xref: /openbmc/linux/drivers/iommu/virtio-iommu.c (revision bef7a78d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Virtio driver for the paravirtualized IOMMU
4  *
5  * Copyright (C) 2019 Arm Limited
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/amba/bus.h>
11 #include <linux/delay.h>
12 #include <linux/dma-iommu.h>
13 #include <linux/freezer.h>
14 #include <linux/interval_tree.h>
15 #include <linux/iommu.h>
16 #include <linux/module.h>
17 #include <linux/of_iommu.h>
18 #include <linux/of_platform.h>
19 #include <linux/pci.h>
20 #include <linux/platform_device.h>
21 #include <linux/virtio.h>
22 #include <linux/virtio_config.h>
23 #include <linux/virtio_ids.h>
24 #include <linux/wait.h>
25 
26 #include <uapi/linux/virtio_iommu.h>
27 
28 #define MSI_IOVA_BASE			0x8000000
29 #define MSI_IOVA_LENGTH			0x100000
30 
31 #define VIOMMU_REQUEST_VQ		0
32 #define VIOMMU_EVENT_VQ			1
33 #define VIOMMU_NR_VQS			2
34 
35 struct viommu_dev {
36 	struct iommu_device		iommu;
37 	struct device			*dev;
38 	struct virtio_device		*vdev;
39 
40 	struct ida			domain_ids;
41 
42 	struct virtqueue		*vqs[VIOMMU_NR_VQS];
43 	spinlock_t			request_lock;
44 	struct list_head		requests;
45 	void				*evts;
46 
47 	/* Device configuration */
48 	struct iommu_domain_geometry	geometry;
49 	u64				pgsize_bitmap;
50 	u32				first_domain;
51 	u32				last_domain;
52 	/* Supported MAP flags */
53 	u32				map_flags;
54 	u32				probe_size;
55 };
56 
57 struct viommu_mapping {
58 	phys_addr_t			paddr;
59 	struct interval_tree_node	iova;
60 	u32				flags;
61 };
62 
63 struct viommu_domain {
64 	struct iommu_domain		domain;
65 	struct viommu_dev		*viommu;
66 	struct mutex			mutex; /* protects viommu pointer */
67 	unsigned int			id;
68 	u32				map_flags;
69 
70 	spinlock_t			mappings_lock;
71 	struct rb_root_cached		mappings;
72 
73 	unsigned long			nr_endpoints;
74 };
75 
76 struct viommu_endpoint {
77 	struct device			*dev;
78 	struct viommu_dev		*viommu;
79 	struct viommu_domain		*vdomain;
80 	struct list_head		resv_regions;
81 };
82 
83 struct viommu_request {
84 	struct list_head		list;
85 	void				*writeback;
86 	unsigned int			write_offset;
87 	unsigned int			len;
88 	char				buf[];
89 };
90 
91 #define VIOMMU_FAULT_RESV_MASK		0xffffff00
92 
93 struct viommu_event {
94 	union {
95 		u32			head;
96 		struct virtio_iommu_fault fault;
97 	};
98 };
99 
100 #define to_viommu_domain(domain)	\
101 	container_of(domain, struct viommu_domain, domain)
102 
103 static int viommu_get_req_errno(void *buf, size_t len)
104 {
105 	struct virtio_iommu_req_tail *tail = buf + len - sizeof(*tail);
106 
107 	switch (tail->status) {
108 	case VIRTIO_IOMMU_S_OK:
109 		return 0;
110 	case VIRTIO_IOMMU_S_UNSUPP:
111 		return -ENOSYS;
112 	case VIRTIO_IOMMU_S_INVAL:
113 		return -EINVAL;
114 	case VIRTIO_IOMMU_S_RANGE:
115 		return -ERANGE;
116 	case VIRTIO_IOMMU_S_NOENT:
117 		return -ENOENT;
118 	case VIRTIO_IOMMU_S_FAULT:
119 		return -EFAULT;
120 	case VIRTIO_IOMMU_S_NOMEM:
121 		return -ENOMEM;
122 	case VIRTIO_IOMMU_S_IOERR:
123 	case VIRTIO_IOMMU_S_DEVERR:
124 	default:
125 		return -EIO;
126 	}
127 }
128 
129 static void viommu_set_req_status(void *buf, size_t len, int status)
130 {
131 	struct virtio_iommu_req_tail *tail = buf + len - sizeof(*tail);
132 
133 	tail->status = status;
134 }
135 
136 static off_t viommu_get_write_desc_offset(struct viommu_dev *viommu,
137 					  struct virtio_iommu_req_head *req,
138 					  size_t len)
139 {
140 	size_t tail_size = sizeof(struct virtio_iommu_req_tail);
141 
142 	if (req->type == VIRTIO_IOMMU_T_PROBE)
143 		return len - viommu->probe_size - tail_size;
144 
145 	return len - tail_size;
146 }
147 
148 /*
149  * __viommu_sync_req - Complete all in-flight requests
150  *
151  * Wait for all added requests to complete. When this function returns, all
152  * requests that were in-flight at the time of the call have completed.
