1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2018-2020 Intel Corporation.
4 * Copyright (C) 2020 Red Hat, Inc.
6 * Author: Tiwei Bie <tiwei.bie@intel.com>
7 * Jason Wang <jasowang@redhat.com>
9 * Thanks Michael S. Tsirkin for the valuable comments and
10 * suggestions. And thanks to Cunming Liang and Zhihong Wang for all
14 #include <linux/kernel.h>
15 #include <linux/module.h>
16 #include <linux/cdev.h>
17 #include <linux/device.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>
29 VHOST_VDPA_BACKEND_FEATURES =
30 (1ULL << VHOST_BACKEND_F_IOTLB_MSG_V2) |
31 (1ULL << VHOST_BACKEND_F_IOTLB_BATCH) |
32 (1ULL << VHOST_BACKEND_F_IOTLB_ASID),
35 #define VHOST_VDPA_DEV_MAX (1U << MINORBITS)
37 #define VHOST_VDPA_IOTLB_BUCKETS 16
39 struct vhost_vdpa_as {
40 struct hlist_node hash_link;
41 struct vhost_iotlb iotlb;
46 struct vhost_dev vdev;
47 struct iommu_domain *domain;
48 struct vhost_virtqueue *vqs;
49 struct completion completion;
50 struct vdpa_device *vdpa;
51 struct hlist_head as[VHOST_VDPA_IOTLB_BUCKETS];
58 struct eventfd_ctx *config_ctx;
60 struct vdpa_iova_range range;
64 static DEFINE_IDA(vhost_vdpa_ida);
66 static dev_t vhost_vdpa_major;
68 static void vhost_vdpa_iotlb_unmap(struct vhost_vdpa *v,
69 struct vhost_iotlb *iotlb, u64 start,
72 static inline u32 iotlb_to_asid(struct vhost_iotlb *iotlb)
74 struct vhost_vdpa_as *as = container_of(iotlb, struct
75 vhost_vdpa_as, iotlb);
79 static struct vhost_vdpa_as *asid_to_as(struct vhost_vdpa *v, u32 asid)
81 struct hlist_head *head = &v->as[asid % VHOST_VDPA_IOTLB_BUCKETS];
82 struct vhost_vdpa_as *as;
84 hlist_for_each_entry(as, head, hash_link)
91 static struct vhost_iotlb *asid_to_iotlb(struct vhost_vdpa *v, u32 asid)
93 struct vhost_vdpa_as *as = asid_to_as(v, asid);
101 static struct vhost_vdpa_as *vhost_vdpa_alloc_as(struct vhost_vdpa *v, u32 asid)
103 struct hlist_head *head = &v->as[asid % VHOST_VDPA_IOTLB_BUCKETS];
104 struct vhost_vdpa_as *as;
106 if (asid_to_as(v, asid))
109 if (asid >= v->vdpa->nas)
112 as = kmalloc(sizeof(*as), GFP_KERNEL);
116 vhost_iotlb_init(&as->iotlb, 0, 0);
118 hlist_add_head(&as->hash_link, head);
123 static struct vhost_vdpa_as *vhost_vdpa_find_alloc_as(struct vhost_vdpa *v,
126 struct vhost_vdpa_as *as = asid_to_as(v, asid);
131 return vhost_vdpa_alloc_as(v, asid);
134 static int vhost_vdpa_remove_as(struct vhost_vdpa *v, u32 asid)
136 struct vhost_vdpa_as *as = asid_to_as(v, asid);
141 hlist_del(&as->hash_link);
142 vhost_vdpa_iotlb_unmap(v, &as->iotlb, 0ULL, 0ULL - 1, asid);
148 static void handle_vq_kick(struct vhost_work *work)
150 struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
152 struct vhost_vdpa *v = container_of(vq->dev, struct vhost_vdpa, vdev);
153 const struct vdpa_config_ops *ops = v->vdpa->config;
155 ops->kick_vq(v->vdpa, vq - v->vqs);
158 static irqreturn_t vhost_vdpa_virtqueue_cb(void *private)
160 struct vhost_virtqueue *vq = private;
161 struct eventfd_ctx *call_ctx = vq->call_ctx.ctx;
164 eventfd_signal(call_ctx, 1);
169 static irqreturn_t vhost_vdpa_config_cb(void *private)
171 struct vhost_vdpa *v = private;
172 struct eventfd_ctx *config_ctx = v->config_ctx;
175 eventfd_signal(config_ctx, 1);
180 static void vhost_vdpa_setup_vq_irq(struct vhost_vdpa *v, u16 qid)
182 struct vhost_virtqueue *vq = &v->vqs[qid];
183 const struct vdpa_config_ops *ops = v->vdpa->config;
184 struct vdpa_device *vdpa = v->vdpa;
187 if (!ops->get_vq_irq)
190 irq = ops->get_vq_irq(vdpa, qid);
194 irq_bypass_unregister_producer(&vq->call_ctx.producer);
195 if (!vq->call_ctx.ctx)
198 vq->call_ctx.producer.token = vq->call_ctx.ctx;
199 vq->call_ctx.producer.irq = irq;
200 ret = irq_bypass_register_producer(&vq->call_ctx.producer);
202 dev_info(&v->dev, "vq %u, irq bypass producer (token %p) registration fails, ret = %d\n",
203 qid, vq->call_ctx.producer.token, ret);
206 static void vhost_vdpa_unsetup_vq_irq(struct vhost_vdpa *v, u16 qid)
208 struct vhost_virtqueue *vq = &v->vqs[qid];
210 irq_bypass_unregister_producer(&vq->call_ctx.producer);
213 static int vhost_vdpa_reset(struct vhost_vdpa *v)
