xref: /openbmc/linux/drivers/vhost/vdpa.c (revision b1a79260) 
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2018-2020 Intel Corporation.
4  * Copyright (C) 2020 Red Hat, Inc.
5  *
6  * Author: Tiwei Bie <tiwei.bie@intel.com>
7  *         Jason Wang <jasowang@redhat.com>
8  *
9  * Thanks Michael S. Tsirkin for the valuable comments and
10  * suggestions.  And thanks to Cunming Liang and Zhihong Wang for all
11  * their supports.
12  */
13 
14 #include <linux/kernel.h>
15 #include <linux/module.h>
16 #include <linux/cdev.h>
17 #include <linux/device.h>
18 #include <linux/mm.h>
19 #include <linux/iommu.h>
20 #include <linux/uuid.h>
21 #include <linux/vdpa.h>
22 #include <linux/nospec.h>
23 #include <linux/vhost.h>
24 #include <linux/virtio_net.h>
25 
26 #include "vhost.h"
27 
28 enum {
29 	VHOST_VDPA_BACKEND_FEATURES =
30 	(1ULL << VHOST_BACKEND_F_IOTLB_MSG_V2) |
31 	(1ULL << VHOST_BACKEND_F_IOTLB_BATCH),
32 };
33 
34 #define VHOST_VDPA_DEV_MAX (1U << MINORBITS)
35 
36 struct vhost_vdpa {
37 	struct vhost_dev vdev;
38 	struct iommu_domain *domain;
39 	struct vhost_virtqueue *vqs;
40 	struct completion completion;
41 	struct vdpa_device *vdpa;
42 	struct device dev;
43 	struct cdev cdev;
44 	atomic_t opened;
45 	int nvqs;
46 	int virtio_id;
47 	int minor;
48 	struct eventfd_ctx *config_ctx;
49 	int in_batch;
50 	struct vdpa_iova_range range;
51 };
52 
53 static DEFINE_IDA(vhost_vdpa_ida);
54 
55 static dev_t vhost_vdpa_major;
56 
57 static void handle_vq_kick(struct vhost_work *work)
58 {
59 	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
60 						  poll.work);
61 	struct vhost_vdpa *v = container_of(vq->dev, struct vhost_vdpa, vdev);
62 	const struct vdpa_config_ops *ops = v->vdpa->config;
63 
64 	ops->kick_vq(v->vdpa, vq - v->vqs);
65 }
66 
67 static irqreturn_t vhost_vdpa_virtqueue_cb(void *private)
68 {
69 	struct vhost_virtqueue *vq = private;
70 	struct eventfd_ctx *call_ctx = vq->call_ctx.ctx;
71 
72 	if (call_ctx)
73 		eventfd_signal(call_ctx, 1);
74 
75 	return IRQ_HANDLED;
76 }
77 
78 static irqreturn_t vhost_vdpa_config_cb(void *private)
79 {
80 	struct vhost_vdpa *v = private;
81 	struct eventfd_ctx *config_ctx = v->config_ctx;
82 
83 	if (config_ctx)
84 		eventfd_signal(config_ctx, 1);
85 
86 	return IRQ_HANDLED;
87 }
88 
89 static void vhost_vdpa_setup_vq_irq(struct vhost_vdpa *v, u16 qid)
90 {
91 	struct vhost_virtqueue *vq = &v->vqs[qid];
92 	const struct vdpa_config_ops *ops = v->vdpa->config;
93 	struct vdpa_device *vdpa = v->vdpa;
94 	int ret, irq;
95 
96 	if (!ops->get_vq_irq)
97 		return;
98 
99 	irq = ops->get_vq_irq(vdpa, qid);
100 	irq_bypass_unregister_producer(&vq->call_ctx.producer);
101 	if (!vq->call_ctx.ctx || irq < 0)
102 		return;
103 
104 	vq->call_ctx.producer.token = vq->call_ctx.ctx;
105 	vq->call_ctx.producer.irq = irq;
106 	ret = irq_bypass_register_producer(&vq->call_ctx.producer);
107 	if (unlikely(ret))
108 		dev_info(&v->dev, "vq %u, irq bypass producer (token %p) registration fails, ret =  %d\n",
109 			 qid, vq->call_ctx.producer.token, ret);
110 }
111 
112 static void vhost_vdpa_unsetup_vq_irq(struct vhost_vdpa *v, u16 qid)
113 {
114 	struct vhost_virtqueue *vq = &v->vqs[qid];
115 
116 	irq_bypass_unregister_producer(&vq->call_ctx.producer);
117 }
118 
119 static void vhost_vdpa_reset(struct vhost_vdpa *v)
120 {
121 	struct vdpa_device *vdpa = v->vdpa;
122 
123 	vdpa_reset(vdpa);
124 	v->in_batch = 0;
125 }
126 
127 static long vhost_vdpa_get_device_id(struct vhost_vdpa *v, u8 __user *argp)
128 {
129 	struct vdpa_device *vdpa = v->vdpa;
130 	const struct vdpa_config_ops *ops = vdpa->config;
131 	u32 device_id;
132 
133 	device_id = ops->get_device_id(vdpa);
134 
135 	if (copy_to_user(argp, &device_id, sizeof(device_id)))
136 		return -EFAULT;
137 
138 	return 0;
139 }
140 
141 static long vhost_vdpa_get_status(struct vhost_vdpa *v, u8 __user *statusp)
142 {
143 	struct vdpa_device *vdpa = v->vdpa;
144 	const struct vdpa_config_ops *ops = vdpa->config;
145 	u8 status;
146 
147 	status = ops->get_status(vdpa);
148 
149 	if (copy_to_user(statusp, &status, sizeof(status)))
150 		return -EFAULT;
151 
152 	return 0;
153 }
154 
155 static long vhost_vdpa_set_status(struct vhost_vdpa *v, u8 __user *statusp)
156 {
157 	struct vdpa_device *vdpa = v->vdpa;
158 	const struct vdpa_config_ops *ops = vdpa->config;
159 	u8 status, status_old;
160 	int nvqs = v->nvqs;
161 	u16 i;
162 
163 	if (copy_from_user(&status, statusp, sizeof(status)))
164 		return -EFAULT;
165 
166 	status_old = ops->get_status(vdpa);
167 
168 	/*
169 	 * Userspace shouldn't remove status bits unless reset the
170 	 * status to 0.
