xref: /openbmc/linux/drivers/vhost/vdpa.c (revision a9d85efb) 
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/slab.h>
20 #include <linux/iommu.h>
21 #include <linux/uuid.h>
22 #include <linux/vdpa.h>
23 #include <linux/nospec.h>
24 #include <linux/vhost.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 int vhost_vdpa_reset(struct vhost_vdpa *v)
120 {
121 	struct vdpa_device *vdpa = v->vdpa;
122 
123 	v->in_batch = 0;
124 
125 	return vdpa_reset(vdpa);
126 }
127 
128 static long vhost_vdpa_get_device_id(struct vhost_vdpa *v, u8 __user *argp)
129 {
130 	struct vdpa_device *vdpa = v->vdpa;
131 	const struct vdpa_config_ops *ops = vdpa->config;
132 	u32 device_id;
133 
134 	device_id = ops->get_device_id(vdpa);
135 
136 	if (copy_to_user(argp, &device_id, sizeof(device_id)))
137 		return -EFAULT;
138 
139 	return 0;
140 }
141 
142 static long vhost_vdpa_get_status(struct vhost_vdpa *v, u8 __user *statusp)
143 {
144 	struct vdpa_device *vdpa = v->vdpa;
145 	const struct vdpa_config_ops *ops = vdpa->config;
146 	u8 status;
147 
148 	status = ops->get_status(vdpa);
149 
150 	if (copy_to_user(statusp, &status, sizeof(status)))
151 		return -EFAULT;
152 
153 	return 0;
154 }
155 
156 static long vhost_vdpa_set_status(struct vhost_vdpa *v, u8 __user *statusp)
157 {
158 	struct vdpa_device *vdpa = v->vdpa;
159 	const struct vdpa_config_ops *ops = vdpa->config;
160 	u8 status, status_old;
161 	int ret, nvqs = v->nvqs;
162 	u16 i;
163 
164 	if (copy_from_user(&status, statusp, sizeof(status)))
165 		return -EFAULT;
166 
167 	status_old = ops->get_status(vdpa);
168 
169 	/*
170 	 * Userspace shouldn't remove status bits unless reset the
171 	 * status to 0.
172 	 */
173 	if (status != 0 && (ops->get_status(vdpa) & ~status) != 0)
174 		return -EINVAL;
175 
176 	if (status == 0) {
177 		ret = ops->reset(vdpa);
178 		if (ret)
179 			return ret;
180 	} else
181 		ops->set_status(vdpa, status);
182 
183 	if ((status & VIRTIO_CONFIG_S_DRIVER_OK) && !(status_old & VIRTIO_CONFIG_S_DRIVER_OK))
184 		for (i = 0; i < nvqs; i++)
185 			vhost_vdpa_setup_vq_irq(v, i);
186 
187 	if ((status_old & VIRTIO_CONFIG_S_DRIVER_OK) && !(status & VIRTIO_CONFIG_S_DRIVER_OK))
188 		for (i = 0; i < nvqs; i++)
189 			vhost_vdpa_unsetup_vq_irq(v, i);
190 
191 	return 0;
192 }
193 
194 static int vhost_vdpa_config_validate(struct vhost_vdpa *v,
195 				      struct vhost_vdpa_config *c)
196 {
197 	struct vdpa_device *vdpa = v->vdpa;
198 	long size = vdpa->config->get_config_size(vdpa);
199 
200 	if (c->len == 0)
201 		return -EINVAL;
202 
203 	if (c->len > size - c->off)
204 		return -E2BIG;
205 
206 	return 0;
207 }
208 
209 static long vhost_vdpa_get_config(struct vhost_vdpa *v,
210 				  struct vhost_vdpa_config __user *c)
211 {
212 	struct vdpa_device *vdpa = v->vdpa;
213 	struct vhost_vdpa_config config;
214 	unsigned long size = offsetof(struct vhost_vdpa_config, buf);
215 	u8 *buf;
216 
217 	if (copy_from_user(&config, c, size))
218 		return -EFAULT;
219 	if (vhost_vdpa_config_validate(v, &config))
220 		return -EINVAL;
221 	buf = kvzalloc(config.len, GFP_KERNEL);
222 	if (!buf)
223 		return -ENOMEM;
224 
225 	vdpa_get_config(vdpa, config.off, buf, config.len);
226 
227 	if (copy_to_user(c->buf, buf, config.len)) {
228 		kvfree(buf);
229 		return -EFAULT;
230 	}
231 
232 	kvfree(buf);
233 	return 0;
234 }
235 
236 static long vhost_vdpa_set_config(struct vhost_vdpa *v,
237 				  struct vhost_vdpa_config __user *c)
238 {
239 	struct vdpa_device *vdpa = v->vdpa;
240 	const struct vdpa_config_ops *ops = vdpa->config;
241 	struct vhost_vdpa_config config;
242 	unsigned long size = offsetof(struct vhost_vdpa_config, buf);
243 	u8 *buf;
244 
245 	if (copy_from_user(&config, c, size))
246 		return -EFAULT;
247 	if (vhost_vdpa_config_validate(v, &config))
248 		return -EINVAL;
249 
250 	buf = vmemdup_user(c->buf, config.len);
251 	if (IS_ERR(buf))
252 		return PTR_ERR(buf);
253 
254 	ops->set_config(vdpa, config.off, buf, config.len);
255 
256 	kvfree(buf);
257 	return 0;
258 }
259 
260 static long vhost_vdpa_get_features(struct vhost_vdpa *v, u64 __user *featurep)
261 {
262 	struct vdpa_device *vdpa = v->vdpa;
263 	const struct vdpa_config_ops *ops = vdpa->config;
264 	u64 features;
265 
266 	features = ops->get_features(vdpa);
267 
268 	if (copy_to_user(featurep, &features, sizeof(features)))
269 		return -EFAULT;
270 
271 	return 0;
272 }
273 
274 static long vhost_vdpa_set_features(struct vhost_vdpa *v, u64 __user *featurep)
275 {
276 	struct vdpa_device *vdpa = v->vdpa;
277 	const struct vdpa_config_ops *ops = vdpa->config;
278 	u64 features;
279 
280 	/*
281 	 * It's not allowed to change the features after they have
282 	 * been negotiated.