153  */
154 static int __viommu_sync_req(struct viommu_dev *viommu)
155 {
156 	unsigned int len;
157 	size_t write_len;
158 	struct viommu_request *req;
159 	struct virtqueue *vq = viommu->vqs[VIOMMU_REQUEST_VQ];
160 
161 	assert_spin_locked(&viommu->request_lock);
162 
163 	virtqueue_kick(vq);
164 
165 	while (!list_empty(&viommu->requests)) {
166 		len = 0;
167 		req = virtqueue_get_buf(vq, &len);
168 		if (!req)
169 			continue;
170 
171 		if (!len)
172 			viommu_set_req_status(req->buf, req->len,
173 					      VIRTIO_IOMMU_S_IOERR);
174 
175 		write_len = req->len - req->write_offset;
176 		if (req->writeback && len == write_len)
177 			memcpy(req->writeback, req->buf + req->write_offset,
178 			       write_len);
179 
180 		list_del(&req->list);
181 		kfree(req);
182 	}
183 
184 	return 0;
185 }
186 
187 static int viommu_sync_req(struct viommu_dev *viommu)
188 {
189 	int ret;
190 	unsigned long flags;
191 
192 	spin_lock_irqsave(&viommu->request_lock, flags);
193 	ret = __viommu_sync_req(viommu);
194 	if (ret)
195 		dev_dbg(viommu->dev, "could not sync requests (%d)\n", ret);
196 	spin_unlock_irqrestore(&viommu->request_lock, flags);
197 
198 	return ret;
199 }
200 
201 /*
202  * __viommu_add_request - Add one request to the queue
203  * @buf: pointer to the request buffer
204  * @len: length of the request buffer
205  * @writeback: copy data back to the buffer when the request completes.
206  *
207  * Add a request to the queue. Only synchronize the queue if it's already full.
208  * Otherwise don't kick the queue nor wait for requests to complete.
209  *
210  * When @writeback is true, data written by the device, including the request
211  * status, is copied into @buf after the request completes. This is unsafe if
212  * the caller allocates @buf on stack and drops the lock between add_req() and
213  * sync_req().
214  *
215  * Return 0 if the request was successfully added to the queue.
216  */
217 static int __viommu_add_req(struct viommu_dev *viommu, void *buf, size_t len,
218 			    bool writeback)
219 {
220 	int ret;
221 	off_t write_offset;
222 	struct viommu_request *req;
223 	struct scatterlist top_sg, bottom_sg;
224 	struct scatterlist *sg[2] = { &top_sg, &bottom_sg };
225 	struct virtqueue *vq = viommu->vqs[VIOMMU_REQUEST_VQ];
226 
227 	assert_spin_locked(&viommu->request_lock);
228 
229 	write_offset = viommu_get_write_desc_offset(viommu, buf, len);
230 	if (write_offset <= 0)
231 		return -EINVAL;
232 
233 	req = kzalloc(sizeof(*req) + len, GFP_ATOMIC);
234 	if (!req)
235 		return -ENOMEM;
236 
237 	req->len = len;
238 	if (writeback) {
239 		req->writeback = buf + write_offset;
240 		req->write_offset = write_offset;
241 	}
242 	memcpy(&req->buf, buf, write_offset);
243 
244 	sg_init_one(&top_sg, req->buf, write_offset);
245 	sg_init_one(&bottom_sg, req->buf + write_offset, len - write_offset);
246 
247 	ret = virtqueue_add_sgs(vq, sg, 1, 1, req, GFP_ATOMIC);
248 	if (ret == -ENOSPC) {
249 		/* If the queue is full, sync and retry */
250 		if (!__viommu_sync_req(viommu))
251 			ret = virtqueue_add_sgs(vq, sg, 1, 1, req, GFP_ATOMIC);
252 	}
253 	if (ret)
254 		goto err_free;
255 
256 	list_add_tail(&req->list, &viommu->requests);
257 	return 0;
258 
259 err_free:
260 	kfree(req);
261 	return ret;
262 }
263 
264 static int viommu_add_req(struct viommu_dev *viommu, void *buf, size_t len)
265 {
266 	int ret;
267 	unsigned long flags;
268 
269 	spin_lock_irqsave(&viommu->request_lock, flags);
270 	ret = __viommu_add_req(viommu, buf, len, false);
271 	if (ret)
272 		dev_dbg(viommu->dev, "could not add request: %d\n", ret);
273 	spin_unlock_irqrestore(&viommu->request_lock, flags);
274 
275 	return ret;
276 }
277 
278 /*
279  * Send a request and wait for it to complete. Return the request status (as an
280  * errno)
281  */
282 static int viommu_send_req_sync(struct viommu_dev *viommu, void *buf,
283 				size_t len)
284 {
285 	int ret;
286 	unsigned long flags;
287 
288 	spin_lock_irqsave(&viommu->request_lock, flags);
289 
290 	ret = __viommu_add_req(viommu, buf, len, true);
291 	if (ret) {
292 		dev_dbg(viommu->dev, "could not add request (%d)\n", ret);
293 		goto out_unlock;
294 	}
295 
296 	ret = __viommu_sync_req(viommu);
297 	if (ret) {
298 		dev_dbg(viommu->dev, "could not sync requests (%d)\n", ret);
299 		/* Fall-through (get the actual request status) */
300 	}
301 
302 	ret = viommu_get_req_errno(buf, len);
303 out_unlock:
304 	spin_unlock_irqrestore(&viommu->request_lock, flags);
305 	return ret;
306 }
307 
308 /*
309  * viommu_add_mapping - add a mapping to the internal tree
310  *
311  * On success, return the new mapping. Otherwise return NULL.