171 	 */
172 	if (status != 0 && (ops->get_status(vdpa) & ~status) != 0)
173 		return -EINVAL;
174 
175 	ops->set_status(vdpa, status);
176 
177 	if ((status & VIRTIO_CONFIG_S_DRIVER_OK) && !(status_old & VIRTIO_CONFIG_S_DRIVER_OK))
178 		for (i = 0; i < nvqs; i++)
179 			vhost_vdpa_setup_vq_irq(v, i);
180 
181 	if ((status_old & VIRTIO_CONFIG_S_DRIVER_OK) && !(status & VIRTIO_CONFIG_S_DRIVER_OK))
182 		for (i = 0; i < nvqs; i++)
183 			vhost_vdpa_unsetup_vq_irq(v, i);
184 
185 	return 0;
186 }
187 
188 static int vhost_vdpa_config_validate(struct vhost_vdpa *v,
189 				      struct vhost_vdpa_config *c)
190 {
191 	long size = 0;
192 
193 	switch (v->virtio_id) {
194 	case VIRTIO_ID_NET:
195 		size = sizeof(struct virtio_net_config);
196 		break;
197 	}
198 
199 	if (c->len == 0)
200 		return -EINVAL;
201 
202 	if (c->len > size - c->off)
203 		return -E2BIG;
204 
205 	return 0;
206 }
207 
208 static long vhost_vdpa_get_config(struct vhost_vdpa *v,
209 				  struct vhost_vdpa_config __user *c)
210 {
211 	struct vdpa_device *vdpa = v->vdpa;
212 	struct vhost_vdpa_config config;
213 	unsigned long size = offsetof(struct vhost_vdpa_config, buf);
214 	u8 *buf;
215 
216 	if (copy_from_user(&config, c, size))
217 		return -EFAULT;
218 	if (vhost_vdpa_config_validate(v, &config))
219 		return -EINVAL;
220 	buf = kvzalloc(config.len, GFP_KERNEL);
221 	if (!buf)
222 		return -ENOMEM;
223 
224 	vdpa_get_config(vdpa, config.off, buf, config.len);
225 
226 	if (copy_to_user(c->buf, buf, config.len)) {
227 		kvfree(buf);
228 		return -EFAULT;
229 	}
230 
231 	kvfree(buf);
232 	return 0;
233 }
234 
235 static long vhost_vdpa_set_config(struct vhost_vdpa *v,
236 				  struct vhost_vdpa_config __user *c)
237 {
238 	struct vdpa_device *vdpa = v->vdpa;
239 	const struct vdpa_config_ops *ops = vdpa->config;
240 	struct vhost_vdpa_config config;
241 	unsigned long size = offsetof(struct vhost_vdpa_config, buf);
242 	u8 *buf;
243 
244 	if (copy_from_user(&config, c, size))
245 		return -EFAULT;
246 	if (vhost_vdpa_config_validate(v, &config))
247 		return -EINVAL;
248 
249 	buf = vmemdup_user(c->buf, config.len);
250 	if (IS_ERR(buf))
251 		return PTR_ERR(buf);
252 
253 	ops->set_config(vdpa, config.off, buf, config.len);
254 
255 	kvfree(buf);
256 	return 0;
257 }
258 
259 static long vhost_vdpa_get_features(struct vhost_vdpa *v, u64 __user *featurep)
260 {
261 	struct vdpa_device *vdpa = v->vdpa;
262 	const struct vdpa_config_ops *ops = vdpa->config;
263 	u64 features;
264 
265 	features = ops->get_features(vdpa);
266 
267 	if (copy_to_user(featurep, &features, sizeof(features)))
268 		return -EFAULT;
269 
270 	return 0;
271 }
272 
273 static long vhost_vdpa_set_features(struct vhost_vdpa *v, u64 __user *featurep)
274 {
275 	struct vdpa_device *vdpa = v->vdpa;
276 	const struct vdpa_config_ops *ops = vdpa->config;
277 	u64 features;
278 
279 	/*
280 	 * It's not allowed to change the features after they have
281 	 * been negotiated.