215 struct vdpa_device *vdpa = v->vdpa;
219 return vdpa_reset(vdpa);
222 static long vhost_vdpa_bind_mm(struct vhost_vdpa *v)
224 struct vdpa_device *vdpa = v->vdpa;
225 const struct vdpa_config_ops *ops = vdpa->config;
227 if (!vdpa->use_va || !ops->bind_mm)
230 return ops->bind_mm(vdpa, v->vdev.mm);
233 static void vhost_vdpa_unbind_mm(struct vhost_vdpa *v)
235 struct vdpa_device *vdpa = v->vdpa;
236 const struct vdpa_config_ops *ops = vdpa->config;
238 if (!vdpa->use_va || !ops->unbind_mm)
241 ops->unbind_mm(vdpa);
244 static long vhost_vdpa_get_device_id(struct vhost_vdpa *v, u8 __user *argp)
246 struct vdpa_device *vdpa = v->vdpa;
247 const struct vdpa_config_ops *ops = vdpa->config;
250 device_id = ops->get_device_id(vdpa);
252 if (copy_to_user(argp, &device_id, sizeof(device_id)))
258 static long vhost_vdpa_get_status(struct vhost_vdpa *v, u8 __user *statusp)
260 struct vdpa_device *vdpa = v->vdpa;
261 const struct vdpa_config_ops *ops = vdpa->config;
264 status = ops->get_status(vdpa);
266 if (copy_to_user(statusp, &status, sizeof(status)))
272 static long vhost_vdpa_set_status(struct vhost_vdpa *v, u8 __user *statusp)
274 struct vdpa_device *vdpa = v->vdpa;
275 const struct vdpa_config_ops *ops = vdpa->config;
276 u8 status, status_old;
281 if (copy_from_user(&status, statusp, sizeof(status)))
284 status_old = ops->get_status(vdpa);
287 * Userspace shouldn't remove status bits unless reset the
290 if (status != 0 && (status_old & ~status) != 0)
293 if ((status_old & VIRTIO_CONFIG_S_DRIVER_OK) && !(status & VIRTIO_CONFIG_S_DRIVER_OK))
294 for (i = 0; i < nvqs; i++)
295 vhost_vdpa_unsetup_vq_irq(v, i);
298 ret = vdpa_reset(vdpa);
302 vdpa_set_status(vdpa, status);
304 if ((status & VIRTIO_CONFIG_S_DRIVER_OK) && !(status_old & VIRTIO_CONFIG_S_DRIVER_OK))
305 for (i = 0; i < nvqs; i++)
306 vhost_vdpa_setup_vq_irq(v, i);
311 static int vhost_vdpa_config_validate(struct vhost_vdpa *v,
312 struct vhost_vdpa_config *c)
314 struct vdpa_device *vdpa = v->vdpa;
315 size_t size = vdpa->config->get_config_size(vdpa);
317 if (c->len == 0 || c->off > size)
320 if (c->len > size - c->off)
326 static long vhost_vdpa_get_config(struct vhost_vdpa *v,
327 struct vhost_vdpa_config __user *c)
329 struct vdpa_device *vdpa = v->vdpa;
330 struct vhost_vdpa_config config;
331 unsigned long size = offsetof(struct vhost_vdpa_config, buf);
334 if (copy_from_user(&config, c, size))
336 if (vhost_vdpa_config_validate(v, &config))
338 buf = kvzalloc(config.len, GFP_KERNEL);
342 vdpa_get_config(vdpa, config.off, buf, config.len);
344 if (copy_to_user(c->buf, buf, config.len)) {
353 static long vhost_vdpa_set_config(struct vhost_vdpa *v,
354 struct vhost_vdpa_config __user *c)
356 struct vdpa_device *vdpa = v->vdpa;
357 struct vhost_vdpa_config config;
358 unsigned long size = offsetof(struct vhost_vdpa_config, buf);
361 if (copy_from_user(&config, c, size))
363 if (vhost_vdpa_config_validate(v, &config))
366 buf = vmemdup_user(c->buf, config.len);
370 vdpa_set_config(vdpa, config.off, buf, config.len);
376 static bool vhost_vdpa_can_suspend(const struct vhost_vdpa *v)
378 struct vdpa_device *vdpa = v->vdpa;
379 const struct vdpa_config_ops *ops = vdpa->config;
384 static bool vhost_vdpa_can_resume(const struct vhost_vdpa *v)
386 struct vdpa_device *vdpa = v->vdpa;
387 const struct vdpa_config_ops *ops = vdpa->config;
392 static long vhost_vdpa_get_features(struct vhost_vdpa *v, u64 __user *featurep)
394 struct vdpa_device *vdpa = v->vdpa;
395 const struct vdpa_config_ops *ops = vdpa->config;
398 features = ops->get_device_features(vdpa);
400 if (copy_to_user(featurep, &features, sizeof(features)))
406 static long vhost_vdpa_set_features(struct vhost_vdpa *v, u64 __user *featurep)
408 struct vdpa_device *vdpa = v->vdpa;
409 const struct vdpa_config_ops *ops = vdpa->config;
410 struct vhost_dev *d = &v->vdev;
416 * It's not allowed to change the features after they have
419 if (ops->get_status(vdpa) & VIRTIO_CONFIG_S_FEATURES_OK)
422 if (copy_from_user(&features, featurep, sizeof(features)))