312  */
313 static int viommu_add_mapping(struct viommu_domain *vdomain, unsigned long iova,
314 			      phys_addr_t paddr, size_t size, u32 flags)
315 {
316 	unsigned long irqflags;
317 	struct viommu_mapping *mapping;
318 
319 	mapping = kzalloc(sizeof(*mapping), GFP_ATOMIC);
320 	if (!mapping)
321 		return -ENOMEM;
322 
323 	mapping->paddr		= paddr;
324 	mapping->iova.start	= iova;
325 	mapping->iova.last	= iova + size - 1;
326 	mapping->flags		= flags;
327 
328 	spin_lock_irqsave(&vdomain->mappings_lock, irqflags);
329 	interval_tree_insert(&mapping->iova, &vdomain->mappings);
330 	spin_unlock_irqrestore(&vdomain->mappings_lock, irqflags);
331 
332 	return 0;
333 }
334 
335 /*
336  * viommu_del_mappings - remove mappings from the internal tree
337  *
338  * @vdomain: the domain
339  * @iova: start of the range
340  * @size: size of the range. A size of 0 corresponds to the entire address
341  *	space.
342  *
343  * On success, returns the number of unmapped bytes (>= size)
344  */
345 static size_t viommu_del_mappings(struct viommu_domain *vdomain,
346 				  unsigned long iova, size_t size)
347 {
348 	size_t unmapped = 0;
349 	unsigned long flags;
350 	unsigned long last = iova + size - 1;
351 	struct viommu_mapping *mapping = NULL;
352 	struct interval_tree_node *node, *next;
353 
354 	spin_lock_irqsave(&vdomain->mappings_lock, flags);
355 	next = interval_tree_iter_first(&vdomain->mappings, iova, last);
356 	while (next) {
357 		node = next;
358 		mapping = container_of(node, struct viommu_mapping, iova);
359 		next = interval_tree_iter_next(node, iova, last);
360 
361 		/* Trying to split a mapping? */
362 		if (mapping->iova.start < iova)
363 			break;
364 
365 		/*
366 		 * Virtio-iommu doesn't allow UNMAP to split a mapping created
367 		 * with a single MAP request, so remove the full mapping.
368 		 */
369 		unmapped += mapping->iova.last - mapping->iova.start + 1;
370 
371 		interval_tree_remove(node, &vdomain->mappings);
372 		kfree(mapping);
373 	}
374 	spin_unlock_irqrestore(&vdomain->mappings_lock, flags);
375 
376 	return unmapped;
377 }
378 
379 /*
380  * viommu_replay_mappings - re-send MAP requests
381  *
382  * When reattaching a domain that was previously detached from all endpoints,
383  * mappings were deleted from the device. Re-create the mappings available in
384  * the internal tree.
385  */
386 static int viommu_replay_mappings(struct viommu_domain *vdomain)
387 {
388 	int ret = 0;
389 	unsigned long flags;
390 	struct viommu_mapping *mapping;
391 	struct interval_tree_node *node;
392 	struct virtio_iommu_req_map map;
393 
394 	spin_lock_irqsave(&vdomain->mappings_lock, flags);
395 	node = interval_tree_iter_first(&vdomain->mappings, 0, -1UL);
396 	while (node) {
397 		mapping = container_of(node, struct viommu_mapping, iova);
398 		map = (struct virtio_iommu_req_map) {
399 			.head.type	= VIRTIO_IOMMU_T_MAP,
400 			.domain		= cpu_to_le32(vdomain->id),
401 			.virt_start	= cpu_to_le64(mapping->iova.start),
402 			.virt_end	= cpu_to_le64(mapping->iova.last),
403 			.phys_start	= cpu_to_le64(mapping->paddr),
404 			.flags		= cpu_to_le32(mapping->flags),
405 		};
406 
407 		ret = viommu_send_req_sync(vdomain->viommu, &map, sizeof(map));
408 		if (ret)
409 			break;
410 
411 		node = interval_tree_iter_next(node, 0, -1UL);
412 	}
413 	spin_unlock_irqrestore(&vdomain->mappings_lock, flags);
414 
415 	return ret;
416 }
417 
418 static int viommu_add_resv_mem(struct viommu_endpoint *vdev,
419 			       struct virtio_iommu_probe_resv_mem *mem,
420 			       size_t len)
421 {
422 	size_t size;
423 	u64 start64, end64;
424 	phys_addr_t start, end;
425 	struct iommu_resv_region *region = NULL;
426 	unsigned long prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
427 
428 	start = start64 = le64_to_cpu(mem->start);
429 	end = end64 = le64_to_cpu(mem->end);
430 	size = end64 - start64 + 1;
431 
432 	/* Catch any overflow, including the unlikely end64 - start64 + 1 = 0 */
433 	if (start != start64 || end != end64 || size < end64 - start64)
434 		return -EOVERFLOW;
435 
436 	if (len < sizeof(*mem))
437 		return -EINVAL;
438 
439 	switch (mem->subtype) {
440 	default:
441 		dev_warn(vdev->dev, "unknown resv mem subtype 0x%x\n",
442 			 mem->subtype);
443 		fallthrough;
444 	case VIRTIO_IOMMU_RESV_MEM_T_RESERVED:
445 		region = iommu_alloc_resv_region(start, size, 0,
446 						 IOMMU_RESV_RESERVED);
447 		break;
448 	case VIRTIO_IOMMU_RESV_MEM_T_MSI:
449 		region = iommu_alloc_resv_region(start, size, prot,
450 						 IOMMU_RESV_MSI);
451 		break;
452 	}
453 	if (!region)
454 		return -ENOMEM;
455 
456 	list_add(&region->list, &vdev->resv_regions);
457 	return 0;
458 }
459 
460 static int viommu_probe_endpoint(struct viommu_dev *viommu, struct device *dev)
461 {
462 	int ret;
463 	u16 type, len;
464 	size_t cur = 0;
465 	size_t probe_len;
466 	struct virtio_iommu_req_probe *probe;
467 	struct virtio_iommu_probe_property *prop;
468 	struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
469 	struct viommu_endpoint *vdev = dev_iommu_priv_get(dev);
470 
471 	if (!fwspec->num_ids)
472 		return -EINVAL;
473 
474 	probe_len = sizeof(*probe) + viommu->probe_size +
475 		    sizeof(struct virtio_iommu_req_tail);
476 	probe = kzalloc(probe_len, GFP_KERNEL);
477 	if (!probe)
478 		return -ENOMEM;
479 
480 	probe->head.type = VIRTIO_IOMMU_T_PROBE;
481 	/*
482 	 * For now, assume that properties of an endpoint that outputs multiple
483 	 * IDs are consistent. Only probe the first one.
