FG
89}
90
91/*
92 * Release virtqueues and device memory.
93 */
94static void
95free_device(struct virtio_net *dev)
96{
97 uint32_t i;
11fdf7f2 98 struct vhost_virtqueue *vq;
7c673cae 99
11fdf7f2
TL
100 for (i = 0; i < dev->nr_vring; i++) {
101 vq = dev->virtqueue[i];
7c673cae 102
11fdf7f2 103 rte_free(vq->shadow_used_ring);
7c673cae 104
11fdf7f2 105 rte_free(vq);
7c673cae
FG
106 }
107
108 rte_free(dev);
109}
110
111static void
11fdf7f2 112init_vring_queue(struct vhost_virtqueue *vq)
7c673cae
FG
113{
114 memset(vq, 0, sizeof(struct vhost_virtqueue));
115
116 vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
117 vq->callfd = VIRTIO_UNINITIALIZED_EVENTFD;
118
119 /* Backends are set to -1 indicating an inactive device. */
120 vq->backend = -1;
121
11fdf7f2
TL
122 /*
123 * always set the vq to enabled; this is to keep compatibility
124 * with the old QEMU, whereas there is no SET_VRING_ENABLE message.
125 */
126 vq->enabled = 1;
7c673cae
FG
127
128 TAILQ_INIT(&vq->zmbuf_list);
129}
130
131static void
11fdf7f2 132reset_vring_queue(struct vhost_virtqueue *vq)
7c673cae
FG
133{
134 int callfd;
135
136 callfd = vq->callfd;
11fdf7f2 137 init_vring_queue(vq);
7c673cae
FG
138 vq->callfd = callfd;
139}
140
7c673cae 141int
11fdf7f2 142alloc_vring_queue(struct virtio_net *dev, uint32_t vring_idx)
7c673cae 143{
11fdf7f2 144 struct vhost_virtqueue *vq;
7c673cae 145
11fdf7f2
TL
146 vq = rte_malloc(NULL, sizeof(struct vhost_virtqueue), 0);
147 if (vq == NULL) {
7c673cae 148 RTE_LOG(ERR, VHOST_CONFIG,
11fdf7f2 149 "Failed to allocate memory for vring:%u.\n", vring_idx);
7c673cae
FG
150 return -1;
151 }
152
11fdf7f2
TL
153 dev->virtqueue[vring_idx] = vq;
154 init_vring_queue(vq);
7c673cae 155
11fdf7f2 156 dev->nr_vring += 1;
7c673cae
FG
157
158 return 0;
159}
160
161/*
162 * Reset some variables in device structure, while keeping few
11fdf7f2 163 * others untouched, such as vid, ifname, nr_vring: they
7c673cae
FG
164 * should be same unless the device is removed.
165 */
166void
167reset_device(struct virtio_net *dev)
168{
169 uint32_t i;
170
11fdf7f2 171 dev->negotiated_features = 0;
7c673cae
FG
172 dev->protocol_features = 0;
173 dev->flags = 0;
174
11fdf7f2
TL
175 for (i = 0; i < dev->nr_vring; i++)
176 reset_vring_queue(dev->virtqueue[i]);
7c673cae
FG
177}
178
179/*
180 * Invoked when there is a new vhost-user connection established (when
181 * there is a new virtio device being attached).
182 */
183int
9f95a23c 184vhost_new_device(uint64_t features, struct vhost_device_ops const *ops)
7c673cae
FG
185{
186 struct virtio_net *dev;
187 int i;
188
189 dev = rte_zmalloc(NULL, sizeof(struct virtio_net), 0);
190 if (dev == NULL) {
191 RTE_LOG(ERR, VHOST_CONFIG,
192 "Failed to allocate memory for new dev.\n");
193 return -1;
194 }
195
196 for (i = 0; i < MAX_VHOST_DEVICE; i++) {
197 if (vhost_devices[i] == NULL)
198 break;
199 }
200 if (i == MAX_VHOST_DEVICE) {
201 RTE_LOG(ERR, VHOST_CONFIG,
202 "Failed to find a free slot for new device.\n");
11fdf7f2 203 rte_free(dev);
7c673cae
FG
204 return -1;
205 }
206
207 vhost_devices[i] = dev;
208 dev->vid = i;
11fdf7f2 209 dev->features = features;
9f95a23c 210 dev->notify_ops = ops;
7c673cae
FG
211
212 return i;
213}
214
215/*
216 * Invoked when there is the vhost-user connection is broken (when
217 * the virtio device is being detached).