283 	 */
284 	if (ops->get_status(vdpa) & VIRTIO_CONFIG_S_FEATURES_OK)
285 		return -EBUSY;
286 
287 	if (copy_from_user(&features, featurep, sizeof(features)))
288 		return -EFAULT;
289 
290 	if (vdpa_set_features(vdpa, features))
291 		return -EINVAL;
292 
293 	return 0;
294 }
295 
296 static long vhost_vdpa_get_vring_num(struct vhost_vdpa *v, u16 __user *argp)
297 {
298 	struct vdpa_device *vdpa = v->vdpa;
299 	const struct vdpa_config_ops *ops = vdpa->config;
300 	u16 num;
301 
302 	num = ops->get_vq_num_max(vdpa);
303 
304 	if (copy_to_user(argp, &num, sizeof(num)))
305 		return -EFAULT;
306 
307 	return 0;
308 }
309 
310 static void vhost_vdpa_config_put(struct vhost_vdpa *v)
311 {
312 	if (v->config_ctx) {
313 		eventfd_ctx_put(v->config_ctx);
314 		v->config_ctx = NULL;
315 	}
316 }
317 
318 static long vhost_vdpa_set_config_call(struct vhost_vdpa *v, u32 __user *argp)
319 {
320 	struct vdpa_callback cb;
321 	int fd;
322 	struct eventfd_ctx *ctx;
323 
324 	cb.callback = vhost_vdpa_config_cb;
325 	cb.private = v->vdpa;
326 	if (copy_from_user(&fd, argp, sizeof(fd)))
327 		return  -EFAULT;
328 
329 	ctx = fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(fd);
330 	swap(ctx, v->config_ctx);
331 
332 	if (!IS_ERR_OR_NULL(ctx))
333 		eventfd_ctx_put(ctx);
334 
335 	if (IS_ERR(v->config_ctx)) {
336 		long ret = PTR_ERR(v->config_ctx);
337 
338 		v->config_ctx = NULL;
339 		return ret;
340 	}
341 
342 	v->vdpa->config->set_config_cb(v->vdpa, &cb);
343 
344 	return 0;
345 }
346 
347 static long vhost_vdpa_get_iova_range(struct vhost_vdpa *v, u32 __user *argp)
348 {
349 	struct vhost_vdpa_iova_range range = {
350 		.first = v->range.first,
351 		.last = v->range.last,
352 	};
353 
354 	if (copy_to_user(argp, &range, sizeof(range)))
355 		return -EFAULT;
356 	return 0;
357 }
358 
359 static long vhost_vdpa_vring_ioctl(struct vhost_vdpa *v, unsigned int cmd,
360 				   void __user *argp)
361 {
362 	struct vdpa_device *vdpa = v->vdpa;
363 	const struct vdpa_config_ops *ops = vdpa->config;
364 	struct vdpa_vq_state vq_state;
365 	struct vdpa_callback cb;
366 	struct vhost_virtqueue *vq;
367 	struct vhost_vring_state s;
368 	u32 idx;
369 	long r;
370 
371 	r = get_user(idx, (u32 __user *)argp);
372 	if (r < 0)
373 		return r;
374 
375 	if (idx >= v->nvqs)
376 		return -ENOBUFS;
377 
378 	idx = array_index_nospec(idx, v->nvqs);
379 	vq = &v->vqs[idx];
380 
381 	switch (cmd) {
382 	case VHOST_VDPA_SET_VRING_ENABLE:
383 		if (copy_from_user(&s, argp, sizeof(s)))
384 			return -EFAULT;
385 		ops->set_vq_ready(vdpa, idx, s.num);
386 		return 0;
387 	case VHOST_GET_VRING_BASE:
388 		r = ops->get_vq_state(v->vdpa, idx, &vq_state);
389 		if (r)
390 			return r;
391 
392 		vq->last_avail_idx = vq_state.split.avail_index;
393 		break;
394 	}
395 
396 	r = vhost_vring_ioctl(&v->vdev, cmd, argp);
397 	if (r)
398 		return r;
399 
400 	switch (cmd) {
401 	case VHOST_SET_VRING_ADDR:
402 		if (ops->set_vq_address(vdpa, idx,
403 					(u64)(uintptr_t)vq->desc,
404 					(u64)(uintptr_t)vq->avail,
405 					(u64)(uintptr_t)vq->used))
406 			r = -EINVAL;
407 		break;
408 
409 	case VHOST_SET_VRING_BASE:
410 		vq_state.split.avail_index = vq->last_avail_idx;
411 		if (ops->set_vq_state(vdpa, idx, &vq_state))
412 			r = -EINVAL;
413 		break;
414 
415 	case VHOST_SET_VRING_CALL:
416 		if (vq->call_ctx.ctx) {
417 			cb.callback = vhost_vdpa_virtqueue_cb;
418 			cb.private = vq;
419 		} else {