484 	 */
485 	probe->endpoint = cpu_to_le32(fwspec->ids[0]);
486 
487 	ret = viommu_send_req_sync(viommu, probe, probe_len);
488 	if (ret)
489 		goto out_free;
490 
491 	prop = (void *)probe->properties;
492 	type = le16_to_cpu(prop->type) & VIRTIO_IOMMU_PROBE_T_MASK;
493 
494 	while (type != VIRTIO_IOMMU_PROBE_T_NONE &&
495 	       cur < viommu->probe_size) {
496 		len = le16_to_cpu(prop->length) + sizeof(*prop);
497 
498 		switch (type) {
499 		case VIRTIO_IOMMU_PROBE_T_RESV_MEM:
500 			ret = viommu_add_resv_mem(vdev, (void *)prop, len);
501 			break;
502 		default:
503 			dev_err(dev, "unknown viommu prop 0x%x\n", type);
504 		}
505 
506 		if (ret)
507 			dev_err(dev, "failed to parse viommu prop 0x%x\n", type);
508 
509 		cur += len;
510 		if (cur >= viommu->probe_size)
511 			break;
512 
513 		prop = (void *)probe->properties + cur;
514 		type = le16_to_cpu(prop->type) & VIRTIO_IOMMU_PROBE_T_MASK;
515 	}
516 
517 out_free:
518 	kfree(probe);
519 	return ret;
520 }
521 
522 static int viommu_fault_handler(struct viommu_dev *viommu,
523 				struct virtio_iommu_fault *fault)
524 {
525 	char *reason_str;
526 
527 	u8 reason	= fault->reason;
528 	u32 flags	= le32_to_cpu(fault->flags);
529 	u32 endpoint	= le32_to_cpu(fault->endpoint);
530 	u64 address	= le64_to_cpu(fault->address);
531 
532 	switch (reason) {
533 	case VIRTIO_IOMMU_FAULT_R_DOMAIN:
534 		reason_str = "domain";
535 		break;
536 	case VIRTIO_IOMMU_FAULT_R_MAPPING:
537 		reason_str = "page";
538 		break;
539 	case VIRTIO_IOMMU_FAULT_R_UNKNOWN:
540 	default:
541 		reason_str = "unknown";
542 		break;
543 	}
544 
545 	/* TODO: find EP by ID and report_iommu_fault */
546 	if (flags & VIRTIO_IOMMU_FAULT_F_ADDRESS)
547 		dev_err_ratelimited(viommu->dev, "%s fault from EP %u at %#llx [%s%s%s]\n",
548 				    reason_str, endpoint, address,
549 				    flags & VIRTIO_IOMMU_FAULT_F_READ ? "R" : "",
550 				    flags & VIRTIO_IOMMU_FAULT_F_WRITE ? "W" : "",
551 				    flags & VIRTIO_IOMMU_FAULT_F_EXEC ? "X" : "");
552 	else
553 		dev_err_ratelimited(viommu->dev, "%s fault from EP %u\n",
554 				    reason_str, endpoint);
555 	return 0;
556 }
557 
558 static void viommu_event_handler(struct virtqueue *vq)
559 {
560 	int ret;
561 	unsigned int len;
562 	struct scatterlist sg[1];
563 	struct viommu_event *evt;
564 	struct viommu_dev *viommu = vq->vdev->priv;
565 
566 	while ((evt = virtqueue_get_buf(vq, &len)) != NULL) {
567 		if (len > sizeof(*evt)) {
568 			dev_err(viommu->dev,
569 				"invalid event buffer (len %u != %zu)\n",
570 				len, sizeof(*evt));
571 		} else if (!(evt->head & VIOMMU_FAULT_RESV_MASK)) {
572 			viommu_fault_handler(viommu, &evt->fault);
573 		}
574 
575 		sg_init_one(sg, evt, sizeof(*evt));
576 		ret = virtqueue_add_inbuf(vq, sg, 1, evt, GFP_ATOMIC);
577 		if (ret)
578 			dev_err(viommu->dev, "could not add event buffer\n");
579 	}
580 
581 	virtqueue_kick(vq);
582 }
583 
584 /* IOMMU API */
585 
586 static struct iommu_domain *viommu_domain_alloc(unsigned type)
587 {
588 	struct viommu_domain *vdomain;
589 
590 	if (type != IOMMU_DOMAIN_UNMANAGED && type != IOMMU_DOMAIN_DMA)
591 		return NULL;
592 
593 	vdomain = kzalloc(sizeof(*vdomain), GFP_KERNEL);
594 	if (!vdomain)
595 		return NULL;
596 
597 	mutex_init(&vdomain->mutex);
598 	spin_lock_init(&vdomain->mappings_lock);
599 	vdomain->mappings = RB_ROOT_CACHED;
600 
601 	if (type == IOMMU_DOMAIN_DMA &&
