1 // SPDX-License-Identifier: GPL-2.0-only
3 * VFIO: IOMMU DMA mapping support for Type1 IOMMU
5 * Copyright (C) 2012 Red Hat, Inc. All rights reserved.
6 * Author: Alex Williamson <alex.williamson@redhat.com>
8 * Derived from original vfio:
9 * Copyright 2010 Cisco Systems, Inc. All rights reserved.
10 * Author: Tom Lyon, pugs@cisco.com
12 * We arbitrarily define a Type1 IOMMU as one matching the below code.
13 * It could be called the x86 IOMMU as it's designed for AMD-Vi & Intel
14 * VT-d, but that makes it harder to re-use as theoretically anyone
15 * implementing a similar IOMMU could make use of this. We expect the
16 * IOMMU to support the IOMMU API and have few to no restrictions around
17 * the IOVA range that can be mapped. The Type1 IOMMU is currently
18 * optimized for relatively static mappings of a userspace process with
19 * userspace pages pinned into memory. We also assume devices and IOMMU
20 * domains are PCI based as the IOMMU API is still centered around a
21 * device/bus interface rather than a group interface.
24 #include <linux/compat.h>
25 #include <linux/device.h>
27 #include <linux/highmem.h>
28 #include <linux/iommu.h>
29 #include <linux/module.h>
31 #include <linux/kthread.h>
32 #include <linux/rbtree.h>
33 #include <linux/sched/signal.h>
34 #include <linux/sched/mm.h>
35 #include <linux/slab.h>
36 #include <linux/uaccess.h>
37 #include <linux/vfio.h>
38 #include <linux/workqueue.h>
39 #include <linux/mdev.h>
40 #include <linux/notifier.h>
41 #include <linux/dma-iommu.h>
42 #include <linux/irqdomain.h>
44 #define DRIVER_VERSION "0.2"
45 #define DRIVER_AUTHOR "Alex Williamson <alex.williamson@redhat.com>"
46 #define DRIVER_DESC "Type1 IOMMU driver for VFIO"
48 static bool allow_unsafe_interrupts;
49 module_param_named(allow_unsafe_interrupts,
50 allow_unsafe_interrupts, bool, S_IRUGO | S_IWUSR);
51 MODULE_PARM_DESC(allow_unsafe_interrupts,
52 "Enable VFIO IOMMU support for on platforms without interrupt remapping support.");
54 static bool disable_hugepages;
55 module_param_named(disable_hugepages,
56 disable_hugepages, bool, S_IRUGO | S_IWUSR);
57 MODULE_PARM_DESC(disable_hugepages,
58 "Disable VFIO IOMMU support for IOMMU hugepages.");
60 static unsigned int dma_entry_limit __read_mostly = U16_MAX;
61 module_param_named(dma_entry_limit, dma_entry_limit, uint, 0644);
62 MODULE_PARM_DESC(dma_entry_limit,
63 "Maximum number of user DMA mappings per container (65535).");
66 struct list_head domain_list;
67 struct list_head iova_list;
68 struct vfio_domain *external_domain; /* domain for external user */
70 struct rb_root dma_list;
71 struct blocking_notifier_head notifier;
72 unsigned int dma_avail;
73 unsigned int vaddr_invalid_count;
74 uint64_t pgsize_bitmap;
75 uint64_t num_non_pinned_groups;
76 wait_queue_head_t vaddr_wait;
79 bool dirty_page_tracking;
84 struct iommu_domain *domain;
85 struct list_head next;
86 struct list_head group_list;
87 int prot; /* IOMMU_CACHE */
88 bool fgsp; /* Fine-grained super pages */
93 dma_addr_t iova; /* Device address */
94 unsigned long vaddr; /* Process virtual addr */
95 size_t size; /* Map size (bytes) */
96 int prot; /* IOMMU_READ/WRITE */
98 bool lock_cap; /* capable(CAP_IPC_LOCK) */
100 struct task_struct *task;
101 struct rb_root pfn_list; /* Ex-user pinned pfn list */
102 unsigned long *bitmap;
106 struct page **pages; /* for pin_user_pages_remote */
107 struct page *fallback_page; /* if pages alloc fails */
108 int capacity; /* length of pages array */
109 int size; /* of batch currently */
110 int offset; /* of next entry in pages */
113 struct vfio_iommu_group {
114 struct iommu_group *iommu_group;
115 struct list_head next;
116 bool mdev_group; /* An mdev group */
117 bool pinned_page_dirty_scope;
121 struct list_head list;
127 * Guest RAM pinning working set or DMA target
131 dma_addr_t iova; /* Device address */
132 unsigned long pfn; /* Host pfn */
133 unsigned int ref_count;
136 struct vfio_regions {
137 struct list_head list;
143 #define IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu) \
144 (!list_empty(&iommu->domain_list))
146 #define DIRTY_BITMAP_BYTES(n) (ALIGN(n, BITS_PER_TYPE(u64)) / BITS_PER_BYTE)
149 * Input argument of number of bits to bitmap_set() is unsigned integer, which
150 * further casts to signed integer for unaligned multi-bit operation,
152 * Then maximum bitmap size supported is 2^31 bits divided by 2^3 bits/byte,
153 * that is 2^28 (256 MB) which maps to 2^31 * 2^12 = 2^43 (8TB) on 4K page
156 #define DIRTY_BITMAP_PAGES_MAX ((u64)INT_MAX)
157 #define DIRTY_BITMAP_SIZE_MAX DIRTY_BITMAP_BYTES(DIRTY_BITMAP_PAGES_MAX)
161 static int put_pfn(unsigned long pfn, int prot);
163 static struct vfio_iommu_group*
164 vfio_iommu_find_iommu_group(struct vfio_iommu *iommu,
165 struct iommu_group *iommu_group);
168 * This code handles mapping and unmapping of user data buffers
169 * into DMA'ble space using the IOMMU
172 static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu,
173 dma_addr_t start, size_t size)
175 struct rb_node *node = iommu->dma_list.rb_node;