425 if (vdpa_set_features(vdpa, features))
428 /* let the vqs know what has been configured */
429 actual_features = ops->get_driver_features(vdpa);
430 for (i = 0; i < d->nvqs; ++i) {
431 struct vhost_virtqueue *vq = d->vqs[i];
433 mutex_lock(&vq->mutex);
434 vq->acked_features = actual_features;
435 mutex_unlock(&vq->mutex);
441 static long vhost_vdpa_get_vring_num(struct vhost_vdpa *v, u16 __user *argp)
443 struct vdpa_device *vdpa = v->vdpa;
444 const struct vdpa_config_ops *ops = vdpa->config;
447 num = ops->get_vq_num_max(vdpa);
449 if (copy_to_user(argp, &num, sizeof(num)))
455 static void vhost_vdpa_config_put(struct vhost_vdpa *v)
458 eventfd_ctx_put(v->config_ctx);
459 v->config_ctx = NULL;
463 static long vhost_vdpa_set_config_call(struct vhost_vdpa *v, u32 __user *argp)
465 struct vdpa_callback cb;
467 struct eventfd_ctx *ctx;
469 cb.callback = vhost_vdpa_config_cb;
471 if (copy_from_user(&fd, argp, sizeof(fd)))
474 ctx = fd == VHOST_FILE_UNBIND ? NULL : eventfd_ctx_fdget(fd);
475 swap(ctx, v->config_ctx);
477 if (!IS_ERR_OR_NULL(ctx))
478 eventfd_ctx_put(ctx);
480 if (IS_ERR(v->config_ctx)) {
481 long ret = PTR_ERR(v->config_ctx);
483 v->config_ctx = NULL;
487 v->vdpa->config->set_config_cb(v->vdpa, &cb);
492 static long vhost_vdpa_get_iova_range(struct vhost_vdpa *v, u32 __user *argp)
494 struct vhost_vdpa_iova_range range = {
495 .first = v->range.first,
496 .last = v->range.last,
499 if (copy_to_user(argp, &range, sizeof(range)))
504 static long vhost_vdpa_get_config_size(struct vhost_vdpa *v, u32 __user *argp)
506 struct vdpa_device *vdpa = v->vdpa;
507 const struct vdpa_config_ops *ops = vdpa->config;
510 size = ops->get_config_size(vdpa);
512 if (copy_to_user(argp, &size, sizeof(size)))
518 static long vhost_vdpa_get_vqs_count(struct vhost_vdpa *v, u32 __user *argp)
520 struct vdpa_device *vdpa = v->vdpa;
522 if (copy_to_user(argp, &vdpa->nvqs, sizeof(vdpa->nvqs)))
528 /* After a successful return of ioctl the device must not process more
529 * virtqueue descriptors. The device can answer to read or writes of config
530 * fields as if it were not suspended. In particular, writing to "queue_enable"
531 * with a value of 1 will not make the device start processing buffers.
533 static long vhost_vdpa_suspend(struct vhost_vdpa *v)
535 struct vdpa_device *vdpa = v->vdpa;
536 const struct vdpa_config_ops *ops = vdpa->config;
541 return ops->suspend(vdpa);
544 /* After a successful return of this ioctl the device resumes processing
545 * virtqueue descriptors. The device becomes fully operational the same way it
546 * was before it was suspended.
548 static long vhost_vdpa_resume(struct vhost_vdpa *v)
550 struct vdpa_device *vdpa = v->vdpa;
551 const struct vdpa_config_ops *ops = vdpa->config;
556 return ops->resume(vdpa);
559 static long vhost_vdpa_vring_ioctl(struct vhost_vdpa *v, unsigned int cmd,
562 struct vdpa_device *vdpa = v->vdpa;
563 const struct vdpa_config_ops *ops = vdpa->config;
564 struct vdpa_vq_state vq_state;
565 struct vdpa_callback cb;
566 struct vhost_virtqueue *vq;
567 struct vhost_vring_state s;
571 r = get_user(idx, (u32 __user *)argp);
578 idx = array_index_nospec(idx, v->nvqs);
582 case VHOST_VDPA_SET_VRING_ENABLE:
583 if (copy_from_user(&s, argp, sizeof(s)))
585 ops->set_vq_ready(vdpa, idx, s.num);
587 case VHOST_VDPA_GET_VRING_GROUP:
588 if (!ops->get_vq_group)
591 s.num = ops->get_vq_group(vdpa, idx);
592 if (s.num >= vdpa->ngroups)
594 else if (copy_to_user(argp, &s, sizeof(s)))
597 case VHOST_VDPA_SET_GROUP_ASID:
598 if (copy_from_user(&s, argp, sizeof(s)))
600 if (s.num >= vdpa->nas)
602 if (!ops->set_group_asid)
604 return ops->set_group_asid(vdpa, idx, s.num);
605 case VHOST_GET_VRING_BASE:
606 r = ops->get_vq_state(v->vdpa, idx, &vq_state);
610 if (vhost_has_feature(vq, VIRTIO_F_RING_PACKED)) {
611 vq->last_avail_idx = vq_state.packed.last_avail_idx |
612 (vq_state.packed.last_avail_counter << 15);
613 vq->last_used_idx = vq_state.packed.last_used_idx |
614 (vq_state.packed.last_used_counter << 15);
616 vq->last_avail_idx = vq_state.split.avail_index;