602 	    iommu_get_dma_cookie(&vdomain->domain)) {
603 		kfree(vdomain);
604 		return NULL;
605 	}
606 
607 	return &vdomain->domain;
608 }
609 
610 static int viommu_domain_finalise(struct viommu_endpoint *vdev,
611 				  struct iommu_domain *domain)
612 {
613 	int ret;
614 	unsigned long viommu_page_size;
615 	struct viommu_dev *viommu = vdev->viommu;
616 	struct viommu_domain *vdomain = to_viommu_domain(domain);
617 
618 	viommu_page_size = 1UL << __ffs(viommu->pgsize_bitmap);
619 	if (viommu_page_size > PAGE_SIZE) {
620 		dev_err(vdev->dev,
621 			"granule 0x%lx larger than system page size 0x%lx\n",
622 			viommu_page_size, PAGE_SIZE);
623 		return -EINVAL;
624 	}
625 
626 	ret = ida_alloc_range(&viommu->domain_ids, viommu->first_domain,
627 			      viommu->last_domain, GFP_KERNEL);
628 	if (ret < 0)
629 		return ret;
630 
631 	vdomain->id		= (unsigned int)ret;
632 
633 	domain->pgsize_bitmap	= viommu->pgsize_bitmap;
634 	domain->geometry	= viommu->geometry;
635 
636 	vdomain->map_flags	= viommu->map_flags;
637 	vdomain->viommu		= viommu;
638 
639 	return 0;
640 }
641 
642 static void viommu_domain_free(struct iommu_domain *domain)
643 {
644 	struct viommu_domain *vdomain = to_viommu_domain(domain);
645 
646 	iommu_put_dma_cookie(domain);
647 
648 	/* Free all remaining mappings (size 2^64) */
649 	viommu_del_mappings(vdomain, 0, 0);
650 
651 	if (vdomain->viommu)
652 		ida_free(&vdomain->viommu->domain_ids, vdomain->id);
653 
654 	kfree(vdomain);
655 }
656 
657 static int viommu_attach_dev(struct iommu_domain *domain, struct device *dev)
658 {
659 	int i;
660 	int ret = 0;
661 	struct virtio_iommu_req_attach req;
662 	struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
663 	struct viommu_endpoint *vdev = dev_iommu_priv_get(dev);
664 	struct viommu_domain *vdomain = to_viommu_domain(domain);
665 
666 	mutex_lock(&vdomain->mutex);
667 	if (!vdomain->viommu) {
668 		/*
669 		 * Properly initialize the domain now that we know which viommu
670 		 * owns it.
671 		 */
672 		ret = viommu_domain_finalise(vdev, domain);
673 	} else if (vdomain->viommu != vdev->viommu) {
674 		dev_err(dev, "cannot attach to foreign vIOMMU\n");
675 		ret = -EXDEV;
676 	}
677 	mutex_unlock(&vdomain->mutex);
678 
679 	if (ret)
680 		return ret;
681 
682 	/*
683 	 * In the virtio-iommu device, when attaching the endpoint to a new
684 	 * domain, it is detached from the old one and, if as as a result the
685 	 * old domain isn't attached to any endpoint, all mappings are removed
686 	 * from the old domain and it is freed.
687 	 *
688 	 * In the driver the old domain still exists, and its mappings will be
689 	 * recreated if it gets reattached to an endpoint. Otherwise it will be
690 	 * freed explicitly.
691 	 *
692 	 * vdev->vdomain is protected by group->mutex
693 	 */
694 	if (vdev->vdomain)
695 		vdev->vdomain->nr_endpoints--;
696 
697 	req = (struct virtio_iommu_req_attach) {
698 		.head.type	= VIRTIO_IOMMU_T_ATTACH,
699 		.domain		= cpu_to_le32(vdomain->id),
700 	};
701 
702 	for (i = 0; i < fwspec->num_ids; i++) {
703 		req.endpoint = cpu_to_le32(fwspec->ids[i]);
704 
705 		ret = viommu_send_req_sync(vdomain->viommu, &req, sizeof(req));
706 		if (ret)
707 			return ret;
708 	}
709 
710 	if (!vdomain->nr_endpoints) {
711 		/*
712 		 * This endpoint is the first to be attached to the domain.
713 		 * Replay existing mappings (e.g. SW MSI).