282 	 */
283 	if (ops->get_status(vdpa) & VIRTIO_CONFIG_S_FEATURES_OK)
284 		return -EBUSY;
285 
286 	if (copy_from_user(&features, featurep, sizeof(features)))
287 		return -EFAULT;
288 
289 	if (vdpa_set_features(vdpa, features))
290 		return -EINVAL;
291 
292 	return 0;
293 }
294 
295 static long vhost_vdpa_get_vring_num(struct vhost_vdpa *v, u16 __user *argp)
296 {
297 	struct vdpa_device *vdpa = v->vdpa;
298 	const struct vdpa_config_ops *ops = vdpa->config;
299 	u16 num;
300 
301 	num = ops->get_vq_num_max(vdpa);
302 
303 	if (copy_to_user(argp, &num, sizeof(num)))
304 		return -EFAULT;
305 
306 	return 0;
307 }
308 
309 static void vhost_vdpa_config_put(struct vhost_vdpa *v)
310 {
311 	if (v->config_ctx)
312 		eventfd_ctx_put(v->config_ctx);
313 }
314 
315 static long vhost_vdpa_set_config_call(struct vhost_vdpa *v, u32 __user *argp)
316 {
317 	struct vdpa_callback cb;
318 	int fd;
319 	struct eventfd_ctx *ctx;
320 
321 	cb.callback = vhost_vdpa_config_cb;
322 	cb.private = v->vdpa;
323 	if (copy_from_user(&fd, argp, sizeof(fd)))
324 		return  -EFAULT;
325 
326 	ctx = fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(fd);
327 	swap(ctx, v->config_ctx);
328 
329 	if (!IS_ERR_OR_NULL(ctx))
330 		eventfd_ctx_put(ctx);
331 
332 	if (IS_ERR(v->config_ctx))
333 		return PTR_ERR(v->config_ctx);
334 
335 	v->vdpa->config->set_config_cb(v->vdpa, &cb);
336 
337 	return 0;
338 }
339 
340 static long vhost_vdpa_get_iova_range(struct vhost_vdpa *v, u32 __user *argp)
341 {
342 	struct vhost_vdpa_iova_range range = {
343 		.first = v->range.first,
344 		.last = v->range.last,
345 	};
346 
347 	if (copy_to_user(argp, &range, sizeof(range)))
348 		return -EFAULT;
349 	return 0;
350 }
351 
352 static long vhost_vdpa_vring_ioctl(struct vhost_vdpa *v, unsigned int cmd,
353 				   void __user *argp)
354 {
355 	struct vdpa_device *vdpa = v->vdpa;
356 	const struct vdpa_config_ops *ops = vdpa->config;
357 	struct vdpa_vq_state vq_state;
358 	struct vdpa_callback cb;
359 	struct vhost_virtqueue *vq;
360 	struct vhost_vring_state s;
361 	u32 idx;
362 	long r;
363 
364 	r = get_user(idx, (u32 __user *)argp);
365 	if (r < 0)
366 		return r;
367 
368 	if (idx >= v->nvqs)
369 		return -ENOBUFS;
370 
371 	idx = array_index_nospec(idx, v->nvqs);
372 	vq = &v->vqs[idx];
373 
374 	switch (cmd) {
375 	case VHOST_VDPA_SET_VRING_ENABLE:
376 		if (copy_from_user(&s, argp, sizeof(s)))
377 			return -EFAULT;
378 		ops->set_vq_ready(vdpa, idx, s.num);
379 		return 0;
380 	case VHOST_GET_VRING_BASE:
381 		r = ops->get_vq_state(v->vdpa, idx, &vq_state);
382 		if (r)
383 			return r;
384 
385 		vq->last_avail_idx = vq_state.avail_index;
386 		break;
387 	}
388 
389 	r = vhost_vring_ioctl(&v->vdev, cmd, argp);
390 	if (r)
391 		return r;
392 
393 	switch (cmd) {
394 	case VHOST_SET_VRING_ADDR:
395 		if (ops->set_vq_address(vdpa, idx,
396 					(u64)(uintptr_t)vq->desc,
397 					(u64)(uintptr_t)vq->avail,
398 					(u64)(uintptr_t)vq->used))
399 			r = -EINVAL;
400 		break;
401 
402 	case VHOST_SET_VRING_BASE:
403 		vq_state.avail_index = vq->last_avail_idx;
404 		if (ops->set_vq_state(vdpa, idx, &vq_state))
405 			r = -EINVAL;
406 		break;
407 
408 	case VHOST_SET_VRING_CALL:
409 		if (vq->call_ctx.ctx) {
410 			cb.callback = vhost_vdpa_virtqueue_cb;
411 			cb.private = vq;
412 		} else {
413 			cb.callback = NULL;
414 			cb.private = NULL;
415 		}
416 		ops->set_vq_cb(vdpa, idx, &cb);
417 		vhost_vdpa_setup_vq_irq(v, idx);
418 		break;
419 
420 	case VHOST_SET_VRING_NUM:
421 		ops->set_vq_num(vdpa, idx, vq->num);
422 		break;
423 	}
424 
425 	return r;
426 }
427 
428 static long vhost_vdpa_unlocked_ioctl(struct file *filep,
429 				      unsigned int cmd, unsigned long arg)
430 {
431 	struct vhost_vdpa *v = filep->private_data;
432 	struct vhost_dev *d = &v->vdev;
433 	void __user *argp = (void __user *)arg;
434 	u64 __user *featurep = argp;
435 	u64 features;