621 r = vhost_vring_ioctl(&v->vdev, cmd, argp);
626 case VHOST_SET_VRING_ADDR:
627 if (ops->set_vq_address(vdpa, idx,
628 (u64)(uintptr_t)vq->desc,
629 (u64)(uintptr_t)vq->avail,
630 (u64)(uintptr_t)vq->used))
634 case VHOST_SET_VRING_BASE:
635 if (vhost_has_feature(vq, VIRTIO_F_RING_PACKED)) {
636 vq_state.packed.last_avail_idx = vq->last_avail_idx & 0x7fff;
637 vq_state.packed.last_avail_counter = !!(vq->last_avail_idx & 0x8000);
638 vq_state.packed.last_used_idx = vq->last_used_idx & 0x7fff;
639 vq_state.packed.last_used_counter = !!(vq->last_used_idx & 0x8000);
641 vq_state.split.avail_index = vq->last_avail_idx;
643 r = ops->set_vq_state(vdpa, idx, &vq_state);
646 case VHOST_SET_VRING_CALL:
647 if (vq->call_ctx.ctx) {
648 cb.callback = vhost_vdpa_virtqueue_cb;
650 cb.trigger = vq->call_ctx.ctx;
656 ops->set_vq_cb(vdpa, idx, &cb);
657 vhost_vdpa_setup_vq_irq(v, idx);
660 case VHOST_SET_VRING_NUM:
661 ops->set_vq_num(vdpa, idx, vq->num);
668 static long vhost_vdpa_unlocked_ioctl(struct file *filep,
669 unsigned int cmd, unsigned long arg)
671 struct vhost_vdpa *v = filep->private_data;
672 struct vhost_dev *d = &v->vdev;
673 void __user *argp = (void __user *)arg;
674 u64 __user *featurep = argp;
678 if (cmd == VHOST_SET_BACKEND_FEATURES) {
679 if (copy_from_user(&features, featurep, sizeof(features)))
681 if (features & ~(VHOST_VDPA_BACKEND_FEATURES |
682 BIT_ULL(VHOST_BACKEND_F_SUSPEND) |
683 BIT_ULL(VHOST_BACKEND_F_RESUME)))
685 if ((features & BIT_ULL(VHOST_BACKEND_F_SUSPEND)) &&
686 !vhost_vdpa_can_suspend(v))
688 if ((features & BIT_ULL(VHOST_BACKEND_F_RESUME)) &&
689 !vhost_vdpa_can_resume(v))
691 vhost_set_backend_features(&v->vdev, features);
695 mutex_lock(&d->mutex);
698 case VHOST_VDPA_GET_DEVICE_ID:
699 r = vhost_vdpa_get_device_id(v, argp);
701 case VHOST_VDPA_GET_STATUS:
702 r = vhost_vdpa_get_status(v, argp);
704 case VHOST_VDPA_SET_STATUS:
705 r = vhost_vdpa_set_status(v, argp);
707 case VHOST_VDPA_GET_CONFIG:
708 r = vhost_vdpa_get_config(v, argp);
710 case VHOST_VDPA_SET_CONFIG:
711 r = vhost_vdpa_set_config(v, argp);
713 case VHOST_GET_FEATURES:
714 r = vhost_vdpa_get_features(v, argp);
716 case VHOST_SET_FEATURES:
717 r = vhost_vdpa_set_features(v, argp);
719 case VHOST_VDPA_GET_VRING_NUM:
720 r = vhost_vdpa_get_vring_num(v, argp);
722 case VHOST_VDPA_GET_GROUP_NUM:
723 if (copy_to_user(argp, &v->vdpa->ngroups,
724 sizeof(v->vdpa->ngroups)))
727 case VHOST_VDPA_GET_AS_NUM:
728 if (copy_to_user(argp, &v->vdpa->nas, sizeof(v->vdpa->nas)))
731 case VHOST_SET_LOG_BASE:
732 case VHOST_SET_LOG_FD:
735 case VHOST_VDPA_SET_CONFIG_CALL:
736 r = vhost_vdpa_set_config_call(v, argp);
738 case VHOST_GET_BACKEND_FEATURES:
739 features = VHOST_VDPA_BACKEND_FEATURES;
740 if (vhost_vdpa_can_suspend(v))
741 features |= BIT_ULL(VHOST_BACKEND_F_SUSPEND);
742 if (vhost_vdpa_can_resume(v))
743 features |= BIT_ULL(VHOST_BACKEND_F_RESUME);
744 if (copy_to_user(featurep, &features, sizeof(features)))
747 case VHOST_VDPA_GET_IOVA_RANGE:
748 r = vhost_vdpa_get_iova_range(v, argp);
750 case VHOST_VDPA_GET_CONFIG_SIZE:
751 r = vhost_vdpa_get_config_size(v, argp);
753 case VHOST_VDPA_GET_VQS_COUNT:
754 r = vhost_vdpa_get_vqs_count(v, argp);
756 case VHOST_VDPA_SUSPEND:
757 r = vhost_vdpa_suspend(v);
759 case VHOST_VDPA_RESUME:
760 r = vhost_vdpa_resume(v);
763 r = vhost_dev_ioctl(&v->vdev, cmd, argp);
764 if (r == -ENOIOCTLCMD)
765 r = vhost_vdpa_vring_ioctl(v, cmd, argp);
773 case VHOST_SET_OWNER:
774 r = vhost_vdpa_bind_mm(v);
776 vhost_dev_reset_owner(d, NULL);
780 mutex_unlock(&d->mutex);
783 static void vhost_vdpa_general_unmap(struct vhost_vdpa *v,
784 struct vhost_iotlb_map *map, u32 asid)
786 struct vdpa_device *vdpa = v->vdpa;
787 const struct vdpa_config_ops *ops = vdpa->config;
789 ops->dma_unmap(vdpa, asid, map->start, map->size);
790 } else if (ops->set_map == NULL) {
791 iommu_unmap(v->domain, map->start, map->size);
795 static void vhost_vdpa_pa_unmap(struct vhost_vdpa *v, struct vhost_iotlb *iotlb,
796 u64 start, u64 last, u32 asid)
798 struct vhost_dev *dev = &v->vdev;