714 		 */
715 		ret = viommu_replay_mappings(vdomain);
716 		if (ret)
717 			return ret;
718 	}
719 
720 	vdomain->nr_endpoints++;
721 	vdev->vdomain = vdomain;
722 
723 	return 0;
724 }
725 
726 static int viommu_map(struct iommu_domain *domain, unsigned long iova,
727 		      phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
728 {
729 	int ret;
730 	u32 flags;
731 	struct virtio_iommu_req_map map;
732 	struct viommu_domain *vdomain = to_viommu_domain(domain);
733 
734 	flags = (prot & IOMMU_READ ? VIRTIO_IOMMU_MAP_F_READ : 0) |
735 		(prot & IOMMU_WRITE ? VIRTIO_IOMMU_MAP_F_WRITE : 0) |
736 		(prot & IOMMU_MMIO ? VIRTIO_IOMMU_MAP_F_MMIO : 0);
737 
738 	if (flags & ~vdomain->map_flags)
739 		return -EINVAL;
740 
741 	ret = viommu_add_mapping(vdomain, iova, paddr, size, flags);
742 	if (ret)
743 		return ret;
744 
745 	map = (struct virtio_iommu_req_map) {
746 		.head.type	= VIRTIO_IOMMU_T_MAP,
747 		.domain		= cpu_to_le32(vdomain->id),
748 		.virt_start	= cpu_to_le64(iova),
749 		.phys_start	= cpu_to_le64(paddr),
750 		.virt_end	= cpu_to_le64(iova + size - 1),
751 		.flags		= cpu_to_le32(flags),
752 	};
753 
754 	if (!vdomain->nr_endpoints)
755 		return 0;
756 
757 	ret = viommu_send_req_sync(vdomain->viommu, &map, sizeof(map));
758 	if (ret)
759 		viommu_del_mappings(vdomain, iova, size);
760 
761 	return ret;
762 }
763 
764 static size_t viommu_unmap(struct iommu_domain *domain, unsigned long iova,
765 			   size_t size, struct iommu_iotlb_gather *gather)
766 {
767 	int ret = 0;
768 	size_t unmapped;
769 	struct virtio_iommu_req_unmap unmap;
770 	struct viommu_domain *vdomain = to_viommu_domain(domain);
771 
772 	unmapped = viommu_del_mappings(vdomain, iova, size);
773 	if (unmapped < size)
774 		return 0;
775 
776 	/* Device already removed all mappings after detach. */
777 	if (!vdomain->nr_endpoints)
778 		return unmapped;
779 
780 	unmap = (struct virtio_iommu_req_unmap) {
781 		.head.type	= VIRTIO_IOMMU_T_UNMAP,
782 		.domain		= cpu_to_le32(vdomain->id),
783 		.virt_start	= cpu_to_le64(iova),
784 		.virt_end	= cpu_to_le64(iova + unmapped - 1),
785 	};
786 
787 	ret = viommu_add_req(vdomain->viommu, &unmap, sizeof(unmap));
788 	return ret ? 0 : unmapped;
789 }
790 
791 static phys_addr_t viommu_iova_to_phys(struct iommu_domain *domain,
792 				       dma_addr_t iova)
793 {
794 	u64 paddr = 0;
795 	unsigned long flags;
796 	struct viommu_mapping *mapping;
797 	struct interval_tree_node *node;
798 	struct viommu_domain *vdomain = to_viommu_domain(domain);
799 
800 	spin_lock_irqsave(&vdomain->mappings_lock, flags);
801 	node = interval_tree_iter_first(&vdomain->mappings, iova, iova);
802 	if (node) {
803 		mapping = container_of(node, struct viommu_mapping, iova);
804 		paddr = mapping->paddr + (iova - mapping->iova.start);
805 	}
806 	spin_unlock_irqrestore(&vdomain->mappings_lock, flags);
807 
808 	return paddr;
809 }
810 
811 static void viommu_iotlb_sync(struct iommu_domain *domain,
812 			      struct iommu_iotlb_gather *gather)
813 {
814 	struct viommu_domain *vdomain = to_viommu_domain(domain);
815 
816 	viommu_sync_req(vdomain->viommu);
817 }
818 
819 static void viommu_get_resv_regions(struct device *dev, struct list_head *head)
820 {
821 	struct iommu_resv_region *entry, *new_entry, *msi = NULL;
822 	struct viommu_endpoint *vdev = dev_iommu_priv_get(dev);
823 	int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
824 
825 	list_for_each_entry(entry, &vdev->resv_regions, list) {
826 		if (entry->type == IOMMU_RESV_MSI)
827 			msi = entry;
828 
829 		new_entry = kmemdup(entry, sizeof(*entry), GFP_KERNEL);
830 		if (!new_entry)
831 			return;
832 		list_add_tail(&new_entry->list, head);
833 	}
834 
835 	/*
836 	 * If the device didn't register any bypass MSI window, add a
837 	 * software-mapped region.
838 	 */
839 	if (!msi) {
840 		msi = iommu_alloc_resv_region(MSI_IOVA_BASE, MSI_IOVA_LENGTH,
841 					      prot, IOMMU_RESV_SW_MSI);
842 		if (!msi)
843 			return;
844 
845 		list_add_tail(&msi->list, head);
846 	}
847 
848 	iommu_dma_get_resv_regions(dev, head);
849 }
850 
851 static struct iommu_ops viommu_ops;
852 static struct virtio_driver virtio_iommu_drv;
853 
854 static int viommu_match_node(struct device *dev, const void *data)
855 {
856 	return dev->parent->fwnode == data;
857 }
858 
859 static struct viommu_dev *viommu_get_by_fwnode(struct fwnode_handle *fwnode)
860 {