436 	long r = 0;
437 
438 	if (cmd == VHOST_SET_BACKEND_FEATURES) {
439 		if (copy_from_user(&features, featurep, sizeof(features)))
440 			return -EFAULT;
441 		if (features & ~VHOST_VDPA_BACKEND_FEATURES)
442 			return -EOPNOTSUPP;
443 		vhost_set_backend_features(&v->vdev, features);
444 		return 0;
445 	}
446 
447 	mutex_lock(&d->mutex);
448 
449 	switch (cmd) {
450 	case VHOST_VDPA_GET_DEVICE_ID:
451 		r = vhost_vdpa_get_device_id(v, argp);
452 		break;
453 	case VHOST_VDPA_GET_STATUS:
454 		r = vhost_vdpa_get_status(v, argp);
455 		break;
456 	case VHOST_VDPA_SET_STATUS:
457 		r = vhost_vdpa_set_status(v, argp);
458 		break;
459 	case VHOST_VDPA_GET_CONFIG:
460 		r = vhost_vdpa_get_config(v, argp);
461 		break;
462 	case VHOST_VDPA_SET_CONFIG:
463 		r = vhost_vdpa_set_config(v, argp);
464 		break;
465 	case VHOST_GET_FEATURES:
466 		r = vhost_vdpa_get_features(v, argp);
467 		break;
468 	case VHOST_SET_FEATURES:
469 		r = vhost_vdpa_set_features(v, argp);
470 		break;
471 	case VHOST_VDPA_GET_VRING_NUM:
472 		r = vhost_vdpa_get_vring_num(v, argp);
473 		break;
474 	case VHOST_SET_LOG_BASE:
475 	case VHOST_SET_LOG_FD:
476 		r = -ENOIOCTLCMD;
477 		break;
478 	case VHOST_VDPA_SET_CONFIG_CALL:
479 		r = vhost_vdpa_set_config_call(v, argp);
480 		break;
481 	case VHOST_GET_BACKEND_FEATURES:
482 		features = VHOST_VDPA_BACKEND_FEATURES;
483 		if (copy_to_user(featurep, &features, sizeof(features)))
484 			r = -EFAULT;
485 		break;
486 	case VHOST_VDPA_GET_IOVA_RANGE:
487 		r = vhost_vdpa_get_iova_range(v, argp);
488 		break;
489 	default:
490 		r = vhost_dev_ioctl(&v->vdev, cmd, argp);
491 		if (r == -ENOIOCTLCMD)
492 			r = vhost_vdpa_vring_ioctl(v, cmd, argp);
493 		break;
494 	}
495 
496 	mutex_unlock(&d->mutex);
497 	return r;
498 }
499 
500 static void vhost_vdpa_iotlb_unmap(struct vhost_vdpa *v, u64 start, u64 last)
501 {
502 	struct vhost_dev *dev = &v->vdev;
503 	struct vhost_iotlb *iotlb = dev->iotlb;
504 	struct vhost_iotlb_map *map;
505 	struct page *page;
506 	unsigned long pfn, pinned;
507 
508 	while ((map = vhost_iotlb_itree_first(iotlb, start, last)) != NULL) {
509 		pinned = map->size >> PAGE_SHIFT;
510 		for (pfn = map->addr >> PAGE_SHIFT;
511 		     pinned > 0; pfn++, pinned--) {
512 			page = pfn_to_page(pfn);
513 			if (map->perm & VHOST_ACCESS_WO)
514 				set_page_dirty_lock(page);
515 			unpin_user_page(page);
516 		}
517 		atomic64_sub(map->size >> PAGE_SHIFT, &dev->mm->pinned_vm);
518 		vhost_iotlb_map_free(iotlb, map);
519 	}
520 }
521 
522 static void vhost_vdpa_iotlb_free(struct vhost_vdpa *v)
523 {
524 	struct vhost_dev *dev = &v->vdev;
525 
526 	vhost_vdpa_iotlb_unmap(v, 0ULL, 0ULL - 1);
527 	kfree(dev->iotlb);
528 	dev->iotlb = NULL;
529 }
530 
531 static int perm_to_iommu_flags(u32 perm)
532 {
533 	int flags = 0;
534 
535 	switch (perm) {
536 	case VHOST_ACCESS_WO:
537 		flags |= IOMMU_WRITE;
538 		break;
539 	case VHOST_ACCESS_RO:
540 		flags |= IOMMU_READ;
541 		break;
542 	case VHOST_ACCESS_RW:
543 		flags |= (IOMMU_WRITE | IOMMU_READ);
544 		break;
545 	default:
546 		WARN(1, "invalidate vhost IOTLB permission\n");
547 		break;
548 	}
549 
550 	return flags | IOMMU_CACHE;
551 }
552 
553 static int vhost_vdpa_map(struct vhost_vdpa *v,
554 			  u64 iova, u64 size, u64 pa, u32 perm)
555 {
556 	struct vhost_dev *dev = &v->vdev;
557 	struct vdpa_device *vdpa = v->vdpa;
558 	const struct vdpa_config_ops *ops = vdpa->config;
559 	int r = 0;
560 
561 	r = vhost_iotlb_add_range(dev->iotlb, iova, iova + size - 1,
562 				  pa, perm);
563 	if (r)
564 		return r;
565 
566 	if (ops->dma_map) {
567 		r = ops->dma_map(vdpa, iova, size, pa, perm);
568 	} else if (ops->set_map) {
569 		if (!v->in_batch)
570 			r = ops->set_map(vdpa, dev->iotlb);
571 	} else {
572 		r = iommu_map(v->domain, iova, pa, size,
573 			      perm_to_iommu_flags(perm));
574 	}
575 
576 	if (r)
577 		vhost_iotlb_del_range(dev->iotlb, iova, iova + size - 1);