799 struct vhost_iotlb_map *map;
801 unsigned long pfn, pinned;
803 while ((map = vhost_iotlb_itree_first(iotlb, start, last)) != NULL) {
804 pinned = PFN_DOWN(map->size);
805 for (pfn = PFN_DOWN(map->addr);
806 pinned > 0; pfn++, pinned--) {
807 page = pfn_to_page(pfn);
808 if (map->perm & VHOST_ACCESS_WO)
809 set_page_dirty_lock(page);
810 unpin_user_page(page);
812 atomic64_sub(PFN_DOWN(map->size), &dev->mm->pinned_vm);
813 vhost_vdpa_general_unmap(v, map, asid);
814 vhost_iotlb_map_free(iotlb, map);
818 static void vhost_vdpa_va_unmap(struct vhost_vdpa *v, struct vhost_iotlb *iotlb,
819 u64 start, u64 last, u32 asid)
821 struct vhost_iotlb_map *map;
822 struct vdpa_map_file *map_file;
824 while ((map = vhost_iotlb_itree_first(iotlb, start, last)) != NULL) {
825 map_file = (struct vdpa_map_file *)map->opaque;
826 fput(map_file->file);
828 vhost_vdpa_general_unmap(v, map, asid);
829 vhost_iotlb_map_free(iotlb, map);
833 static void vhost_vdpa_iotlb_unmap(struct vhost_vdpa *v,
834 struct vhost_iotlb *iotlb, u64 start,
837 struct vdpa_device *vdpa = v->vdpa;
840 return vhost_vdpa_va_unmap(v, iotlb, start, last, asid);
842 return vhost_vdpa_pa_unmap(v, iotlb, start, last, asid);
845 static int perm_to_iommu_flags(u32 perm)
850 case VHOST_ACCESS_WO:
851 flags |= IOMMU_WRITE;
853 case VHOST_ACCESS_RO:
856 case VHOST_ACCESS_RW:
857 flags |= (IOMMU_WRITE | IOMMU_READ);
860 WARN(1, "invalidate vhost IOTLB permission\n");
864 return flags | IOMMU_CACHE;
867 static int vhost_vdpa_map(struct vhost_vdpa *v, struct vhost_iotlb *iotlb,
868 u64 iova, u64 size, u64 pa, u32 perm, void *opaque)
870 struct vhost_dev *dev = &v->vdev;
871 struct vdpa_device *vdpa = v->vdpa;
872 const struct vdpa_config_ops *ops = vdpa->config;
873 u32 asid = iotlb_to_asid(iotlb);
876 r = vhost_iotlb_add_range_ctx(iotlb, iova, iova + size - 1,
882 r = ops->dma_map(vdpa, asid, iova, size, pa, perm, opaque);
883 } else if (ops->set_map) {
885 r = ops->set_map(vdpa, asid, iotlb);
887 r = iommu_map(v->domain, iova, pa, size,
888 perm_to_iommu_flags(perm), GFP_KERNEL);
891 vhost_iotlb_del_range(iotlb, iova, iova + size - 1);
896 atomic64_add(PFN_DOWN(size), &dev->mm->pinned_vm);
901 static void vhost_vdpa_unmap(struct vhost_vdpa *v,
902 struct vhost_iotlb *iotlb,
905 struct vdpa_device *vdpa = v->vdpa;
906 const struct vdpa_config_ops *ops = vdpa->config;
907 u32 asid = iotlb_to_asid(iotlb);
909 vhost_vdpa_iotlb_unmap(v, iotlb, iova, iova + size - 1, asid);
913 ops->set_map(vdpa, asid, iotlb);
918 static int vhost_vdpa_va_map(struct vhost_vdpa *v,
919 struct vhost_iotlb *iotlb,
920 u64 iova, u64 size, u64 uaddr, u32 perm)
922 struct vhost_dev *dev = &v->vdev;
923 u64 offset, map_size, map_iova = iova;
924 struct vdpa_map_file *map_file;
925 struct vm_area_struct *vma;
928 mmap_read_lock(dev->mm);
931 vma = find_vma(dev->mm, uaddr);
936 map_size = min(size, vma->vm_end - uaddr);
937 if (!(vma->vm_file && (vma->vm_flags & VM_SHARED) &&
938 !(vma->vm_flags & (VM_IO | VM_PFNMAP))))
941 map_file = kzalloc(sizeof(*map_file), GFP_KERNEL);
946 offset = (vma->vm_pgoff << PAGE_SHIFT) + uaddr - vma->vm_start;
947 map_file->offset = offset;
948 map_file->file = get_file(vma->vm_file);
949 ret = vhost_vdpa_map(v, iotlb, map_iova, map_size, uaddr,
952 fput(map_file->file);
959 map_iova += map_size;
962 vhost_vdpa_unmap(v, iotlb, iova, map_iova - iova);
964 mmap_read_unlock(dev->mm);
969 static int vhost_vdpa_pa_map(struct vhost_vdpa *v,
970 struct vhost_iotlb *iotlb,
971 u64 iova, u64 size, u64 uaddr, u32 perm)
973 struct vhost_dev *dev = &v->vdev;
974 struct page **page_list;
975 unsigned long list_size = PAGE_SIZE / sizeof(struct page *);
976 unsigned int gup_flags = FOLL_LONGTERM;
977 unsigned long npages, cur_base, map_pfn, last_pfn = 0;
978 unsigned long lock_limit, sz2pin, nchunks, i;
983 /* Limit the use of memory for bookkeeping */
984 page_list = (struct page **) __get_free_page(GFP_KERNEL);
988 if (perm & VHOST_ACCESS_WO)
989 gup_flags |= FOLL_WRITE;
991 npages = PFN_UP(size + (iova & ~PAGE_MASK));
997 mmap_read_lock(dev->mm);