178 struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node);
180 if (start + size <= dma->iova)
181 node = node->rb_left;
182 else if (start >= dma->iova + dma->size)
183 node = node->rb_right;
191 static struct rb_node *vfio_find_dma_first_node(struct vfio_iommu *iommu,
192 dma_addr_t start, u64 size)
194 struct rb_node *res = NULL;
195 struct rb_node *node = iommu->dma_list.rb_node;
196 struct vfio_dma *dma_res = NULL;
199 struct vfio_dma *dma = rb_entry(node, struct vfio_dma, node);
201 if (start < dma->iova + dma->size) {
204 if (start >= dma->iova)
206 node = node->rb_left;
208 node = node->rb_right;
211 if (res && size && dma_res->iova >= start + size)
216 static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new)
218 struct rb_node **link = &iommu->dma_list.rb_node, *parent = NULL;
219 struct vfio_dma *dma;
223 dma = rb_entry(parent, struct vfio_dma, node);
225 if (new->iova + new->size <= dma->iova)
226 link = &(*link)->rb_left;
228 link = &(*link)->rb_right;
231 rb_link_node(&new->node, parent, link);
232 rb_insert_color(&new->node, &iommu->dma_list);
235 static void vfio_unlink_dma(struct vfio_iommu *iommu, struct vfio_dma *old)
237 rb_erase(&old->node, &iommu->dma_list);
241 static int vfio_dma_bitmap_alloc(struct vfio_dma *dma, size_t pgsize)
243 uint64_t npages = dma->size / pgsize;
245 if (npages > DIRTY_BITMAP_PAGES_MAX)
249 * Allocate extra 64 bits that are used to calculate shift required for
250 * bitmap_shift_left() to manipulate and club unaligned number of pages
251 * in adjacent vfio_dma ranges.
253 dma->bitmap = kvzalloc(DIRTY_BITMAP_BYTES(npages) + sizeof(u64),
261 static void vfio_dma_bitmap_free(struct vfio_dma *dma)
267 static void vfio_dma_populate_bitmap(struct vfio_dma *dma, size_t pgsize)
270 unsigned long pgshift = __ffs(pgsize);
272 for (p = rb_first(&dma->pfn_list); p; p = rb_next(p)) {
273 struct vfio_pfn *vpfn = rb_entry(p, struct vfio_pfn, node);
275 bitmap_set(dma->bitmap, (vpfn->iova - dma->iova) >> pgshift, 1);
279 static void vfio_iommu_populate_bitmap_full(struct vfio_iommu *iommu)
282 unsigned long pgshift = __ffs(iommu->pgsize_bitmap);
284 for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
285 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
287 bitmap_set(dma->bitmap, 0, dma->size >> pgshift);
291 static int vfio_dma_bitmap_alloc_all(struct vfio_iommu *iommu, size_t pgsize)
295 for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
296 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
299 ret = vfio_dma_bitmap_alloc(dma, pgsize);
303 for (p = rb_prev(n); p; p = rb_prev(p)) {
304 struct vfio_dma *dma = rb_entry(n,
305 struct vfio_dma, node);
307 vfio_dma_bitmap_free(dma);
311 vfio_dma_populate_bitmap(dma, pgsize);
316 static void vfio_dma_bitmap_free_all(struct vfio_iommu *iommu)
320 for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
321 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
323 vfio_dma_bitmap_free(dma);
328 * Helper Functions for host iova-pfn list
330 static struct vfio_pfn *vfio_find_vpfn(struct vfio_dma *dma, dma_addr_t iova)
332 struct vfio_pfn *vpfn;
333 struct rb_node *node = dma->pfn_list.rb_node;
336 vpfn = rb_entry(node, struct vfio_pfn, node);
338 if (iova < vpfn->iova)
339 node = node->rb_left;
340 else if (iova > vpfn->iova)
341 node = node->rb_right;
348 static void vfio_link_pfn(struct vfio_dma *dma,
349 struct vfio_pfn *new)
351 struct rb_node **link, *parent = NULL;
352 struct vfio_pfn *vpfn;
354 link = &dma->pfn_list.rb_node;
357 vpfn = rb_entry(parent, struct vfio_pfn, node);
359 if (new->iova < vpfn->iova)
360 link = &(*link)->rb_left;
362 link = &(*link)->rb_right;
365 rb_link_node(&new->node, parent, link);
366 rb_insert_color(&new->node, &dma->pfn_list);
369 static void vfio_unlink_pfn(struct vfio_dma *dma, struct vfio_pfn *old)
371 rb_erase(&old->node, &dma->pfn_list);
374 static int vfio_add_to_pfn_list(struct vfio_dma *dma, dma_addr_t iova,
377 struct vfio_pfn *vpfn;
379 vpfn = kzalloc(sizeof(*vpfn), GFP_KERNEL);
386 vfio_link_pfn(dma, vpfn);
390 static void vfio_remove_from_pfn_list(struct vfio_dma *dma,
391 struct vfio_pfn *vpfn)
393 vfio_unlink_pfn(dma, vpfn);
397 static struct vfio_pfn *vfio_iova_get_vfio_pfn(struct vfio_dma *dma,
400 struct vfio_pfn *vpfn = vfio_find_vpfn(dma, iova);
407 static int vfio_iova_put_vfio_pfn(struct vfio_dma *dma, struct vfio_pfn *vpfn)
412 if (!vpfn->ref_count) {
413 ret = put_pfn(vpfn->pfn, dma->prot);
414 vfio_remove_from_pfn_list(dma, vpfn);
419 static int vfio_lock_acct(struct vfio_dma *dma, long npage, bool async)
421 struct mm_struct *mm;
427 mm = async ? get_task_mm(dma->task) : dma->task->mm;
429 return -ESRCH; /* process exited */
431 ret = mmap_write_lock_killable(mm);
433 ret = __account_locked_vm(mm, abs(npage), npage > 0, dma->task,
435 mmap_write_unlock(mm);
445 * Some mappings aren't backed by a struct page, for example an mmap'd
446 * MMIO range for our own or another device. These use a different
447 * pfn conversion and shouldn't be tracked as locked pages.