420 			cb.callback = NULL;
421 			cb.private = NULL;
422 		}
423 		ops->set_vq_cb(vdpa, idx, &cb);
424 		vhost_vdpa_setup_vq_irq(v, idx);
425 		break;
426 
427 	case VHOST_SET_VRING_NUM:
428 		ops->set_vq_num(vdpa, idx, vq->num);
429 		break;
430 	}
431 
432 	return r;
433 }
434 
435 static long vhost_vdpa_unlocked_ioctl(struct file *filep,
436 				      unsigned int cmd, unsigned long arg)
437 {
438 	struct vhost_vdpa *v = filep->private_data;
439 	struct vhost_dev *d = &v->vdev;
440 	void __user *argp = (void __user *)arg;
441 	u64 __user *featurep = argp;
442 	u64 features;
443 	long r = 0;
444 
445 	if (cmd == VHOST_SET_BACKEND_FEATURES) {
446 		if (copy_from_user(&features, featurep, sizeof(features)))
447 			return -EFAULT;
448 		if (features & ~VHOST_VDPA_BACKEND_FEATURES)
449 			return -EOPNOTSUPP;
450 		vhost_set_backend_features(&v->vdev, features);
451 		return 0;
452 	}
453 
454 	mutex_lock(&d->mutex);
455 
456 	switch (cmd) {
457 	case VHOST_VDPA_GET_DEVICE_ID:
458 		r = vhost_vdpa_get_device_id(v, argp);
459 		break;
460 	case VHOST_VDPA_GET_STATUS:
461 		r = vhost_vdpa_get_status(v, argp);
462 		break;
463 	case VHOST_VDPA_SET_STATUS:
464 		r = vhost_vdpa_set_status(v, argp);
465 		break;
466 	case VHOST_VDPA_GET_CONFIG:
467 		r = vhost_vdpa_get_config(v, argp);
468 		break;
469 	case VHOST_VDPA_SET_CONFIG:
470 		r = vhost_vdpa_set_config(v, argp);
471 		break;
472 	case VHOST_GET_FEATURES:
473 		r = vhost_vdpa_get_features(v, argp);
474 		break;
475 	case VHOST_SET_FEATURES:
476 		r = vhost_vdpa_set_features(v, argp);
477 		break;
478 	case VHOST_VDPA_GET_VRING_NUM:
479 		r = vhost_vdpa_get_vring_num(v, argp);
480 		break;
481 	case VHOST_SET_LOG_BASE:
482 	case VHOST_SET_LOG_FD:
483 		r = -ENOIOCTLCMD;
484 		break;
485 	case VHOST_VDPA_SET_CONFIG_CALL:
486 		r = vhost_vdpa_set_config_call(v, argp);
487 		break;
488 	case VHOST_GET_BACKEND_FEATURES:
489 		features = VHOST_VDPA_BACKEND_FEATURES;
490 		if (copy_to_user(featurep, &features, sizeof(features)))
491 			r = -EFAULT;
492 		break;
493 	case VHOST_VDPA_GET_IOVA_RANGE:
494 		r = vhost_vdpa_get_iova_range(v, argp);
495 		break;
496 	default:
497 		r = vhost_dev_ioctl(&v->vdev, cmd, argp);
498 		if (r == -ENOIOCTLCMD)
499 			r = vhost_vdpa_vring_ioctl(v, cmd, argp);
500 		break;
501 	}
502 
503 	mutex_unlock(&d->mutex);
504 	return r;
505 }
506 
507 static void vhost_vdpa_pa_unmap(struct vhost_vdpa *v, u64 start, u64 last)
508 {
509 	struct vhost_dev *dev = &v->vdev;
510 	struct vhost_iotlb *iotlb = dev->iotlb;
511 	struct vhost_iotlb_map *map;
512 	struct page *page;
513 	unsigned long pfn, pinned;
514 
515 	while ((map = vhost_iotlb_itree_first(iotlb, start, last)) != NULL) {
516 		pinned = PFN_DOWN(map->size);
517 		for (pfn = PFN_DOWN(map->addr);
518 		     pinned > 0; pfn++, pinned--) {
519 			page = pfn_to_page(pfn);
520 			if (map->perm & VHOST_ACCESS_WO)
521 				set_page_dirty_lock(page);
522 			unpin_user_page(page);
523 		}
524 		atomic64_sub(PFN_DOWN(map->size), &dev->mm->pinned_vm);
525 		vhost_iotlb_map_free(iotlb, map);
526 	}
527 }
528 
529 static void vhost_vdpa_va_unmap(struct vhost_vdpa *v, u64 start, u64 last)
530 {
531 	struct vhost_dev *dev = &v->vdev;
532 	struct vhost_iotlb *iotlb = dev->iotlb;
533 	struct vhost_iotlb_map *map;
534 	struct vdpa_map_file *map_file;
535 
536 	while ((map = vhost_iotlb_itree_first(iotlb, start, last)) != NULL) {
537 		map_file = (struct vdpa_map_file *)map->opaque;