218 */
219void
220vhost_destroy_device(int vid)
221{
222 struct virtio_net *dev = get_device(vid);
223
224 if (dev == NULL)
225 return;
226
227 if (dev->flags & VIRTIO_DEV_RUNNING) {
228 dev->flags &= ~VIRTIO_DEV_RUNNING;
11fdf7f2 229 dev->notify_ops->destroy_device(vid);
7c673cae
FG
230 }
231
232 cleanup_device(dev, 1);
233 free_device(dev);
234
235 vhost_devices[vid] = NULL;
236}
237
238void
239vhost_set_ifname(int vid, const char *if_name, unsigned int if_len)
240{
241 struct virtio_net *dev;
242 unsigned int len;
243
244 dev = get_device(vid);
245 if (dev == NULL)
246 return;
247
248 len = if_len > sizeof(dev->ifname) ?
249 sizeof(dev->ifname) : if_len;
250
251 strncpy(dev->ifname, if_name, len);
252 dev->ifname[sizeof(dev->ifname) - 1] = '\0';
253}
254
255void
256vhost_enable_dequeue_zero_copy(int vid)
257{
258 struct virtio_net *dev = get_device(vid);
259
260 if (dev == NULL)
261 return;
262
263 dev->dequeue_zero_copy = 1;
264}
265
11fdf7f2
TL
266int
267rte_vhost_get_mtu(int vid, uint16_t *mtu)
268{
269 struct virtio_net *dev = get_device(vid);
270
271 if (!dev)
272 return -ENODEV;
273
274 if (!(dev->flags & VIRTIO_DEV_READY))
275 return -EAGAIN;
276
277 if (!(dev->negotiated_features & VIRTIO_NET_F_MTU))
278 return -ENOTSUP;
279
280 *mtu = dev->mtu;
281
282 return 0;
283}
284
7c673cae
FG
285int
286rte_vhost_get_numa_node(int vid)
287{
288#ifdef RTE_LIBRTE_VHOST_NUMA
289 struct virtio_net *dev = get_device(vid);
290 int numa_node;
291 int ret;
292
293 if (dev == NULL)
294 return -1;
295
296 ret = get_mempolicy(&numa_node, NULL, 0, dev,
297 MPOL_F_NODE | MPOL_F_ADDR);
298 if (ret < 0) {
299 RTE_LOG(ERR, VHOST_CONFIG,
300 "(%d) failed to query numa node: %d\n", vid, ret);
301 return -1;
302 }
303
304 return numa_node;
305#else
306 RTE_SET_USED(vid);
307 return -1;
308#endif
309}
310
11fdf7f2
TL
311int
312rte_vhost_get_ifname(int vid, char *buf, size_t len)
7c673cae
FG
313{
314 struct virtio_net *dev = get_device(vid);
315
316 if (dev == NULL)
11fdf7f2
TL
317 return -1;
318
319 len = RTE_MIN(len, sizeof(dev->ifname));
320
321 strncpy(buf, dev->ifname, len);
322 buf[len - 1] = '\0';
7c673cae 323
11fdf7f2 324 return 0;
7c673cae
FG
325}
326
327int
11fdf7f2 328rte_vhost_get_negotiated_features(int vid, uint64_t *features)
7c673cae 329{
11fdf7f2 330 struct virtio_net *dev;
7c673cae 331
11fdf7f2
TL
332 dev = get_device(vid);
333 if (!dev)
7c673cae
FG
334 return -1;
335
11fdf7f2
TL
336 *features = dev->negotiated_features;
337 return 0;
338}
7c673cae 339
11fdf7f2
TL
340int
341rte_vhost_get_mem_table(int vid, struct rte_vhost_memory **mem)
342{
343 struct virtio_net *dev;
344 struct rte_vhost_memory *m;
345 size_t size;
346
347 dev = get_device(vid);
348 if (!dev)
349 return -1;
350
351 size = dev->mem->nregions * sizeof(struct rte_vhost_mem_region);
352 m = malloc(sizeof(struct rte_vhost_memory) + size);
353 if (!m)
354 return -1;
355
356 m->nregions = dev->mem->nregions;
357 memcpy(m->regions, dev->mem->regions, size);
358 *mem = m;
359
360 return 0;
361}
362
363int
364rte_vhost_get_vhost_vring(int vid, uint16_t vring_idx,
365 struct rte_vhost_vring *vring)
366{
367 struct virtio_net *dev;
368 struct vhost_virtqueue *vq;
369
370 dev = get_device(vid);
371 if (!dev)
372 return -1;
373