448 * For compound pages, any driver that sets the reserved bit in head
449 * page needs to set the reserved bit in all subpages to be safe.
451 static bool is_invalid_reserved_pfn(unsigned long pfn)
454 return PageReserved(pfn_to_page(pfn));
459 static int put_pfn(unsigned long pfn, int prot)
461 if (!is_invalid_reserved_pfn(pfn)) {
462 struct page *page = pfn_to_page(pfn);
464 unpin_user_pages_dirty_lock(&page, 1, prot & IOMMU_WRITE);
470 #define VFIO_BATCH_MAX_CAPACITY (PAGE_SIZE / sizeof(struct page *))
472 static void vfio_batch_init(struct vfio_batch *batch)
477 if (unlikely(disable_hugepages))
480 batch->pages = (struct page **) __get_free_page(GFP_KERNEL);
484 batch->capacity = VFIO_BATCH_MAX_CAPACITY;
488 batch->pages = &batch->fallback_page;
492 static void vfio_batch_unpin(struct vfio_batch *batch, struct vfio_dma *dma)
494 while (batch->size) {
495 unsigned long pfn = page_to_pfn(batch->pages[batch->offset]);
497 put_pfn(pfn, dma->prot);
503 static void vfio_batch_fini(struct vfio_batch *batch)
505 if (batch->capacity == VFIO_BATCH_MAX_CAPACITY)
506 free_page((unsigned long)batch->pages);
509 static int follow_fault_pfn(struct vm_area_struct *vma, struct mm_struct *mm,
510 unsigned long vaddr, unsigned long *pfn,
517 ret = follow_pte(vma->vm_mm, vaddr, &ptep, &ptl);
519 bool unlocked = false;
521 ret = fixup_user_fault(mm, vaddr,
523 (write_fault ? FAULT_FLAG_WRITE : 0),
531 ret = follow_pte(vma->vm_mm, vaddr, &ptep, &ptl);
536 if (write_fault && !pte_write(*ptep))
539 *pfn = pte_pfn(*ptep);
541 pte_unmap_unlock(ptep, ptl);
546 * Returns the positive number of pfns successfully obtained or a negative
549 static int vaddr_get_pfns(struct mm_struct *mm, unsigned long vaddr,
550 long npages, int prot, unsigned long *pfn,
553 struct vm_area_struct *vma;
554 unsigned int flags = 0;
557 if (prot & IOMMU_WRITE)
561 ret = pin_user_pages_remote(mm, vaddr, npages, flags | FOLL_LONGTERM,
567 * The zero page is always resident, we don't need to pin it
568 * and it falls into our invalid/reserved test so we don't
569 * unpin in put_pfn(). Unpin all zero pages in the batch here.
571 for (i = 0 ; i < ret; i++) {
572 if (unlikely(is_zero_pfn(page_to_pfn(pages[i]))))
573 unpin_user_page(pages[i]);
576 *pfn = page_to_pfn(pages[0]);
580 vaddr = untagged_addr(vaddr);
583 vma = vma_lookup(mm, vaddr);
585 if (vma && vma->vm_flags & VM_PFNMAP) {
586 ret = follow_fault_pfn(vma, mm, vaddr, pfn, prot & IOMMU_WRITE);
591 if (is_invalid_reserved_pfn(*pfn))
598 mmap_read_unlock(mm);
602 static int vfio_wait(struct vfio_iommu *iommu)
606 prepare_to_wait(&iommu->vaddr_wait, &wait, TASK_KILLABLE);
607 mutex_unlock(&iommu->lock);
609 mutex_lock(&iommu->lock);
610 finish_wait(&iommu->vaddr_wait, &wait);
611 if (kthread_should_stop() || !iommu->container_open ||
612 fatal_signal_pending(current)) {
619 * Find dma struct and wait for its vaddr to be valid. iommu lock is dropped
620 * if the task waits, but is re-locked on return. Return result in *dma_p.
621 * Return 0 on success with no waiting, WAITED on success if waited, and -errno
624 static int vfio_find_dma_valid(struct vfio_iommu *iommu, dma_addr_t start,
625 size_t size, struct vfio_dma **dma_p)
630 *dma_p = vfio_find_dma(iommu, start, size);
633 else if (!(*dma_p)->vaddr_invalid)
636 ret = vfio_wait(iommu);
637 } while (ret == WAITED);
643 * Wait for all vaddr in the dma_list to become valid. iommu lock is dropped
644 * if the task waits, but is re-locked on return. Return 0 on success with no
645 * waiting, WAITED on success if waited, and -errno on error.