578 	else
579 		atomic64_add(size >> PAGE_SHIFT, &dev->mm->pinned_vm);
580 
581 	return r;
582 }
583 
584 static void vhost_vdpa_unmap(struct vhost_vdpa *v, u64 iova, u64 size)
585 {
586 	struct vhost_dev *dev = &v->vdev;
587 	struct vdpa_device *vdpa = v->vdpa;
588 	const struct vdpa_config_ops *ops = vdpa->config;
589 
590 	vhost_vdpa_iotlb_unmap(v, iova, iova + size - 1);
591 
592 	if (ops->dma_map) {
593 		ops->dma_unmap(vdpa, iova, size);
594 	} else if (ops->set_map) {
595 		if (!v->in_batch)
596 			ops->set_map(vdpa, dev->iotlb);
597 	} else {
598 		iommu_unmap(v->domain, iova, size);
599 	}
600 }
601 
602 static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
603 					   struct vhost_iotlb_msg *msg)
604 {
605 	struct vhost_dev *dev = &v->vdev;
606 	struct vhost_iotlb *iotlb = dev->iotlb;
607 	struct page **page_list;
608 	unsigned long list_size = PAGE_SIZE / sizeof(struct page *);
609 	unsigned int gup_flags = FOLL_LONGTERM;
610 	unsigned long npages, cur_base, map_pfn, last_pfn = 0;
611 	unsigned long lock_limit, sz2pin, nchunks, i;
612 	u64 iova = msg->iova;
613 	long pinned;
614 	int ret = 0;
615 
616 	if (msg->iova < v->range.first ||
617 	    msg->iova + msg->size - 1 > v->range.last)
618 		return -EINVAL;
619 
620 	if (vhost_iotlb_itree_first(iotlb, msg->iova,
621 				    msg->iova + msg->size - 1))
622 		return -EEXIST;
623 
624 	/* Limit the use of memory for bookkeeping */
625 	page_list = (struct page **) __get_free_page(GFP_KERNEL);
626 	if (!page_list)
627 		return -ENOMEM;
628 
629 	if (msg->perm & VHOST_ACCESS_WO)
630 		gup_flags |= FOLL_WRITE;
631 
632 	npages = PAGE_ALIGN(msg->size + (iova & ~PAGE_MASK)) >> PAGE_SHIFT;
633 	if (!npages) {
634 		ret = -EINVAL;
635 		goto free;
636 	}
637 
638 	mmap_read_lock(dev->mm);
639 
640 	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
641 	if (npages + atomic64_read(&dev->mm->pinned_vm) > lock_limit) {
642 		ret = -ENOMEM;
643 		goto unlock;
644 	}
645 
646 	cur_base = msg->uaddr & PAGE_MASK;
647 	iova &= PAGE_MASK;
648 	nchunks = 0;
649 
650 	while (npages) {
651 		sz2pin = min_t(unsigned long, npages, list_size);
652 		pinned = pin_user_pages(cur_base, sz2pin,
653 					gup_flags, page_list, NULL);
654 		if (sz2pin != pinned) {
655 			if (pinned < 0) {
656 				ret = pinned;
657 			} else {
658 				unpin_user_pages(page_list, pinned);
659 				ret = -ENOMEM;
660 			}
661 			goto out;
662 		}
663 		nchunks++;
664 
665 		if (!last_pfn)
666 			map_pfn = page_to_pfn(page_list[0]);
667 
668 		for (i = 0; i < pinned; i++) {
669 			unsigned long this_pfn = page_to_pfn(page_list[i]);
670 			u64 csize;
671 
672 			if (last_pfn && (this_pfn != last_pfn + 1)) {
673 				/* Pin a contiguous chunk of memory */
674 				csize = (last_pfn - map_pfn + 1) << PAGE_SHIFT;
675 				ret = vhost_vdpa_map(v, iova, csize,
676 						     map_pfn << PAGE_SHIFT,
677 						     msg->perm);
678 				if (ret) {
679 					/*
680 					 * Unpin the pages that are left unmapped
681 					 * from this point on in the current
682 					 * page_list. The remaining outstanding
683 					 * ones which may stride across several
684 					 * chunks will be covered in the common
685 					 * error path subsequently.
686 					 */
687 					unpin_user_pages(&page_list[i],
688 							 pinned - i);
689 					goto out;
690 				}
691 
692 				map_pfn = this_pfn;
693 				iova += csize;
694 				nchunks = 0;
695 			}
696 
697 			last_pfn = this_pfn;
698 		}
699 
700 		cur_base += pinned << PAGE_SHIFT;
701 		npages -= pinned;
702 	}
703 
704 	/* Pin the rest chunk */
705 	ret = vhost_vdpa_map(v, iova, (last_pfn - map_pfn + 1) << PAGE_SHIFT,
706 			     map_pfn << PAGE_SHIFT, msg->perm);
707 out:
708 	if (ret) {
709 		if (nchunks) {
710 			unsigned long pfn;
711 
712 			/*
713 			 * Unpin the outstanding pages which are yet to be
714 			 * mapped but haven't due to vdpa_map() or
715 			 * pin_user_pages() failure.