861 	struct device *dev = driver_find_device(&virtio_iommu_drv.driver, NULL,
862 						fwnode, viommu_match_node);
863 	put_device(dev);
864 
865 	return dev ? dev_to_virtio(dev)->priv : NULL;
866 }
867 
868 static struct iommu_device *viommu_probe_device(struct device *dev)
869 {
870 	int ret;
871 	struct viommu_endpoint *vdev;
872 	struct viommu_dev *viommu = NULL;
873 	struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
874 
875 	if (!fwspec || fwspec->ops != &viommu_ops)
876 		return ERR_PTR(-ENODEV);
877 
878 	viommu = viommu_get_by_fwnode(fwspec->iommu_fwnode);
879 	if (!viommu)
880 		return ERR_PTR(-ENODEV);
881 
882 	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
883 	if (!vdev)
884 		return ERR_PTR(-ENOMEM);
885 
886 	vdev->dev = dev;
887 	vdev->viommu = viommu;
888 	INIT_LIST_HEAD(&vdev->resv_regions);
889 	dev_iommu_priv_set(dev, vdev);
890 
891 	if (viommu->probe_size) {
892 		/* Get additional information for this endpoint */
893 		ret = viommu_probe_endpoint(viommu, dev);
894 		if (ret)
895 			goto err_free_dev;
896 	}
897 
898 	return &viommu->iommu;
899 
900 err_free_dev:
901 	generic_iommu_put_resv_regions(dev, &vdev->resv_regions);
902 	kfree(vdev);
903 
904 	return ERR_PTR(ret);
905 }
906 
907 static void viommu_release_device(struct device *dev)
908 {
909 	struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
910 	struct viommu_endpoint *vdev;
911 
912 	if (!fwspec || fwspec->ops != &viommu_ops)
913 		return;
914 
915 	vdev = dev_iommu_priv_get(dev);
916 
917 	generic_iommu_put_resv_regions(dev, &vdev->resv_regions);
918 	kfree(vdev);
919 }
920 
921 static struct iommu_group *viommu_device_group(struct device *dev)
922 {
923 	if (dev_is_pci(dev))
924 		return pci_device_group(dev);
925 	else
926 		return generic_device_group(dev);
927 }
928 
929 static int viommu_of_xlate(struct device *dev, struct of_phandle_args *args)
930 {
931 	return iommu_fwspec_add_ids(dev, args->args, 1);
932 }
933 
934 static struct iommu_ops viommu_ops = {
935 	.domain_alloc		= viommu_domain_alloc,
936 	.domain_free		= viommu_domain_free,
937 	.attach_dev		= viommu_attach_dev,
938 	.map			= viommu_map,
939 	.unmap			= viommu_unmap,
940 	.iova_to_phys		= viommu_iova_to_phys,
941 	.iotlb_sync		= viommu_iotlb_sync,
942 	.probe_device		= viommu_probe_device,
943 	.release_device		= viommu_release_device,
944 	.device_group		= viommu_device_group,
945 	.get_resv_regions	= viommu_get_resv_regions,
946 	.put_resv_regions	= generic_iommu_put_resv_regions,
947 	.of_xlate		= viommu_of_xlate,
948 };
949 
950 static int viommu_init_vqs(struct viommu_dev *viommu)
951 {
952 	struct virtio_device *vdev = dev_to_virtio(viommu->dev);
953 	const char *names[] = { "request", "event" };
954 	vq_callback_t *callbacks[] = {
955 		NULL, /* No async requests */
956 		viommu_event_handler,
957 	};
958 
959 	return virtio_find_vqs(vdev, VIOMMU_NR_VQS, viommu->vqs, callbacks,
960 			       names, NULL);
961 }
962 
963 static int viommu_fill_evtq(struct viommu_dev *viommu)
964 {
965 	int i, ret;
966 	struct scatterlist sg[1];
967 	struct viommu_event *evts;
968 	struct virtqueue *vq = viommu->vqs[VIOMMU_EVENT_VQ];
969 	size_t nr_evts = vq->num_free;
970 
971 	viommu->evts = evts = devm_kmalloc_array(viommu->dev, nr_evts,
972 						 sizeof(*evts), GFP_KERNEL);
973 	if (!evts)
974 		return -ENOMEM;
975 
976 	for (i = 0; i < nr_evts; i++) {
977 		sg_init_one(sg, &evts[i], sizeof(*evts));
978 		ret = virtqueue_add_inbuf(vq, sg, 1, &evts[i], GFP_KERNEL);
979 		if (ret)
980 			return ret;
981 	}
982 
983 	return 0;
984 }
985 
986 static int viommu_probe(struct virtio_device *vdev)
987 {
988 	struct device *parent_dev = vdev->dev.parent;
989 	struct viommu_dev *viommu = NULL;
990 	struct device *dev = &vdev->dev;
991 	u64 input_start = 0;
992 	u64 input_end = -1UL;
993 	int ret;
994 
995 	if (!virtio_has_feature(vdev, VIRTIO_F_VERSION_1) ||
996 	    !virtio_has_feature(vdev, VIRTIO_IOMMU_F_MAP_UNMAP))
997 		return -ENODEV;
998 
999 	viommu = devm_kzalloc(dev, sizeof(*viommu), GFP_KERNEL);
1000 	if (!viommu)
1001 		return -ENOMEM;
1002 
1003 	spin_lock_init(&viommu->request_lock);
1004 	ida_init(&viommu->domain_ids);
1005 	viommu->dev = dev;
1006 	viommu->vdev = vdev;
1007 	INIT_LIST_HEAD(&viommu->requests);
1008 
1009 	ret = viommu_init_vqs(viommu);
1010 	if (ret)
1011 		return ret;
1012 