647 static int vfio_wait_all_valid(struct vfio_iommu *iommu)
651 while (iommu->vaddr_invalid_count && ret >= 0)
652 ret = vfio_wait(iommu);
658 * Attempt to pin pages. We really don't want to track all the pfns and
659 * the iommu can only map chunks of consecutive pfns anyway, so get the
660 * first page and all consecutive pages with the same locking.
662 static long vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
663 long npage, unsigned long *pfn_base,
664 unsigned long limit, struct vfio_batch *batch)
667 struct mm_struct *mm = current->mm;
668 long ret, pinned = 0, lock_acct = 0;
670 dma_addr_t iova = vaddr - dma->vaddr + dma->iova;
672 /* This code path is only user initiated */
677 /* Leftover pages in batch from an earlier call. */
678 *pfn_base = page_to_pfn(batch->pages[batch->offset]);
680 rsvd = is_invalid_reserved_pfn(*pfn_base);
687 /* Empty batch, so refill it. */
688 long req_pages = min_t(long, npage, batch->capacity);
690 ret = vaddr_get_pfns(mm, vaddr, req_pages, dma->prot,
700 rsvd = is_invalid_reserved_pfn(*pfn_base);
705 * pfn is preset for the first iteration of this inner loop and
706 * updated at the end to handle a VM_PFNMAP pfn. In that case,
707 * batch->pages isn't valid (there's no struct page), so allow
708 * batch->pages to be touched only when there's more than one
709 * pfn to check, which guarantees the pfns are from a
713 if (pfn != *pfn_base + pinned ||
714 rsvd != is_invalid_reserved_pfn(pfn))
718 * Reserved pages aren't counted against the user,
719 * externally pinned pages are already counted against
722 if (!rsvd && !vfio_find_vpfn(dma, iova)) {
723 if (!dma->lock_cap &&
724 mm->locked_vm + lock_acct + 1 > limit) {
725 pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n",
726 __func__, limit << PAGE_SHIFT);
743 pfn = page_to_pfn(batch->pages[batch->offset]);
746 if (unlikely(disable_hugepages))
751 ret = vfio_lock_acct(dma, lock_acct, false);
754 if (batch->size == 1 && !batch->offset) {
755 /* May be a VM_PFNMAP pfn, which the batch can't remember. */
756 put_pfn(pfn, dma->prot);
761 if (pinned && !rsvd) {
762 for (pfn = *pfn_base ; pinned ; pfn++, pinned--)
763 put_pfn(pfn, dma->prot);
765 vfio_batch_unpin(batch, dma);
773 static long vfio_unpin_pages_remote(struct vfio_dma *dma, dma_addr_t iova,
774 unsigned long pfn, long npage,
777 long unlocked = 0, locked = 0;
780 for (i = 0; i < npage; i++, iova += PAGE_SIZE) {
781 if (put_pfn(pfn++, dma->prot)) {
783 if (vfio_find_vpfn(dma, iova))
789 vfio_lock_acct(dma, locked - unlocked, true);
794 static int vfio_pin_page_external(struct vfio_dma *dma, unsigned long vaddr,
795 unsigned long *pfn_base, bool do_accounting)
797 struct page *pages[1];
798 struct mm_struct *mm;
801 mm = get_task_mm(dma->task);
805 ret = vaddr_get_pfns(mm, vaddr, 1, dma->prot, pfn_base, pages);
811 if (do_accounting && !is_invalid_reserved_pfn(*pfn_base)) {
812 ret = vfio_lock_acct(dma, 1, true);
814 put_pfn(*pfn_base, dma->prot);
816 pr_warn("%s: Task %s (%d) RLIMIT_MEMLOCK "
817 "(%ld) exceeded\n", __func__,
818 dma->task->comm, task_pid_nr(dma->task),
819 task_rlimit(dma->task, RLIMIT_MEMLOCK));
828 static int vfio_unpin_page_external(struct vfio_dma *dma, dma_addr_t iova,
832 struct vfio_pfn *vpfn = vfio_find_vpfn(dma, iova);
837 unlocked = vfio_iova_put_vfio_pfn(dma, vpfn);
840 vfio_lock_acct(dma, -unlocked, true);
845 static int vfio_iommu_type1_pin_pages(void *iommu_data,
846 struct iommu_group *iommu_group,
847 unsigned long *user_pfn,
849 unsigned long *phys_pfn)
851 struct vfio_iommu *iommu = iommu_data;
852 struct vfio_iommu_group *group;
854 unsigned long remote_vaddr;
855 struct vfio_dma *dma;
859 if (!iommu || !user_pfn || !phys_pfn)
862 /* Supported for v2 version only */
866 mutex_lock(&iommu->lock);
869 * Wait for all necessary vaddr's to be valid so they can be used in
870 * the main loop without dropping the lock, to avoid racing vs unmap.
873 if (iommu->vaddr_invalid_count) {
874 for (i = 0; i < npage; i++) {
875 iova = user_pfn[i] << PAGE_SHIFT;
876 ret = vfio_find_dma_valid(iommu, iova, PAGE_SIZE, &dma);
884 /* Fail if notifier list is empty */
885 if (!iommu->notifier.head) {
891 * If iommu capable domain exist in the container then all pages are
892 * already pinned and accounted. Accounting should be done if there is no
893 * iommu capable domain in the container.