538 		fput(map_file->file);
539 		kfree(map_file);
540 		vhost_iotlb_map_free(iotlb, map);
541 	}
542 }
543 
544 static void vhost_vdpa_iotlb_unmap(struct vhost_vdpa *v, u64 start, u64 last)
545 {
546 	struct vdpa_device *vdpa = v->vdpa;
547 
548 	if (vdpa->use_va)
549 		return vhost_vdpa_va_unmap(v, start, last);
550 
551 	return vhost_vdpa_pa_unmap(v, start, last);
552 }
553 
554 static void vhost_vdpa_iotlb_free(struct vhost_vdpa *v)
555 {
556 	struct vhost_dev *dev = &v->vdev;
557 
558 	vhost_vdpa_iotlb_unmap(v, 0ULL, 0ULL - 1);
559 	kfree(dev->iotlb);
560 	dev->iotlb = NULL;
561 }
562 
563 static int perm_to_iommu_flags(u32 perm)
564 {
565 	int flags = 0;
566 
567 	switch (perm) {
568 	case VHOST_ACCESS_WO:
569 		flags |= IOMMU_WRITE;
570 		break;
571 	case VHOST_ACCESS_RO:
572 		flags |= IOMMU_READ;
573 		break;
574 	case VHOST_ACCESS_RW:
575 		flags |= (IOMMU_WRITE | IOMMU_READ);
576 		break;
577 	default:
578 		WARN(1, "invalidate vhost IOTLB permission\n");
579 		break;
580 	}
581 
582 	return flags | IOMMU_CACHE;
583 }
584 
585 static int vhost_vdpa_map(struct vhost_vdpa *v, u64 iova,
586 			  u64 size, u64 pa, u32 perm, void *opaque)
587 {
588 	struct vhost_dev *dev = &v->vdev;
589 	struct vdpa_device *vdpa = v->vdpa;
590 	const struct vdpa_config_ops *ops = vdpa->config;
591 	int r = 0;
592 
593 	r = vhost_iotlb_add_range_ctx(dev->iotlb, iova, iova + size - 1,
594 				      pa, perm, opaque);
595 	if (r)
596 		return r;
597 
598 	if (ops->dma_map) {
599 		r = ops->dma_map(vdpa, iova, size, pa, perm, opaque);
600 	} else if (ops->set_map) {
601 		if (!v->in_batch)
602 			r = ops->set_map(vdpa, dev->iotlb);
603 	} else {
604 		r = iommu_map(v->domain, iova, pa, size,
605 			      perm_to_iommu_flags(perm));
606 	}
607 	if (r) {
608 		vhost_iotlb_del_range(dev->iotlb, iova, iova + size - 1);
609 		return r;
610 	}
611 
612 	if (!vdpa->use_va)
613 		atomic64_add(PFN_DOWN(size), &dev->mm->pinned_vm);
614 
615 	return 0;
616 }
617 
618 static void vhost_vdpa_unmap(struct vhost_vdpa *v, u64 iova, u64 size)
619 {
620 	struct vhost_dev *dev = &v->vdev;
621 	struct vdpa_device *vdpa = v->vdpa;
622 	const struct vdpa_config_ops *ops = vdpa->config;
623 
624 	vhost_vdpa_iotlb_unmap(v, iova, iova + size - 1);
625 
626 	if (ops->dma_map) {
627 		ops->dma_unmap(vdpa, iova, size);
628 	} else if (ops->set_map) {
629 		if (!v->in_batch)
630 			ops->set_map(vdpa, dev->iotlb);
631 	} else {
632 		iommu_unmap(v->domain, iova, size);
633 	}
634 }
635 
636 static int vhost_vdpa_va_map(struct vhost_vdpa *v,
637 			     u64 iova, u64 size, u64 uaddr, u32 perm)
638 {
639 	struct vhost_dev *dev = &v->vdev;
640 	u64 offset, map_size, map_iova = iova;
641 	struct vdpa_map_file *map_file;
642 	struct vm_area_struct *vma;
643 	int ret = 0;
644 
645 	mmap_read_lock(dev->mm);
646 
647 	while (size) {
648 		vma = find_vma(dev->mm, uaddr);
649 		if (!vma) {
650 			ret = -EINVAL;
651 			break;
652 		}
653 		map_size = min(size, vma->vm_end - uaddr);
654 		if (!(vma->vm_file && (vma->vm_flags & VM_SHARED) &&
655 			!(vma->vm_flags & (VM_IO | VM_PFNMAP))))
656 			goto next;
657 
658 		map_file = kzalloc(sizeof(*map_file), GFP_KERNEL);
659 		if (!map_file) {
660 			ret = -ENOMEM;
661 			break;
662 		}
663 		offset = (vma->vm_pgoff << PAGE_SHIFT) + uaddr - vma->vm_start;
664 		map_file->offset = offset;
665 		map_file->file = get_file(vma->vm_file);
666 		ret = vhost_vdpa_map(v, map_iova, map_size, uaddr,