374 if (vring_idx >= VHOST_MAX_VRING)
375 return -1;
376
377 vq = dev->virtqueue[vring_idx];
378 if (!vq)
379 return -1;
380
381 vring->desc = vq->desc;
382 vring->avail = vq->avail;
383 vring->used = vq->used;
384 vring->log_guest_addr = vq->log_guest_addr;
385
386 vring->callfd = vq->callfd;
387 vring->kickfd = vq->kickfd;
388 vring->size = vq->size;
389
7c673cae
FG
390 return 0;
391}
392
393uint16_t
394rte_vhost_avail_entries(int vid, uint16_t queue_id)
395{
396 struct virtio_net *dev;
397 struct vhost_virtqueue *vq;
398
399 dev = get_device(vid);
400 if (!dev)
401 return 0;
402
403 vq = dev->virtqueue[queue_id];
404 if (!vq->enabled)
405 return 0;
406
407 return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx;
408}
409
410int
411rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
412{
413 struct virtio_net *dev = get_device(vid);
414
415 if (dev == NULL)
416 return -1;
417
418 if (enable) {
419 RTE_LOG(ERR, VHOST_CONFIG,
420 "guest notification isn't supported.\n");
421 return -1;
422 }
423
424 dev->virtqueue[queue_id]->used->flags = VRING_USED_F_NO_NOTIFY;
425 return 0;
426}
427
11fdf7f2
TL
428void
429rte_vhost_log_write(int vid, uint64_t addr, uint64_t len)
7c673cae 430{
11fdf7f2 431 struct virtio_net *dev = get_device(vid);
7c673cae 432
11fdf7f2
TL
433 if (dev == NULL)
434 return;
435
436 vhost_log_write(dev, addr, len);
7c673cae
FG
437}
438
11fdf7f2
TL
439void
440rte_vhost_log_used_vring(int vid, uint16_t vring_idx,
441 uint64_t offset, uint64_t len)
7c673cae 442{
11fdf7f2
TL
443 struct virtio_net *dev;
444 struct vhost_virtqueue *vq;
445
446 dev = get_device(vid);
447 if (dev == NULL)
448 return;
449
450 if (vring_idx >= VHOST_MAX_VRING)
451 return;
452 vq = dev->virtqueue[vring_idx];
453 if (!vq)
454 return;
455
456 vhost_log_used_vring(dev, vq, offset, len);
7c673cae
FG
457}
458
7c673cae 459int
9f95a23c
TL
460rte_vhost_set_vring_base(int vid, uint16_t vring_idx,
461 uint16_t last_avail_idx, uint16_t last_used_idx)
462{
11fdf7f2
TL
463 struct virtio_net *dev;
464 struct vhost_virtqueue *vq;
465
466 dev = get_device(vid);
467 if (!dev)
468 return -1;
469
470 if (vring_idx >= VHOST_MAX_VRING)
471 return -1;
472
473 vq = dev->virtqueue[vring_idx];
474 if (!vq)
475 return -1;
476
477 vq->last_avail_idx = last_avail_idx;
478 vq->last_used_idx = last_used_idx;
7c673cae
FG
479
480 return 0;
481}
9f95a23c
TL
482
483int
484rte_vhost_get_vring_base(int vid, uint16_t vring_idx,
485 uint16_t *last_avail_idx, uint16_t *last_used_idx)
486{
487 struct virtio_net *dev;
488 struct vhost_virtqueue *vq;
489
490 dev = get_device(vid);
491 if (!dev)
492 return -1;
493
494 if (vring_idx >= VHOST_MAX_VRING)
495 return -1;
496
497 vq = dev->virtqueue[vring_idx];
498 if (!vq)
499 return -1;
500
501 *last_avail_idx = vq->last_avail_idx;
502 *last_used_idx = vq->last_used_idx;
503
504 return 0;
505}
506
507int
508rte_vhost_vring_call(int vid, uint16_t vring_idx)
509{
510 struct virtio_net *dev;
511 struct vhost_virtqueue *vq;
512
513 dev = get_device(vid);
514 if(!dev)
515 return -1;
516
517 if (vring_idx >= VHOST_MAX_VRING)
518 return -1;
519
520 vq = dev->virtqueue[vring_idx];
521 if (!vq)
522 return -1;
523
524 if (vq->callfd != -1) {
525 eventfd_write(vq->callfd, (eventfd_t)1);
526 return 0;
527 }
528
529 return -1;
530}