895 do_accounting = !IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu);
897 for (i = 0; i < npage; i++) {
898 struct vfio_pfn *vpfn;
900 iova = user_pfn[i] << PAGE_SHIFT;
901 dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
907 if ((dma->prot & prot) != prot) {
912 vpfn = vfio_iova_get_vfio_pfn(dma, iova);
914 phys_pfn[i] = vpfn->pfn;
918 remote_vaddr = dma->vaddr + (iova - dma->iova);
919 ret = vfio_pin_page_external(dma, remote_vaddr, &phys_pfn[i],
924 ret = vfio_add_to_pfn_list(dma, iova, phys_pfn[i]);
926 if (put_pfn(phys_pfn[i], dma->prot) && do_accounting)
927 vfio_lock_acct(dma, -1, true);
931 if (iommu->dirty_page_tracking) {
932 unsigned long pgshift = __ffs(iommu->pgsize_bitmap);
935 * Bitmap populated with the smallest supported page
938 bitmap_set(dma->bitmap,
939 (iova - dma->iova) >> pgshift, 1);
944 group = vfio_iommu_find_iommu_group(iommu, iommu_group);
945 if (!group->pinned_page_dirty_scope) {
946 group->pinned_page_dirty_scope = true;
947 iommu->num_non_pinned_groups--;
954 for (j = 0; j < i; j++) {
957 iova = user_pfn[j] << PAGE_SHIFT;
958 dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
959 vfio_unpin_page_external(dma, iova, do_accounting);
963 mutex_unlock(&iommu->lock);
967 static int vfio_iommu_type1_unpin_pages(void *iommu_data,
968 unsigned long *user_pfn,
971 struct vfio_iommu *iommu = iommu_data;
975 if (!iommu || !user_pfn || npage <= 0)
978 /* Supported for v2 version only */
982 mutex_lock(&iommu->lock);
984 do_accounting = !IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu);
985 for (i = 0; i < npage; i++) {
986 struct vfio_dma *dma;
989 iova = user_pfn[i] << PAGE_SHIFT;
990 dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
994 vfio_unpin_page_external(dma, iova, do_accounting);
997 mutex_unlock(&iommu->lock);
998 return i > 0 ? i : -EINVAL;
1001 static long vfio_sync_unpin(struct vfio_dma *dma, struct vfio_domain *domain,
1002 struct list_head *regions,
1003 struct iommu_iotlb_gather *iotlb_gather)
1006 struct vfio_regions *entry, *next;
1008 iommu_iotlb_sync(domain->domain, iotlb_gather);
1010 list_for_each_entry_safe(entry, next, regions, list) {
1011 unlocked += vfio_unpin_pages_remote(dma,
1013 entry->phys >> PAGE_SHIFT,
1014 entry->len >> PAGE_SHIFT,
1016 list_del(&entry->list);
1026 * Generally, VFIO needs to unpin remote pages after each IOTLB flush.
1027 * Therefore, when using IOTLB flush sync interface, VFIO need to keep track
1028 * of these regions (currently using a list).
1030 * This value specifies maximum number of regions for each IOTLB flush sync.
1032 #define VFIO_IOMMU_TLB_SYNC_MAX 512
1034 static size_t unmap_unpin_fast(struct vfio_domain *domain,
1035 struct vfio_dma *dma, dma_addr_t *iova,
1036 size_t len, phys_addr_t phys, long *unlocked,
1037 struct list_head *unmapped_list,
1039 struct iommu_iotlb_gather *iotlb_gather)
1041 size_t unmapped = 0;
1042 struct vfio_regions *entry = kzalloc(sizeof(*entry), GFP_KERNEL);
1045 unmapped = iommu_unmap_fast(domain->domain, *iova, len,
1051 entry->iova = *iova;
1053 entry->len = unmapped;
1054 list_add_tail(&entry->list, unmapped_list);
1062 * Sync if the number of fast-unmap regions hits the limit
1063 * or in case of errors.
1065 if (*unmapped_cnt >= VFIO_IOMMU_TLB_SYNC_MAX || !unmapped) {
1066 *unlocked += vfio_sync_unpin(dma, domain, unmapped_list,
1074 static size_t unmap_unpin_slow(struct vfio_domain *domain,
1075 struct vfio_dma *dma, dma_addr_t *iova,
1076 size_t len, phys_addr_t phys,
1079 size_t unmapped = iommu_unmap(domain->domain, *iova, len);
1082 *unlocked += vfio_unpin_pages_remote(dma, *iova,
1084 unmapped >> PAGE_SHIFT,
1092 static long vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
1095 dma_addr_t iova = dma->iova, end = dma->iova + dma->size;
1096 struct vfio_domain *domain, *d;
1097 LIST_HEAD(unmapped_region_list);
1098 struct iommu_iotlb_gather iotlb_gather;
1099 int unmapped_region_cnt = 0;
1105 if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu))
1109 * We use the IOMMU to track the physical addresses, otherwise we'd
1110 * need a much more complicated tracking system. Unfortunately that
1111 * means we need to use one of the iommu domains to figure out the
1112 * pfns to unpin. The rest need to be unmapped in advance so we have
1113 * no iommu translations remaining when the pages are unpinned.
1115 domain = d = list_first_entry(&iommu->domain_list,
1116 struct vfio_domain, next);
1118 list_for_each_entry_continue(d, &iommu->domain_list, next) {
1119 iommu_unmap(d->domain, dma->iova, dma->size);
1123 iommu_iotlb_gather_init(&iotlb_gather);
1124 while (iova < end) {
1125 size_t unmapped, len;
1126 phys_addr_t phys, next;
1128 phys = iommu_iova_to_phys(domain->domain, iova);
1129 if (WARN_ON(!phys)) {
1135 * To optimize for fewer iommu_unmap() calls, each of which
1136 * may require hardware cache flushing, try to find the
1137 * largest contiguous physical memory chunk to unmap.