667 				     perm, map_file);
668 		if (ret) {
669 			fput(map_file->file);
670 			kfree(map_file);
671 			break;
672 		}
673 next:
674 		size -= map_size;
675 		uaddr += map_size;
676 		map_iova += map_size;
677 	}
678 	if (ret)
679 		vhost_vdpa_unmap(v, iova, map_iova - iova);
680 
681 	mmap_read_unlock(dev->mm);
682 
683 	return ret;
684 }
685 
686 static int vhost_vdpa_pa_map(struct vhost_vdpa *v,
687 			     u64 iova, u64 size, u64 uaddr, u32 perm)
688 {
689 	struct vhost_dev *dev = &v->vdev;
690 	struct page **page_list;
691 	unsigned long list_size = PAGE_SIZE / sizeof(struct page *);
692 	unsigned int gup_flags = FOLL_LONGTERM;
693 	unsigned long npages, cur_base, map_pfn, last_pfn = 0;
694 	unsigned long lock_limit, sz2pin, nchunks, i;
695 	u64 start = iova;
696 	long pinned;
697 	int ret = 0;
698 
699 	/* Limit the use of memory for bookkeeping */
700 	page_list = (struct page **) __get_free_page(GFP_KERNEL);
701 	if (!page_list)
702 		return -ENOMEM;
703 
704 	if (perm & VHOST_ACCESS_WO)
705 		gup_flags |= FOLL_WRITE;
706 
707 	npages = PFN_UP(size + (iova & ~PAGE_MASK));
708 	if (!npages) {
709 		ret = -EINVAL;
710 		goto free;
711 	}
712 
713 	mmap_read_lock(dev->mm);
714 
715 	lock_limit = PFN_DOWN(rlimit(RLIMIT_MEMLOCK));
716 	if (npages + atomic64_read(&dev->mm->pinned_vm) > lock_limit) {
717 		ret = -ENOMEM;
718 		goto unlock;
719 	}
720 
721 	cur_base = uaddr & PAGE_MASK;
722 	iova &= PAGE_MASK;
723 	nchunks = 0;
724 
725 	while (npages) {
726 		sz2pin = min_t(unsigned long, npages, list_size);
727 		pinned = pin_user_pages(cur_base, sz2pin,
728 					gup_flags, page_list, NULL);
729 		if (sz2pin != pinned) {
730 			if (pinned < 0) {
731 				ret = pinned;
732 			} else {
733 				unpin_user_pages(page_list, pinned);
734 				ret = -ENOMEM;
735 			}
736 			goto out;
737 		}
738 		nchunks++;
739 
740 		if (!last_pfn)
741 			map_pfn = page_to_pfn(page_list[0]);
742 
743 		for (i = 0; i < pinned; i++) {
744 			unsigned long this_pfn = page_to_pfn(page_list[i]);
745 			u64 csize;
746 
747 			if (last_pfn && (this_pfn != last_pfn + 1)) {
748 				/* Pin a contiguous chunk of memory */
749 				csize = PFN_PHYS(last_pfn - map_pfn + 1);
750 				ret = vhost_vdpa_map(v, iova, csize,
751 						     PFN_PHYS(map_pfn),
752 						     perm, NULL);
753 				if (ret) {
754 					/*
755 					 * Unpin the pages that are left unmapped
756 					 * from this point on in the current
757 					 * page_list. The remaining outstanding
758 					 * ones which may stride across several
759 					 * chunks will be covered in the common
760 					 * error path subsequently.
761 					 */
762 					unpin_user_pages(&page_list[i],
763 							 pinned - i);
764 					goto out;
765 				}
766 
767 				map_pfn = this_pfn;
768 				iova += csize;
769 				nchunks = 0;
770 			}
771 
772 			last_pfn = this_pfn;
773 		}
774 
775 		cur_base += PFN_PHYS(pinned);
776 		npages -= pinned;
777 	}
778 
779 	/* Pin the rest chunk */
780 	ret = vhost_vdpa_map(v, iova, PFN_PHYS(last_pfn - map_pfn + 1),
781 			     PFN_PHYS(map_pfn), perm, NULL);
782 out:
783 	if (ret) {
784 		if (nchunks) {
785 			unsigned long pfn;
786 
787 			/*
788 			 * Unpin the outstanding pages which are yet to be
789 			 * mapped but haven't due to vdpa_map() or
790 			 * pin_user_pages() failure.
791 			 *
792 			 * Mapped pages are accounted in vdpa_map(), hence
793 			 * the corresponding unpinning will be handled by
794 			 * vdpa_unmap().