999 lock_limit = PFN_DOWN(rlimit(RLIMIT_MEMLOCK));
1000 if (npages + atomic64_read(&dev->mm->pinned_vm) > lock_limit) {
1005 cur_base = uaddr & PAGE_MASK;
1010 sz2pin = min_t(unsigned long, npages, list_size);
1011 pinned = pin_user_pages(cur_base, sz2pin,
1012 gup_flags, page_list);
1013 if (sz2pin != pinned) {
1017 unpin_user_pages(page_list, pinned);
1025 map_pfn = page_to_pfn(page_list[0]);
1027 for (i = 0; i < pinned; i++) {
1028 unsigned long this_pfn = page_to_pfn(page_list[i]);
1031 if (last_pfn && (this_pfn != last_pfn + 1)) {
1032 /* Pin a contiguous chunk of memory */
1033 csize = PFN_PHYS(last_pfn - map_pfn + 1);
1034 ret = vhost_vdpa_map(v, iotlb, iova, csize,
1039 * Unpin the pages that are left unmapped
1040 * from this point on in the current
1041 * page_list. The remaining outstanding
1042 * ones which may stride across several
1043 * chunks will be covered in the common
1044 * error path subsequently.
1046 unpin_user_pages(&page_list[i],
1056 last_pfn = this_pfn;
1059 cur_base += PFN_PHYS(pinned);
1063 /* Pin the rest chunk */
1064 ret = vhost_vdpa_map(v, iotlb, iova, PFN_PHYS(last_pfn - map_pfn + 1),
1065 PFN_PHYS(map_pfn), perm, NULL);
1072 * Unpin the outstanding pages which are yet to be
1073 * mapped but haven't due to vdpa_map() or
1074 * pin_user_pages() failure.
1076 * Mapped pages are accounted in vdpa_map(), hence
1077 * the corresponding unpinning will be handled by
1081 for (pfn = map_pfn; pfn <= last_pfn; pfn++)
1082 unpin_user_page(pfn_to_page(pfn));
1084 vhost_vdpa_unmap(v, iotlb, start, size);
1087 mmap_read_unlock(dev->mm);
1089 free_page((unsigned long)page_list);
1094 static int vhost_vdpa_process_iotlb_update(struct vhost_vdpa *v,
1095 struct vhost_iotlb *iotlb,
1096 struct vhost_iotlb_msg *msg)
1098 struct vdpa_device *vdpa = v->vdpa;
1100 if (msg->iova < v->range.first || !msg->size ||
1101 msg->iova > U64_MAX - msg->size + 1 ||
1102 msg->iova + msg->size - 1 > v->range.last)
1105 if (vhost_iotlb_itree_first(iotlb, msg->iova,
1106 msg->iova + msg->size - 1))
1110 return vhost_vdpa_va_map(v, iotlb, msg->iova, msg->size,
1111 msg->uaddr, msg->perm);
1113 return vhost_vdpa_pa_map(v, iotlb, msg->iova, msg->size, msg->uaddr,
1117 static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
1118 struct vhost_iotlb_msg *msg)
1120 struct vhost_vdpa *v = container_of(dev, struct vhost_vdpa, vdev);
1121 struct vdpa_device *vdpa = v->vdpa;
1122 const struct vdpa_config_ops *ops = vdpa->config;
1123 struct vhost_iotlb *iotlb = NULL;
1124 struct vhost_vdpa_as *as = NULL;
1127 mutex_lock(&dev->mutex);
1129 r = vhost_dev_check_owner(dev);
1133 if (msg->type == VHOST_IOTLB_UPDATE ||
1134 msg->type == VHOST_IOTLB_BATCH_BEGIN) {
1135 as = vhost_vdpa_find_alloc_as(v, asid);
1137 dev_err(&v->dev, "can't find and alloc asid %d\n",
1144 iotlb = asid_to_iotlb(v, asid);
1146 if ((v->in_batch && v->batch_asid != asid) || !iotlb) {
1147 if (v->in_batch && v->batch_asid != asid) {
1148 dev_info(&v->dev, "batch id %d asid %d\n",
1149 v->batch_asid, asid);
1152 dev_err(&v->dev, "no iotlb for asid %d\n", asid);
1157 switch (msg->type) {
1158 case VHOST_IOTLB_UPDATE:
1159 r = vhost_vdpa_process_iotlb_update(v, iotlb, msg);
1161 case VHOST_IOTLB_INVALIDATE:
1162 vhost_vdpa_unmap(v, iotlb, msg->iova, msg->size);
1164 case VHOST_IOTLB_BATCH_BEGIN:
1165 v->batch_asid = asid;
1168 case VHOST_IOTLB_BATCH_END:
1169 if (v->in_batch && ops->set_map)
1170 ops->set_map(vdpa, asid, iotlb);
1171 v->in_batch = false;
1178 mutex_unlock(&dev->mutex);
1183 static ssize_t vhost_vdpa_chr_write_iter(struct kiocb *iocb,
1184 struct iov_iter *from)
1186 struct file *file = iocb->ki_filp;
1187 struct vhost_vdpa *v = file->private_data;
1188 struct vhost_dev *dev = &v->vdev;
1190 return vhost_chr_write_iter(dev, from);
1193 static int vhost_vdpa_alloc_domain(struct vhost_vdpa *v)
1195 struct vdpa_device *vdpa = v->vdpa;
1196 const struct vdpa_config_ops *ops = vdpa->config;
1197 struct device *dma_dev = vdpa_get_dma_dev(vdpa);
1198 const struct bus_type *bus;
1201 /* Device want to do DMA by itself */
1202 if (ops->set_map || ops->dma_map)