1013 	virtio_cread_le(vdev, struct virtio_iommu_config, page_size_mask,
1014 			&viommu->pgsize_bitmap);
1015 
1016 	if (!viommu->pgsize_bitmap) {
1017 		ret = -EINVAL;
1018 		goto err_free_vqs;
1019 	}
1020 
1021 	viommu->map_flags = VIRTIO_IOMMU_MAP_F_READ | VIRTIO_IOMMU_MAP_F_WRITE;
1022 	viommu->last_domain = ~0U;
1023 
1024 	/* Optional features */
1025 	virtio_cread_le_feature(vdev, VIRTIO_IOMMU_F_INPUT_RANGE,
1026 				struct virtio_iommu_config, input_range.start,
1027 				&input_start);
1028 
1029 	virtio_cread_le_feature(vdev, VIRTIO_IOMMU_F_INPUT_RANGE,
1030 				struct virtio_iommu_config, input_range.end,
1031 				&input_end);
1032 
1033 	virtio_cread_le_feature(vdev, VIRTIO_IOMMU_F_DOMAIN_RANGE,
1034 				struct virtio_iommu_config, domain_range.start,
1035 				&viommu->first_domain);
1036 
1037 	virtio_cread_le_feature(vdev, VIRTIO_IOMMU_F_DOMAIN_RANGE,
1038 				struct virtio_iommu_config, domain_range.end,
1039 				&viommu->last_domain);
1040 
1041 	virtio_cread_le_feature(vdev, VIRTIO_IOMMU_F_PROBE,
1042 				struct virtio_iommu_config, probe_size,
1043 				&viommu->probe_size);
1044 
1045 	viommu->geometry = (struct iommu_domain_geometry) {
1046 		.aperture_start	= input_start,
1047 		.aperture_end	= input_end,
1048 		.force_aperture	= true,
1049 	};
1050 
1051 	if (virtio_has_feature(vdev, VIRTIO_IOMMU_F_MMIO))
1052 		viommu->map_flags |= VIRTIO_IOMMU_MAP_F_MMIO;
1053 
1054 	viommu_ops.pgsize_bitmap = viommu->pgsize_bitmap;
1055 
1056 	virtio_device_ready(vdev);
1057 
1058 	/* Populate the event queue with buffers */
1059 	ret = viommu_fill_evtq(viommu);
1060 	if (ret)
1061 		goto err_free_vqs;
1062 
1063 	ret = iommu_device_sysfs_add(&viommu->iommu, dev, NULL, "%s",
1064 				     virtio_bus_name(vdev));
1065 	if (ret)
1066 		goto err_free_vqs;
1067 
1068 	iommu_device_set_ops(&viommu->iommu, &viommu_ops);
1069 	iommu_device_set_fwnode(&viommu->iommu, parent_dev->fwnode);
1070 
1071 	iommu_device_register(&viommu->iommu);
1072 
1073 #ifdef CONFIG_PCI
1074 	if (pci_bus_type.iommu_ops != &viommu_ops) {
1075 		ret = bus_set_iommu(&pci_bus_type, &viommu_ops);
1076 		if (ret)
1077 			goto err_unregister;
1078 	}
1079 #endif
1080 #ifdef CONFIG_ARM_AMBA
1081 	if (amba_bustype.iommu_ops != &viommu_ops) {
1082 		ret = bus_set_iommu(&amba_bustype, &viommu_ops);
1083 		if (ret)
1084 			goto err_unregister;
1085 	}
1086 #endif
1087 	if (platform_bus_type.iommu_ops != &viommu_ops) {
1088 		ret = bus_set_iommu(&platform_bus_type, &viommu_ops);
1089 		if (ret)
1090 			goto err_unregister;
1091 	}
1092 
1093 	vdev->priv = viommu;
1094 
1095 	dev_info(dev, "input address: %u bits\n",
1096 		 order_base_2(viommu->geometry.aperture_end));
1097 	dev_info(dev, "page mask: %#llx\n", viommu->pgsize_bitmap);
1098 
1099 	return 0;
1100 
1101 err_unregister:
1102 	iommu_device_sysfs_remove(&viommu->iommu);
1103 	iommu_device_unregister(&viommu->iommu);
1104 err_free_vqs:
1105 	vdev->config->del_vqs(vdev);
1106 
1107 	return ret;
1108 }
1109 
1110 static void viommu_remove(struct virtio_device *vdev)
1111 {
1112 	struct viommu_dev *viommu = vdev->priv;
1113 
1114 	iommu_device_sysfs_remove(&viommu->iommu);
1115 	iommu_device_unregister(&viommu->iommu);
1116 
1117 	/* Stop all virtqueues */
1118 	vdev->config->reset(vdev);
1119 	vdev->config->del_vqs(vdev);
1120 
1121 	dev_info(&vdev->dev, "device removed\n");
1122 }
1123 
1124 static void viommu_config_changed(struct virtio_device *vdev)
1125 {
1126 	dev_warn(&vdev->dev, "config changed\n");
1127 }
1128 
1129 static unsigned int features[] = {
1130 	VIRTIO_IOMMU_F_MAP_UNMAP,
1131 	VIRTIO_IOMMU_F_INPUT_RANGE,
1132 	VIRTIO_IOMMU_F_DOMAIN_RANGE,
1133 	VIRTIO_IOMMU_F_PROBE,
1134 	VIRTIO_IOMMU_F_MMIO,
1135 };
1136 
1137 static struct virtio_device_id id_table[] = {
1138 	{ VIRTIO_ID_IOMMU, VIRTIO_DEV_ANY_ID },
1139 	{ 0 },
1140 };
1141 
1142 static struct virtio_driver virtio_iommu_drv = {
1143 	.driver.name		= KBUILD_MODNAME,
1144 	.driver.owner		= THIS_MODULE,
1145 	.id_table		= id_table,
1146 	.feature_table		= features,
1147 	.feature_table_size	= ARRAY_SIZE(features),
1148 	.probe			= viommu_probe,
1149 	.remove			= viommu_remove,
1150 	.config_changed		= viommu_config_changed,
1151 };
1152 
1153 module_virtio_driver(virtio_iommu_drv);
1154 
1155 MODULE_DESCRIPTION("Virtio IOMMU driver");
1156 MODULE_AUTHOR("Jean-Philippe Brucker <jean-philippe.brucker@arm.com>");
1157 MODULE_LICENSE("GPL v2");
1158