795 			 */
796 			WARN_ON(!last_pfn);
797 			for (pfn = map_pfn; pfn <= last_pfn; pfn++)
798 				unpin_user_page(pfn_to_page(pfn));
799 		}
800 		vhost_vdpa_unmap(v, start, size);
801 	}
802 unlock:
803 	mmap_read_unlock(dev->mm);
804 free:
805 	free_page((unsigned long)page_list);
806 	return ret;
807 
808 }
809 
810 static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
811 					   struct vhost_iotlb_msg *msg)
812 {
813 	struct vhost_dev *dev = &v->vdev;
814 	struct vdpa_device *vdpa = v->vdpa;
815 	struct vhost_iotlb *iotlb = dev->iotlb;
816 
817 	if (msg->iova < v->range.first || !msg->size ||
818 	    msg->iova > U64_MAX - msg->size + 1 ||
819 	    msg->iova + msg->size - 1 > v->range.last)
820 		return -EINVAL;
821 
822 	if (vhost_iotlb_itree_first(iotlb, msg->iova,
823 				    msg->iova + msg->size - 1))
824 		return -EEXIST;
825 
826 	if (vdpa->use_va)
827 		return vhost_vdpa_va_map(v, msg->iova, msg->size,
828 					 msg->uaddr, msg->perm);
829 
830 	return vhost_vdpa_pa_map(v, msg->iova, msg->size, msg->uaddr,
831 				 msg->perm);
832 }
833 
834 static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev,
835 					struct vhost_iotlb_msg *msg)
836 {
837 	struct vhost_vdpa *v = container_of(dev, struct vhost_vdpa, vdev);
838 	struct vdpa_device *vdpa = v->vdpa;
839 	const struct vdpa_config_ops *ops = vdpa->config;
840 	int r = 0;
841 
842 	mutex_lock(&dev->mutex);
843 
844 	r = vhost_dev_check_owner(dev);
845 	if (r)
846 		goto unlock;
847 
848 	switch (msg->type) {
849 	case VHOST_IOTLB_UPDATE:
850 		r = vhost_vdpa_process_iotlb_update(v, msg);
851 		break;
852 	case VHOST_IOTLB_INVALIDATE:
853 		vhost_vdpa_unmap(v, msg->iova, msg->size);
854 		break;
855 	case VHOST_IOTLB_BATCH_BEGIN:
856 		v->in_batch = true;
857 		break;
858 	case VHOST_IOTLB_BATCH_END:
859 		if (v->in_batch && ops->set_map)
860 			ops->set_map(vdpa, dev->iotlb);
861 		v->in_batch = false;
862 		break;
863 	default:
864 		r = -EINVAL;
865 		break;
866 	}
867 unlock:
868 	mutex_unlock(&dev->mutex);
869 
870 	return r;
871 }
872 
873 static ssize_t vhost_vdpa_chr_write_iter(struct kiocb *iocb,
874 					 struct iov_iter *from)
875 {
876 	struct file *file = iocb->ki_filp;
877 	struct vhost_vdpa *v = file->private_data;
878 	struct vhost_dev *dev = &v->vdev;
879 
880 	return vhost_chr_write_iter(dev, from);
881 }
882 
883 static int vhost_vdpa_alloc_domain(struct vhost_vdpa *v)
884 {
885 	struct vdpa_device *vdpa = v->vdpa;
886 	const struct vdpa_config_ops *ops = vdpa->config;
887 	struct device *dma_dev = vdpa_get_dma_dev(vdpa);
888 	struct bus_type *bus;
889 	int ret;
890 
891 	/* Device want to do DMA by itself */
892 	if (ops->set_map || ops->dma_map)
893 		return 0;
894 
895 	bus = dma_dev->bus;
896 	if (!bus)
897 		return -EFAULT;
898 
899 	if (!iommu_capable(bus, IOMMU_CAP_CACHE_COHERENCY))
900 		return -ENOTSUPP;
901 
902 	v->domain = iommu_domain_alloc(bus);
903 	if (!v->domain)
904 		return -EIO;
905 
906 	ret = iommu_attach_device(v->domain, dma_dev);
907 	if (ret)
908 		goto err_attach;
909 
910 	return 0;
911 
912 err_attach:
913 	iommu_domain_free(v->domain);
914 	return ret;
915 }
916 
917 static void vhost_vdpa_free_domain(struct vhost_vdpa *v)
918 {
919 	struct vdpa_device *vdpa = v->vdpa;
920 	struct device *dma_dev = vdpa_get_dma_dev(vdpa);
921 
922 	if (v->domain) {
923 		iommu_detach_device(v->domain, dma_dev);
924 		iommu_domain_free(v->domain);
925 	}
926 
927 	v->domain = NULL;
928 }
929 
930 static void vhost_vdpa_set_iova_range(struct vhost_vdpa *v)
931 {
932 	struct vdpa_iova_range *range = &v->range;
933 	struct vdpa_device *vdpa = v->vdpa;
934 	const struct vdpa_config_ops *ops = vdpa->config;
935 
936 	if (ops->get_iova_range) {
937 		*range = ops->get_iova_range(vdpa);
938 	} else if (v->domain && v->domain->geometry.force_aperture) {
939 		range->first = v->domain->geometry.aperture_start;