1209 if (!device_iommu_capable(dma_dev, IOMMU_CAP_CACHE_COHERENCY)) {
1210 dev_warn_once(&v->dev,
1211 "Failed to allocate domain, device is not IOMMU cache coherent capable\n");
1215 v->domain = iommu_domain_alloc(bus);
1219 ret = iommu_attach_device(v->domain, dma_dev);
1226 iommu_domain_free(v->domain);
1231 static void vhost_vdpa_free_domain(struct vhost_vdpa *v)
1233 struct vdpa_device *vdpa = v->vdpa;
1234 struct device *dma_dev = vdpa_get_dma_dev(vdpa);
1237 iommu_detach_device(v->domain, dma_dev);
1238 iommu_domain_free(v->domain);
1244 static void vhost_vdpa_set_iova_range(struct vhost_vdpa *v)
1246 struct vdpa_iova_range *range = &v->range;
1247 struct vdpa_device *vdpa = v->vdpa;
1248 const struct vdpa_config_ops *ops = vdpa->config;
1250 if (ops->get_iova_range) {
1251 *range = ops->get_iova_range(vdpa);
1252 } else if (v->domain && v->domain->geometry.force_aperture) {
1253 range->first = v->domain->geometry.aperture_start;
1254 range->last = v->domain->geometry.aperture_end;
1257 range->last = ULLONG_MAX;
1261 static void vhost_vdpa_cleanup(struct vhost_vdpa *v)
1263 struct vhost_vdpa_as *as;
1266 for (asid = 0; asid < v->vdpa->nas; asid++) {
1267 as = asid_to_as(v, asid);
1269 vhost_vdpa_remove_as(v, asid);
1272 vhost_vdpa_free_domain(v);
1273 vhost_dev_cleanup(&v->vdev);
1277 static int vhost_vdpa_open(struct inode *inode, struct file *filep)
1279 struct vhost_vdpa *v;
1280 struct vhost_dev *dev;
1281 struct vhost_virtqueue **vqs;
1285 v = container_of(inode->i_cdev, struct vhost_vdpa, cdev);
1287 opened = atomic_cmpxchg(&v->opened, 0, 1);
1292 r = vhost_vdpa_reset(v);
1296 vqs = kmalloc_array(nvqs, sizeof(*vqs), GFP_KERNEL);
1303 for (i = 0; i < nvqs; i++) {
1304 vqs[i] = &v->vqs[i];
1305 vqs[i]->handle_kick = handle_vq_kick;
1307 vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, false,
1308 vhost_vdpa_process_iotlb_msg);
1310 r = vhost_vdpa_alloc_domain(v);
1312 goto err_alloc_domain;
1314 vhost_vdpa_set_iova_range(v);
1316 filep->private_data = v;
1321 vhost_vdpa_cleanup(v);
1323 atomic_dec(&v->opened);
1327 static void vhost_vdpa_clean_irq(struct vhost_vdpa *v)
1331 for (i = 0; i < v->nvqs; i++)
1332 vhost_vdpa_unsetup_vq_irq(v, i);
1335 static int vhost_vdpa_release(struct inode *inode, struct file *filep)
1337 struct vhost_vdpa *v = filep->private_data;
1338 struct vhost_dev *d = &v->vdev;
1340 mutex_lock(&d->mutex);
1341 filep->private_data = NULL;
1342 vhost_vdpa_clean_irq(v);
1343 vhost_vdpa_reset(v);
1344 vhost_dev_stop(&v->vdev);
1345 vhost_vdpa_unbind_mm(v);
1346 vhost_vdpa_config_put(v);
1347 vhost_vdpa_cleanup(v);
1348 mutex_unlock(&d->mutex);
1350 atomic_dec(&v->opened);
1351 complete(&v->completion);
1357 static vm_fault_t vhost_vdpa_fault(struct vm_fault *vmf)
1359 struct vhost_vdpa *v = vmf->vma->vm_file->private_data;
1360 struct vdpa_device *vdpa = v->vdpa;
1361 const struct vdpa_config_ops *ops = vdpa->config;
1362 struct vdpa_notification_area notify;
1363 struct vm_area_struct *vma = vmf->vma;
1364 u16 index = vma->vm_pgoff;
1366 notify = ops->get_vq_notification(vdpa, index);
1368 vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
1369 if (remap_pfn_range(vma, vmf->address & PAGE_MASK,
1370 PFN_DOWN(notify.addr), PAGE_SIZE,
1372 return VM_FAULT_SIGBUS;
1374 return VM_FAULT_NOPAGE;
1377 static const struct vm_operations_struct vhost_vdpa_vm_ops = {
1378 .fault = vhost_vdpa_fault,
1381 static int vhost_vdpa_mmap(struct file *file, struct vm_area_struct *vma)
1383 struct vhost_vdpa *v = vma->vm_file->private_data;
1384 struct vdpa_device *vdpa = v->vdpa;
1385 const struct vdpa_config_ops *ops = vdpa->config;
1386 struct vdpa_notification_area notify;
1387 unsigned long index = vma->vm_pgoff;
1389 if (vma->vm_end - vma->vm_start != PAGE_SIZE)
1391 if ((vma->vm_flags & VM_SHARED) == 0)
1393 if (vma->vm_flags & VM_READ)
1397 if (!ops->get_vq_notification)
1400 /* To be safe and easily modelled by userspace, We only
1401 * support the doorbell which sits on the page boundary and
1402 * does not share the page with other registers.