1139 for (len = PAGE_SIZE;
1140 !domain->fgsp && iova + len < end; len += PAGE_SIZE) {
1141 next = iommu_iova_to_phys(domain->domain, iova + len);
1142 if (next != phys + len)
1147 * First, try to use fast unmap/unpin. In case of failure,
1148 * switch to slow unmap/unpin path.
1150 unmapped = unmap_unpin_fast(domain, dma, &iova, len, phys,
1151 &unlocked, &unmapped_region_list,
1152 &unmapped_region_cnt,
1155 unmapped = unmap_unpin_slow(domain, dma, &iova, len,
1157 if (WARN_ON(!unmapped))
1162 dma->iommu_mapped = false;
1164 if (unmapped_region_cnt) {
1165 unlocked += vfio_sync_unpin(dma, domain, &unmapped_region_list,
1169 if (do_accounting) {
1170 vfio_lock_acct(dma, -unlocked, true);
1176 static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *dma)
1178 WARN_ON(!RB_EMPTY_ROOT(&dma->pfn_list));
1179 vfio_unmap_unpin(iommu, dma, true);
1180 vfio_unlink_dma(iommu, dma);
1181 put_task_struct(dma->task);
1182 vfio_dma_bitmap_free(dma);
1183 if (dma->vaddr_invalid) {
1184 iommu->vaddr_invalid_count--;
1185 wake_up_all(&iommu->vaddr_wait);
1191 static void vfio_update_pgsize_bitmap(struct vfio_iommu *iommu)
1193 struct vfio_domain *domain;
1195 iommu->pgsize_bitmap = ULONG_MAX;
1197 list_for_each_entry(domain, &iommu->domain_list, next)
1198 iommu->pgsize_bitmap &= domain->domain->pgsize_bitmap;
1201 * In case the IOMMU supports page sizes smaller than PAGE_SIZE
1202 * we pretend PAGE_SIZE is supported and hide sub-PAGE_SIZE sizes.
1203 * That way the user will be able to map/unmap buffers whose size/
1204 * start address is aligned with PAGE_SIZE. Pinning code uses that
1205 * granularity while iommu driver can use the sub-PAGE_SIZE size
1206 * to map the buffer.
1208 if (iommu->pgsize_bitmap & ~PAGE_MASK) {
1209 iommu->pgsize_bitmap &= PAGE_MASK;
1210 iommu->pgsize_bitmap |= PAGE_SIZE;
1214 static int update_user_bitmap(u64 __user *bitmap, struct vfio_iommu *iommu,
1215 struct vfio_dma *dma, dma_addr_t base_iova,
1218 unsigned long pgshift = __ffs(pgsize);
1219 unsigned long nbits = dma->size >> pgshift;
1220 unsigned long bit_offset = (dma->iova - base_iova) >> pgshift;
1221 unsigned long copy_offset = bit_offset / BITS_PER_LONG;
1222 unsigned long shift = bit_offset % BITS_PER_LONG;
1223 unsigned long leftover;
1226 * mark all pages dirty if any IOMMU capable device is not able
1227 * to report dirty pages and all pages are pinned and mapped.
1229 if (iommu->num_non_pinned_groups && dma->iommu_mapped)
1230 bitmap_set(dma->bitmap, 0, nbits);
1233 bitmap_shift_left(dma->bitmap, dma->bitmap, shift,
1236 if (copy_from_user(&leftover,
1237 (void __user *)(bitmap + copy_offset),
1241 bitmap_or(dma->bitmap, dma->bitmap, &leftover, shift);
1244 if (copy_to_user((void __user *)(bitmap + copy_offset), dma->bitmap,
1245 DIRTY_BITMAP_BYTES(nbits + shift)))
1251 static int vfio_iova_dirty_bitmap(u64 __user *bitmap, struct vfio_iommu *iommu,
1252 dma_addr_t iova, size_t size, size_t pgsize)
1254 struct vfio_dma *dma;
1256 unsigned long pgshift = __ffs(pgsize);
1260 * GET_BITMAP request must fully cover vfio_dma mappings. Multiple
1261 * vfio_dma mappings may be clubbed by specifying large ranges, but
1262 * there must not be any previous mappings bisected by the range.
1263 * An error will be returned if these conditions are not met.