940 		range->last = v->domain->geometry.aperture_end;
941 	} else {
942 		range->first = 0;
943 		range->last = ULLONG_MAX;
944 	}
945 }
946 
947 static int vhost_vdpa_open(struct inode *inode, struct file *filep)
948 {
949 	struct vhost_vdpa *v;
950 	struct vhost_dev *dev;
951 	struct vhost_virtqueue **vqs;
952 	int nvqs, i, r, opened;
953 
954 	v = container_of(inode->i_cdev, struct vhost_vdpa, cdev);
955 
956 	opened = atomic_cmpxchg(&v->opened, 0, 1);
957 	if (opened)
958 		return -EBUSY;
959 
960 	nvqs = v->nvqs;
961 	r = vhost_vdpa_reset(v);
962 	if (r)
963 		goto err;
964 
965 	vqs = kmalloc_array(nvqs, sizeof(*vqs), GFP_KERNEL);
966 	if (!vqs) {
967 		r = -ENOMEM;
968 		goto err;
969 	}
970 
971 	dev = &v->vdev;
972 	for (i = 0; i < nvqs; i++) {
973 		vqs[i] = &v->vqs[i];
974 		vqs[i]->handle_kick = handle_vq_kick;
975 	}
976 	vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, false,
977 		       vhost_vdpa_process_iotlb_msg);
978 
979 	dev->iotlb = vhost_iotlb_alloc(0, 0);
980 	if (!dev->iotlb) {
981 		r = -ENOMEM;
982 		goto err_init_iotlb;
983 	}
984 
985 	r = vhost_vdpa_alloc_domain(v);
986 	if (r)
987 		goto err_init_iotlb;
988 
989 	vhost_vdpa_set_iova_range(v);
990 
991 	filep->private_data = v;
992 
993 	return 0;
994 
995 err_init_iotlb:
996 	vhost_dev_cleanup(&v->vdev);
997 	kfree(vqs);
998 err:
999 	atomic_dec(&v->opened);
1000 	return r;
1001 }
1002 
1003 static void vhost_vdpa_clean_irq(struct vhost_vdpa *v)
1004 {
1005 	int i;
1006 
1007 	for (i = 0; i < v->nvqs; i++)
1008 		vhost_vdpa_unsetup_vq_irq(v, i);
1009 }
1010 
1011 static int vhost_vdpa_release(struct inode *inode, struct file *filep)
1012 {
1013 	struct vhost_vdpa *v = filep->private_data;
1014 	struct vhost_dev *d = &v->vdev;
1015 
1016 	mutex_lock(&d->mutex);
1017 	filep->private_data = NULL;
1018 	vhost_vdpa_reset(v);
1019 	vhost_dev_stop(&v->vdev);
1020 	vhost_vdpa_iotlb_free(v);
1021 	vhost_vdpa_free_domain(v);
1022 	vhost_vdpa_config_put(v);
1023 	vhost_vdpa_clean_irq(v);
1024 	vhost_dev_cleanup(&v->vdev);
1025 	kfree(v->vdev.vqs);
1026 	mutex_unlock(&d->mutex);
1027 
1028 	atomic_dec(&v->opened);
1029 	complete(&v->completion);
1030 
1031 	return 0;
1032 }
1033 
1034 #ifdef CONFIG_MMU
1035 static vm_fault_t vhost_vdpa_fault(struct vm_fault *vmf)
1036 {
1037 	struct vhost_vdpa *v = vmf->vma->vm_file->private_data;
1038 	struct vdpa_device *vdpa = v->vdpa;
1039 	const struct vdpa_config_ops *ops = vdpa->config;
1040 	struct vdpa_notification_area notify;
1041 	struct vm_area_struct *vma = vmf->vma;
1042 	u16 index = vma->vm_pgoff;
1043 
1044 	notify = ops->get_vq_notification(vdpa, index);
1045 
1046 	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
1047 	if (remap_pfn_range(vma, vmf->address & PAGE_MASK,
1048 			    PFN_DOWN(notify.addr), PAGE_SIZE,
1049 			    vma->vm_page_prot))
1050 		return VM_FAULT_SIGBUS;
1051 
1052 	return VM_FAULT_NOPAGE;
1053 }
1054 
1055 static const struct vm_operations_struct vhost_vdpa_vm_ops = {
1056 	.fault = vhost_vdpa_fault,
1057 };
1058 
1059 static int vhost_vdpa_mmap(struct file *file, struct vm_area_struct *vma)
1060 {
1061 	struct vhost_vdpa *v = vma->vm_file->private_data;
1062 	struct vdpa_device *vdpa = v->vdpa;
1063 	const struct vdpa_config_ops *ops = vdpa->config;
1064 	struct vdpa_notification_area notify;
1065 	unsigned long index = vma->vm_pgoff;
1066 
1067 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
1068 		return -EINVAL;
1069 	if ((vma->vm_flags & VM_SHARED) == 0)
1070 		return -EINVAL;
1071 	if (vma->vm_flags & VM_READ)
1072 		return -EINVAL;
1073 	if (index > 65535)
1074 		return -EINVAL;
1075 	if (!ops->get_vq_notification)
1076 		return -ENOTSUPP;
1077 
1078 	/* To be safe and easily modelled by userspace, We only
1079 	 * support the doorbell which sits on the page boundary and
1080 	 * does not share the page with other registers.