1404 notify = ops->get_vq_notification(vdpa, index);
1405 if (notify.addr & (PAGE_SIZE - 1))
1407 if (vma->vm_end - vma->vm_start != notify.size)
1410 vm_flags_set(vma, VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP);
1411 vma->vm_ops = &vhost_vdpa_vm_ops;
1414 #endif /* CONFIG_MMU */
1416 static const struct file_operations vhost_vdpa_fops = {
1417 .owner = THIS_MODULE,
1418 .open = vhost_vdpa_open,
1419 .release = vhost_vdpa_release,
1420 .write_iter = vhost_vdpa_chr_write_iter,
1421 .unlocked_ioctl = vhost_vdpa_unlocked_ioctl,
1423 .mmap = vhost_vdpa_mmap,
1424 #endif /* CONFIG_MMU */
1425 .compat_ioctl = compat_ptr_ioctl,
1428 static void vhost_vdpa_release_dev(struct device *device)
1430 struct vhost_vdpa *v =
1431 container_of(device, struct vhost_vdpa, dev);
1433 ida_simple_remove(&vhost_vdpa_ida, v->minor);
1438 static int vhost_vdpa_probe(struct vdpa_device *vdpa)
1440 const struct vdpa_config_ops *ops = vdpa->config;
1441 struct vhost_vdpa *v;
1445 /* We can't support platform IOMMU device with more than 1
1448 if (!ops->set_map && !ops->dma_map &&
1449 (vdpa->ngroups > 1 || vdpa->nas > 1))
1452 v = kzalloc(sizeof(*v), GFP_KERNEL | __GFP_RETRY_MAYFAIL);
1456 minor = ida_simple_get(&vhost_vdpa_ida, 0,
1457 VHOST_VDPA_DEV_MAX, GFP_KERNEL);
1463 atomic_set(&v->opened, 0);
1466 v->nvqs = vdpa->nvqs;
1467 v->virtio_id = ops->get_device_id(vdpa);
1469 device_initialize(&v->dev);
1470 v->dev.release = vhost_vdpa_release_dev;
1471 v->dev.parent = &vdpa->dev;
1472 v->dev.devt = MKDEV(MAJOR(vhost_vdpa_major), minor);
1473 v->vqs = kmalloc_array(v->nvqs, sizeof(struct vhost_virtqueue),
1480 r = dev_set_name(&v->dev, "vhost-vdpa-%u", minor);
1484 cdev_init(&v->cdev, &vhost_vdpa_fops);
1485 v->cdev.owner = THIS_MODULE;
1487 r = cdev_device_add(&v->cdev, &v->dev);
1491 init_completion(&v->completion);
1492 vdpa_set_drvdata(vdpa, v);
1494 for (i = 0; i < VHOST_VDPA_IOTLB_BUCKETS; i++)
1495 INIT_HLIST_HEAD(&v->as[i]);
1500 put_device(&v->dev);
1501 ida_simple_remove(&vhost_vdpa_ida, v->minor);
1505 static void vhost_vdpa_remove(struct vdpa_device *vdpa)
1507 struct vhost_vdpa *v = vdpa_get_drvdata(vdpa);
1510 cdev_device_del(&v->cdev, &v->dev);
1513 opened = atomic_cmpxchg(&v->opened, 0, 1);
1516 wait_for_completion(&v->completion);
1519 put_device(&v->dev);
1522 static struct vdpa_driver vhost_vdpa_driver = {
1524 .name = "vhost_vdpa",
1526 .probe = vhost_vdpa_probe,
1527 .remove = vhost_vdpa_remove,
1530 static int __init vhost_vdpa_init(void)
1534 r = alloc_chrdev_region(&vhost_vdpa_major, 0, VHOST_VDPA_DEV_MAX,
1537 goto err_alloc_chrdev;
1539 r = vdpa_register_driver(&vhost_vdpa_driver);
1541 goto err_vdpa_register_driver;
1545 err_vdpa_register_driver:
1546 unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1550 module_init(vhost_vdpa_init);
1552 static void __exit vhost_vdpa_exit(void)
1554 vdpa_unregister_driver(&vhost_vdpa_driver);
1555 unregister_chrdev_region(vhost_vdpa_major, VHOST_VDPA_DEV_MAX);
1557 module_exit(vhost_vdpa_exit);
1559 MODULE_VERSION("0.0.1");
1560 MODULE_LICENSE("GPL v2");
1561 MODULE_AUTHOR("Intel Corporation");
1562 MODULE_DESCRIPTION("vDPA-based vhost backend for virtio");