716 			 *
717 			 * Mapped pages are accounted in vdpa_map(), hence
718 			 * the corresponding unpinning will be handled by
719 			 * vdpa_unmap().
720 			 */
721 			WARN_ON(!last_pfn);
722 			for (pfn = map_pfn; pfn <= last_pfn; pfn++)
723 				unpin_user_page(pfn_to_page(pfn));
724 		}
725 		vhost_vdpa_unmap(v, msg->iova, msg->size);
726 	}
727 unlock:
728 	mmap_read_unlock(dev->mm);
729 free:
730 	free_page((unsigned long)page_list);
731 	return ret;
732 }
733 
734 static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev,
735 					struct vhost_iotlb_msg *msg)
736 {
737 	struct vhost_vdpa *v = container_of(dev, struct vhost_vdpa, vdev);
738 	struct vdpa_device *vdpa = v->vdpa;
739 	const struct vdpa_config_ops *ops = vdpa->config;
740 	int r = 0;
741 
742 	r = vhost_dev_check_owner(dev);
743 	if (r)
744 		return r;
745 
746 	switch (msg->type) {
747 	case VHOST_IOTLB_UPDATE:
748 		r = vhost_vdpa_process_iotlb_update(v, msg);
749 		break;
750 	case VHOST_IOTLB_INVALIDATE:
751 		vhost_vdpa_unmap(v, msg->iova, msg->size);
752 		break;
753 	case VHOST_IOTLB_BATCH_BEGIN:
754 		v->in_batch = true;
755 		break;
756 	case VHOST_IOTLB_BATCH_END:
757 		if (v->in_batch && ops->set_map)
758 			ops->set_map(vdpa, dev->iotlb);
759 		v->in_batch = false;
760 		break;
761 	default:
762 		r = -EINVAL;
763 		break;
764 	}
765 
766 	return r;
767 }
768 
769 static ssize_t vhost_vdpa_chr_write_iter(struct kiocb *iocb,
770 					 struct iov_iter *from)
771 {
772 	struct file *file = iocb->ki_filp;
773 	struct vhost_vdpa *v = file->private_data;
774 	struct vhost_dev *dev = &v->vdev;
775 
776 	return vhost_chr_write_iter(dev, from);
777 }
778 
779 static int vhost_vdpa_alloc_domain(struct vhost_vdpa *v)
780 {
781 	struct vdpa_device *vdpa = v->vdpa;
782 	const struct vdpa_config_ops *ops = vdpa->config;
783 	struct device *dma_dev = vdpa_get_dma_dev(vdpa);
784 	struct bus_type *bus;
785 	int ret;
786 
787 	/* Device want to do DMA by itself */
788 	if (ops->set_map || ops->dma_map)
789 		return 0;
790 
791 	bus = dma_dev->bus;
792 	if (!bus)
793 		return -EFAULT;
794 
795 	if (!iommu_capable(bus, IOMMU_CAP_CACHE_COHERENCY))
796 		return -ENOTSUPP;
797 
798 	v->domain = iommu_domain_alloc(bus);
799 	if (!v->domain)
800 		return -EIO;
801 
802 	ret = iommu_attach_device(v->domain, dma_dev);
803 	if (ret)
804 		goto err_attach;
805 
806 	return 0;
807 
808 err_attach:
809 	iommu_domain_free(v->domain);
810 	return ret;
811 }
812 
813 static void vhost_vdpa_free_domain(struct vhost_vdpa *v)
814 {
815 	struct vdpa_device *vdpa = v->vdpa;
816 	struct device *dma_dev = vdpa_get_dma_dev(vdpa);
817 
818 	if (v->domain) {
819 		iommu_detach_device(v->domain, dma_dev);
820 		iommu_domain_free(v->domain);
821 	}
822 
823 	v->domain = NULL;
824 }
825 
826 static void vhost_vdpa_set_iova_range(struct vhost_vdpa *v)
827 {
828 	struct vdpa_iova_range *range = &v->range;
829 	struct iommu_domain_geometry geo;
830 	struct vdpa_device *vdpa = v->vdpa;
831 	const struct vdpa_config_ops *ops = vdpa->config;
832 
833 	if (ops->get_iova_range) {
834 		*range = ops->get_iova_range(vdpa);
835 	} else if (v->domain &&
836 		   !iommu_domain_get_attr(v->domain,
837 		   DOMAIN_ATTR_GEOMETRY, &geo) &&
838 		   geo.force_aperture) {
839 		range->first = geo.aperture_start;
840 		range->last = geo.aperture_end;
841 	} else {
842 		range->first = 0;
843 		range->last = ULLONG_MAX;
844 	}
845 }
846 
847 static int vhost_vdpa_open(struct inode *inode, struct file *filep)
848 {
849 	struct vhost_vdpa *v;
850 	struct vhost_dev *dev;
851 	struct vhost_virtqueue **vqs;