1081 	 */
1082 	notify = ops->get_vq_notification(vdpa, index);
1083 	if (notify.addr & (PAGE_SIZE - 1))
1084 		return -EINVAL;
1085 	if (vma->vm_end - vma->vm_start != notify.size)
1086 		return -ENOTSUPP;
1087 
1088 	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
1089 	vma->vm_ops = &vhost_vdpa_vm_ops;
1090 	return 0;
1091 }
1092 #endif /* CONFIG_MMU */
1093 
1094 static const struct file_operations vhost_vdpa_fops = {
1095 	.owner		= THIS_MODULE,
1096 	.open		= vhost_vdpa_open,
1097 	.release	= vhost_vdpa_release,
1098 	.write_iter	= vhost_vdpa_chr_write_iter,
1099 	.unlocked_ioctl	= vhost_vdpa_unlocked_ioctl,
1100 #ifdef CONFIG_MMU
1101 	.mmap		= vhost_vdpa_mmap,
1102 #endif /* CONFIG_MMU */
1103 	.compat_ioctl	= compat_ptr_ioctl,
1104 };
1105 
1106 static void vhost_vdpa_release_dev(struct device *device)
1107 {
1108 	struct vhost_vdpa *v =
1109 	       container_of(device, struct vhost_vdpa, dev);
1110 
1111 	ida_simple_remove(&vhost_vdpa_ida, v->minor);
1112 	kfree(v->vqs);
1113 	kfree(v);
1114 }
1115 
1116 static int vhost_vdpa_probe(struct vdpa_device *vdpa)
1117 {
1118 	const struct vdpa_config_ops *ops = vdpa->config;
1119 	struct vhost_vdpa *v;
1120 	int minor;
1121 	int r;
1122 
1123 	v = kzalloc(sizeof(*v), GFP_KERNEL | __GFP_RETRY_MAYFAIL);
1124 	if (!v)
1125 		return -ENOMEM;
1126 
1127 	minor = ida_simple_get(&vhost_vdpa_ida, 0,
1128 			       VHOST_VDPA_DEV_MAX, GFP_KERNEL);
1129 	if (minor < 0) {
1130 		kfree(v);
1131 		return minor;
1132 	}
1133 
1134 	atomic_set(&v->opened, 0);
1135 	v->minor = minor;
1136 	v->vdpa = vdpa;
1137 	v->nvqs = vdpa->nvqs;
1138 	v->virtio_id = ops->get_device_id(vdpa);
1139 
1140 	device_initialize(&v->dev);
1141 	v->dev.release = vhost_vdpa_release_dev;
1142 	v->dev.parent = &vdpa->dev;
1143 	v->dev.devt = MKDEV(MAJOR(vhost_vdpa_major), minor);
1144 	v->vqs = kmalloc_array(v->nvqs, sizeof(struct vhost_virtqueue),
1145 			       GFP_KERNEL);
1146 	if (!v->vqs) {
1147 		r = -ENOMEM;
1148 		goto err;
1149 	}
1150 
1151 	r = dev_set_name(&v->dev, "vhost-vdpa-%u", minor);
1152 	if (r)
1153 		goto err;
1154 
1155 	cdev_init(&v->cdev, &vhost_vdpa_fops);
1156 	v->cdev.owner = THIS_MODULE;
1157 
1158 	r = cdev_device_add(&v->cdev, &v->dev);
1159 	if (r)
1160 		goto err;
1161 
1162 	init_completion(&v->completion);
1163 	vdpa_set_drvdata(vdpa, v);
1164 
1165 	return 0;
1166 
1167 err:
1168 	put_device(&v->dev);
1169 	return r;
1170 }
1171 
1172 static void vhost_vdpa_remove(struct vdpa_device *vdpa)
1173 {
1174 	struct vhost_vdpa *v = vdpa_get_drvdata(vdpa);
1175 	int opened;
1176 
1177 	cdev_device_del(&v->cdev, &v->dev);
1178 
1179 	do {
1180 		opened = atomic_cmpxchg(&v->opened, 0, 1);
1181 		if (!opened)
1182 			break;
1183 		wait_for_completion(&v->completion);
1184 	} while (1);
1185 
1186 	put_device(&v->dev);
1187 }
1188 
1189 static struct vdpa_driver vhost_vdpa_driver = {
1190 	.driver = {
1191 		.name	= "vhost_vdpa",
1192 	},
1193 	.probe	= vhost_vdpa_probe,
1194 	.remove	= vhost_vdpa_remove,
1195 };
1196 
1197 static int __init vhost_vdpa_init(void)
1198 {
1199 	int r;
1200 
1201 	r = alloc_chrdev_region(&vhost_vdpa_major, 0, VHOST_VDPA_DEV_MAX,
1202 				"vhost-vdpa");
1203 	if (r)
1204 		goto err_alloc_chrdev;
1205 
1206 	r = vdpa_register_driver(&vhost_vdpa_driver);
1207 	if (r)
1208 		goto err_vdpa_register_driver;
1209 
1210 	return 0;
1211 
1212 err_vdpa_register_driver:
1213 	unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1214 err_alloc_chrdev:
1215 	return r;
1216 }
1217 module_init(vhost_vdpa_init);
1218 
1219 static void __exit vhost_vdpa_exit(void)
1220 {
1221 	vdpa_unregister_driver(&vhost_vdpa_driver);
1222 	unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1223 }
1224 module_exit(vhost_vdpa_exit);
1225 
1226 MODULE_VERSION("0.0.1");
1227 MODULE_LICENSE("GPL v2");
1228 MODULE_AUTHOR("Intel Corporation");
1229 MODULE_DESCRIPTION("vDPA-based vhost backend for virtio");
1230