1265 dma = vfio_find_dma(iommu, iova, 1);
1266 if (dma && dma->iova != iova)
1269 dma = vfio_find_dma(iommu, iova + size - 1, 0);
1270 if (dma && dma->iova + dma->size != iova + size)
1273 for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
1274 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
1276 if (dma->iova < iova)
1279 if (dma->iova > iova + size - 1)
1282 ret = update_user_bitmap(bitmap, iommu, dma, iova, pgsize);
1287 * Re-populate bitmap to include all pinned pages which are
1288 * considered as dirty but exclude pages which are unpinned and
1289 * pages which are marked dirty by vfio_dma_rw()
1291 bitmap_clear(dma->bitmap, 0, dma->size >> pgshift);
1292 vfio_dma_populate_bitmap(dma, pgsize);
1297 static int verify_bitmap_size(uint64_t npages, uint64_t bitmap_size)
1299 if (!npages || !bitmap_size || (bitmap_size > DIRTY_BITMAP_SIZE_MAX) ||
1300 (bitmap_size < DIRTY_BITMAP_BYTES(npages)))
1306 static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
1307 struct vfio_iommu_type1_dma_unmap *unmap,
1308 struct vfio_bitmap *bitmap)
1310 struct vfio_dma *dma, *dma_last = NULL;
1311 size_t unmapped = 0, pgsize;
1312 int ret = -EINVAL, retries = 0;
1313 unsigned long pgshift;
1314 dma_addr_t iova = unmap->iova;
1315 u64 size = unmap->size;
1316 bool unmap_all = unmap->flags & VFIO_DMA_UNMAP_FLAG_ALL;
1317 bool invalidate_vaddr = unmap->flags & VFIO_DMA_UNMAP_FLAG_VADDR;
1318 struct rb_node *n, *first_n;
1320 mutex_lock(&iommu->lock);
1322 pgshift = __ffs(iommu->pgsize_bitmap);
1323 pgsize = (size_t)1 << pgshift;
1325 if (iova & (pgsize - 1))
1332 } else if (!size || size & (pgsize - 1) ||
1333 iova + size - 1 < iova || size > SIZE_MAX) {
1337 /* When dirty tracking is enabled, allow only min supported pgsize */
1338 if ((unmap->flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) &&
1339 (!iommu->dirty_page_tracking || (bitmap->pgsize != pgsize))) {
1343 WARN_ON((pgsize - 1) & PAGE_MASK);
1346 * vfio-iommu-type1 (v1) - User mappings were coalesced together to
1347 * avoid tracking individual mappings. This means that the granularity
1348 * of the original mapping was lost and the user was allowed to attempt
1349 * to unmap any range. Depending on the contiguousness of physical
1350 * memory and page sizes supported by the IOMMU, arbitrary unmaps may
1351 * or may not have worked. We only guaranteed unmap granularity
1352 * matching the original mapping; even though it was untracked here,
1353 * the original mappings are reflected in IOMMU mappings. This
1354 * resulted in a couple unusual behaviors. First, if a range is not
1355 * able to be unmapped, ex. a set of 4k pages that was mapped as a
1356 * 2M hugepage into the IOMMU, the unmap ioctl returns success but with
1357 * a zero sized unmap. Also, if an unmap request overlaps the first
1358 * address of a hugepage, the IOMMU will unmap the entire hugepage.
1359 * This also returns success and the returned unmap size reflects the
1360 * actual size unmapped.
1362 * We attempt to maintain compatibility with this "v1" interface, but
1363 * we take control out of the hands of the IOMMU. Therefore, an unmap
1364 * request offset from the beginning of the original mapping will
1365 * return success with zero sized unmap. And an unmap request covering
1366 * the first iova of mapping will unmap the entire range.
1368 * The v2 version of this interface intends to be more deterministic.
1369 * Unmap requests must fully cover previous mappings. Multiple
1370 * mappings may still be unmaped by specifying large ranges, but there
1371 * must not be any previous mappings bisected by the range. An error
1372 * will be returned if these conditions are not met. The v2 interface
1373 * will only return success and a size of zero if there were no
1374 * mappings within the range.
1376 if (iommu->v2 && !unmap_all) {
1377 dma = vfio_find_dma(iommu, iova, 1);
1378 if (dma && dma->iova != iova)
1381 dma = vfio_find_dma(iommu, iova + size - 1, 0);
1382 if (dma && dma->iova + dma->size != iova + size)
1387 n = first_n = vfio_find_dma_first_node(iommu, iova, size);
1390 dma = rb_entry(n, struct vfio_dma, node);
1391 if (dma->iova >= iova + size)
1394 if (!iommu->v2 && iova > dma->iova)
1397 * Task with same address space who mapped this iova range is
1398 * allowed to unmap the iova range.
1400 if (dma->task->mm != current->mm)
1403 if (invalidate_vaddr) {
1404 if (dma->vaddr_invalid) {
1405 struct rb_node *last_n = n;
1407 for (n = first_n; n != last_n; n = rb_next(n)) {
1409 struct vfio_dma, node);
1410 dma->vaddr_invalid = false;
1411 iommu->vaddr_invalid_count--;
1417 dma->vaddr_invalid = true;
1418 iommu->vaddr_invalid_count++;
1419 unmapped += dma->size;
1424 if (!RB_EMPTY_ROOT(&dma->pfn_list)) {
1425 struct vfio_iommu_type1_dma_unmap nb_unmap;
1427 if (dma_last == dma) {
1428 BUG_ON(++retries > 10);
1434 nb_unmap.iova = dma->iova;
1435 nb_unmap.size = dma->size;
1438 * Notify anyone (mdev vendor drivers) to invalidate and
1439 * unmap iovas within the range we're about to unmap.