852 	int nvqs, i, r, opened;
853 
854 	v = container_of(inode->i_cdev, struct vhost_vdpa, cdev);
855 
856 	opened = atomic_cmpxchg(&v->opened, 0, 1);
857 	if (opened)
858 		return -EBUSY;
859 
860 	nvqs = v->nvqs;
861 	vhost_vdpa_reset(v);
862 
863 	vqs = kmalloc_array(nvqs, sizeof(*vqs), GFP_KERNEL);
864 	if (!vqs) {
865 		r = -ENOMEM;
866 		goto err;
867 	}
868 
869 	dev = &v->vdev;
870 	for (i = 0; i < nvqs; i++) {
871 		vqs[i] = &v->vqs[i];
872 		vqs[i]->handle_kick = handle_vq_kick;
873 	}
874 	vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, false,
875 		       vhost_vdpa_process_iotlb_msg);
876 
877 	dev->iotlb = vhost_iotlb_alloc(0, 0);
878 	if (!dev->iotlb) {
879 		r = -ENOMEM;
880 		goto err_init_iotlb;
881 	}
882 
883 	r = vhost_vdpa_alloc_domain(v);
884 	if (r)
885 		goto err_init_iotlb;
886 
887 	vhost_vdpa_set_iova_range(v);
888 
889 	filep->private_data = v;
890 
891 	return 0;
892 
893 err_init_iotlb:
894 	vhost_dev_cleanup(&v->vdev);
895 	kfree(vqs);
896 err:
897 	atomic_dec(&v->opened);
898 	return r;
899 }
900 
901 static void vhost_vdpa_clean_irq(struct vhost_vdpa *v)
902 {
903 	struct vhost_virtqueue *vq;
904 	int i;
905 
906 	for (i = 0; i < v->nvqs; i++) {
907 		vq = &v->vqs[i];
908 		if (vq->call_ctx.producer.irq)
909 			irq_bypass_unregister_producer(&vq->call_ctx.producer);
910 	}
911 }
912 
913 static int vhost_vdpa_release(struct inode *inode, struct file *filep)
914 {
915 	struct vhost_vdpa *v = filep->private_data;
916 	struct vhost_dev *d = &v->vdev;
917 
918 	mutex_lock(&d->mutex);
919 	filep->private_data = NULL;
920 	vhost_vdpa_reset(v);
921 	vhost_dev_stop(&v->vdev);
922 	vhost_vdpa_iotlb_free(v);
923 	vhost_vdpa_free_domain(v);
924 	vhost_vdpa_config_put(v);
925 	vhost_vdpa_clean_irq(v);
926 	vhost_dev_cleanup(&v->vdev);
927 	kfree(v->vdev.vqs);
928 	mutex_unlock(&d->mutex);
929 
930 	atomic_dec(&v->opened);
931 	complete(&v->completion);
932 
933 	return 0;
934 }
935 
936 #ifdef CONFIG_MMU
937 static vm_fault_t vhost_vdpa_fault(struct vm_fault *vmf)
938 {
939 	struct vhost_vdpa *v = vmf->vma->vm_file->private_data;
940 	struct vdpa_device *vdpa = v->vdpa;
941 	const struct vdpa_config_ops *ops = vdpa->config;
942 	struct vdpa_notification_area notify;
943 	struct vm_area_struct *vma = vmf->vma;
944 	u16 index = vma->vm_pgoff;
945 
946 	notify = ops->get_vq_notification(vdpa, index);
947 
948 	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
949 	if (remap_pfn_range(vma, vmf->address & PAGE_MASK,
950 			    notify.addr >> PAGE_SHIFT, PAGE_SIZE,
951 			    vma->vm_page_prot))
952 		return VM_FAULT_SIGBUS;
953 
954 	return VM_FAULT_NOPAGE;
955 }
956 
957 static const struct vm_operations_struct vhost_vdpa_vm_ops = {
958 	.fault = vhost_vdpa_fault,
959 };
960 
961 static int vhost_vdpa_mmap(struct file *file, struct vm_area_struct *vma)
962 {
963 	struct vhost_vdpa *v = vma->vm_file->private_data;
964 	struct vdpa_device *vdpa = v->vdpa;
965 	const struct vdpa_config_ops *ops = vdpa->config;
966 	struct vdpa_notification_area notify;
967 	unsigned long index = vma->vm_pgoff;
968 
969 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
970 		return -EINVAL;
971 	if ((vma->vm_flags & VM_SHARED) == 0)
972 		return -EINVAL;
973 	if (vma->vm_flags & VM_READ)
974 		return -EINVAL;
975 	if (index > 65535)
976 		return -EINVAL;
977 	if (!ops->get_vq_notification)
978 		return -ENOTSUPP;
979 
980 	/* To be safe and easily modelled by userspace, We only
981 	 * support the doorbell which sits on the page boundary and
982 	 * does not share the page with other registers.
983 	 */
984 	notify = ops->get_vq_notification(vdpa, index);
985 	if (notify.addr & (PAGE_SIZE - 1))
986 		return -EINVAL;
987 	if (vma->vm_end - vma->vm_start != notify.size)
988 		return -ENOTSUPP;
989 
990 	vma->vm_ops = &vhost_vdpa_vm_ops;
991 	return 0;
992 }
993 #endif /* CONFIG_MMU */
994 
995 static const struct file_operations vhost_vdpa_fops = {
996 	.owner		= THIS_MODULE,
997 	.open		= vhost_vdpa_open,
998 	.release	= vhost_vdpa_release,
999 	.write_iter	= vhost_vdpa_chr_write_iter,
1000 	.unlocked_ioctl	= vhost_vdpa_unlocked_ioctl,
1001 #ifdef CONFIG_MMU
1002 	.mmap		= vhost_vdpa_mmap,
1003 #endif /* CONFIG_MMU */
1004 	.compat_ioctl	= compat_ptr_ioctl,
1005 };
1006 
1007 static void vhost_vdpa_release_dev(struct device *device)
1008 {
1009 	struct vhost_vdpa *v =
1010 	       container_of(device, struct vhost_vdpa, dev);
1011 
1012 	ida_simple_remove(&vhost_vdpa_ida, v->minor);
1013 	kfree(v->vqs);
1014 	kfree(v);
1015 }
1016 
1017 static int vhost_vdpa_probe(struct vdpa_device *vdpa)
1018 {
1019 	const struct vdpa_config_ops *ops = vdpa->config;
1020 	struct vhost_vdpa *v;
1021 	int minor;
1022 	int r;
1023 
1024 	/* Currently, we only accept the network devices. */
1025 	if (ops->get_device_id(vdpa) != VIRTIO_ID_NET)
1026 		return -ENOTSUPP;
1027 
1028 	v = kzalloc(sizeof(*v), GFP_KERNEL | __GFP_RETRY_MAYFAIL);
1029 	if (!v)
1030 		return -ENOMEM;
1031 
1032 	minor = ida_simple_get(&vhost_vdpa_ida, 0,
1033 			       VHOST_VDPA_DEV_MAX, GFP_KERNEL);
1034 	if (minor < 0) {
1035 		kfree(v);
1036 		return minor;
1037 	}
1038 
1039 	atomic_set(&v->opened, 0);
1040 	v->minor = minor;
1041 	v->vdpa = vdpa;
1042 	v->nvqs = vdpa->nvqs;
1043 	v->virtio_id = ops->get_device_id(vdpa);
1044 
1045 	device_initialize(&v->dev);
1046 	v->dev.release = vhost_vdpa_release_dev;
1047 	v->dev.parent = &vdpa->dev;
1048 	v->dev.devt = MKDEV(MAJOR(vhost_vdpa_major), minor);
1049 	v->vqs = kmalloc_array(v->nvqs, sizeof(struct vhost_virtqueue),
1050 			       GFP_KERNEL);
1051 	if (!v->vqs) {
1052 		r = -ENOMEM;
1053 		goto err;
1054 	}
1055 
1056 	r = dev_set_name(&v->dev, "vhost-vdpa-%u", minor);
1057 	if (r)
1058 		goto err;
1059 
1060 	cdev_init(&v->cdev, &vhost_vdpa_fops);
1061 	v->cdev.owner = THIS_MODULE;
1062 
1063 	r = cdev_device_add(&v->cdev, &v->dev);
1064 	if (r)
1065 		goto err;
1066 
1067 	init_completion(&v->completion);
1068 	vdpa_set_drvdata(vdpa, v);
1069 
1070 	return 0;
1071 
1072 err:
1073 	put_device(&v->dev);
1074 	return r;
1075 }
1076 
1077 static void vhost_vdpa_remove(struct vdpa_device *vdpa)
1078 {
1079 	struct vhost_vdpa *v = vdpa_get_drvdata(vdpa);
1080 	int opened;
1081 
1082 	cdev_device_del(&v->cdev, &v->dev);
1083 
1084 	do {
1085 		opened = atomic_cmpxchg(&v->opened, 0, 1);
1086 		if (!opened)
1087 			break;
1088 		wait_for_completion(&v->completion);
1089 	} while (1);
1090 
1091 	put_device(&v->dev);
1092 }
1093 
1094 static struct vdpa_driver vhost_vdpa_driver = {
1095 	.driver = {
1096 		.name	= "vhost_vdpa",
1097 	},
1098 	.probe	= vhost_vdpa_probe,
1099 	.remove	= vhost_vdpa_remove,
1100 };
1101 
1102 static int __init vhost_vdpa_init(void)
1103 {
1104 	int r;
1105 
1106 	r = alloc_chrdev_region(&vhost_vdpa_major, 0, VHOST_VDPA_DEV_MAX,
1107 				"vhost-vdpa");
1108 	if (r)
1109 		goto err_alloc_chrdev;
1110 
1111 	r = vdpa_register_driver(&vhost_vdpa_driver);
1112 	if (r)
1113 		goto err_vdpa_register_driver;
1114 
1115 	return 0;
1116 
1117 err_vdpa_register_driver:
1118 	unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1119 err_alloc_chrdev:
1120 	return r;
1121 }
1122 module_init(vhost_vdpa_init);
1123 
1124 static void __exit vhost_vdpa_exit(void)
1125 {
1126 	vdpa_unregister_driver(&vhost_vdpa_driver);
1127 	unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1128 }
1129 module_exit(vhost_vdpa_exit);
1130 
1131 MODULE_VERSION("0.0.1");
1132 MODULE_LICENSE("GPL v2");
1133 MODULE_AUTHOR("Intel Corporation");
1134 MODULE_DESCRIPTION("vDPA-based vhost backend for virtio");
1135