1440 * Vendor drivers MUST unpin pages in response to an
1443 mutex_unlock(&iommu->lock);
1444 blocking_notifier_call_chain(&iommu->notifier,
1445 VFIO_IOMMU_NOTIFY_DMA_UNMAP,
1447 mutex_lock(&iommu->lock);
1451 if (unmap->flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) {
1452 ret = update_user_bitmap(bitmap->data, iommu, dma,
1458 unmapped += dma->size;
1460 vfio_remove_dma(iommu, dma);
1464 mutex_unlock(&iommu->lock);
1466 /* Report how much was unmapped */
1467 unmap->size = unmapped;
1472 static int vfio_iommu_map(struct vfio_iommu *iommu, dma_addr_t iova,
1473 unsigned long pfn, long npage, int prot)
1475 struct vfio_domain *d;
1478 list_for_each_entry(d, &iommu->domain_list, next) {
1479 ret = iommu_map(d->domain, iova, (phys_addr_t)pfn << PAGE_SHIFT,
1480 npage << PAGE_SHIFT, prot | d->prot);
1490 list_for_each_entry_continue_reverse(d, &iommu->domain_list, next) {
1491 iommu_unmap(d->domain, iova, npage << PAGE_SHIFT);
1498 static int vfio_pin_map_dma(struct vfio_iommu *iommu, struct vfio_dma *dma,
1501 dma_addr_t iova = dma->iova;
1502 unsigned long vaddr = dma->vaddr;
1503 struct vfio_batch batch;
1504 size_t size = map_size;
1506 unsigned long pfn, limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
1509 vfio_batch_init(&batch);
1512 /* Pin a contiguous chunk of memory */
1513 npage = vfio_pin_pages_remote(dma, vaddr + dma->size,
1514 size >> PAGE_SHIFT, &pfn, limit,
1523 ret = vfio_iommu_map(iommu, iova + dma->size, pfn, npage,
1526 vfio_unpin_pages_remote(dma, iova + dma->size, pfn,
1528 vfio_batch_unpin(&batch, dma);
1532 size -= npage << PAGE_SHIFT;
1533 dma->size += npage << PAGE_SHIFT;
1536 vfio_batch_fini(&batch);
1537 dma->iommu_mapped = true;
1540 vfio_remove_dma(iommu, dma);
1546 * Check dma map request is within a valid iova range
1548 static bool vfio_iommu_iova_dma_valid(struct vfio_iommu *iommu,
1549 dma_addr_t start, dma_addr_t end)
1551 struct list_head *iova = &iommu->iova_list;
1552 struct vfio_iova *node;
1554 list_for_each_entry(node, iova, list) {
1555 if (start >= node->start && end <= node->end)
1560 * Check for list_empty() as well since a container with
1561 * a single mdev device will have an empty list.
1563 return list_empty(iova);
1566 static int vfio_dma_do_map(struct vfio_iommu *iommu,
1567 struct vfio_iommu_type1_dma_map *map)
1569 bool set_vaddr = map->flags & VFIO_DMA_MAP_FLAG_VADDR;
1570 dma_addr_t iova = map->iova;
1571 unsigned long vaddr = map->vaddr;
1572 size_t size = map->size;
1573 int ret = 0, prot = 0;
1575 struct vfio_dma *dma;
1577 /* Verify that none of our __u64 fields overflow */
1578 if (map->size != size || map->vaddr != vaddr || map->iova != iova)
1581 /* READ/WRITE from device perspective */
1582 if (map->flags & VFIO_DMA_MAP_FLAG_WRITE)
1583 prot |= IOMMU_WRITE;
1584 if (map->flags & VFIO_DMA_MAP_FLAG_READ)
1587 if ((prot && set_vaddr) || (!prot && !set_vaddr))
1590 mutex_lock(&iommu->lock);
1592 pgsize = (size_t)1 << __ffs(iommu->pgsize_bitmap);
1594 WARN_ON((pgsize - 1) & PAGE_MASK);
1596 if (!size || (size | iova | vaddr) & (pgsize - 1)) {
1601 /* Don't allow IOVA or virtual address wrap */
1602 if (iova + size - 1 < iova || vaddr + size - 1 < vaddr) {
1607 dma = vfio_find_dma(iommu, iova, size);
1611 } else if (!dma->vaddr_invalid || dma->iova != iova ||
1612 dma->size != size) {
1616 dma->vaddr_invalid = false;
1617 iommu->vaddr_invalid_count--;
1618 wake_up_all(&iommu->vaddr_wait);
1626 if (!iommu->dma_avail) {
1631 if (!vfio_iommu_iova_dma_valid(iommu, iova, iova + size - 1)) {
1636 dma = kzalloc(sizeof(*dma), GFP_KERNEL);
1648 * We need to be able to both add to a task's locked memory and test
1649 * against the locked memory limit and we need to be able to do both
1650 * outside of this call path as pinning can be asynchronous via the
1651 * external interfaces for mdev devices. RLIMIT_MEMLOCK requires a
1652 * task_struct and VM locked pages requires an mm_struct, however
1653 * holding an indefinite mm reference is not recommended, therefore we
1654 * only hold a reference to a task. We could hold a reference to
1655 * current, however QEMU uses this call path through vCPU threads,
1656 * which can be killed resulting in a NULL mm and failure in the unmap
1657 * path when called via a different thread. Avoid this problem by
1658 * using the group_leader as threads within the same group require
1659 * both CLONE_THREAD and CLONE_VM and will therefore use the same
1662 * Previously we also used the task for testing CAP_IPC_LOCK at the
1663 * time of pinning and accounting, however has_capability() makes use
1664 * of real_cred, a copy-on-write field, so we can't guarantee that it
1665 * matches group_leader, or in fact that it might not change by the
1666 * time it's evaluated. If a process were to call MAP_DMA with
1667 * CAP_IPC_LOCK but later drop it, it doesn't make sense that they
1668 * possibly see different results for an iommu_mapped vfio_dma vs
1669 * externally mapped. Therefore track CAP_IPC_LOCK in vfio_dma at the
1670 * time of calling MAP_DMA.
1672 get_task_struct(current->group_leader);
1673 dma->task = current->group_leader;
1674 dma->lock_cap = capable(CAP_IPC_LOCK);
1676 dma->pfn_list = RB_ROOT;
1678 /* Insert zero-sized and grow as we map chunks of it */
1679 vfio_link_dma(iommu, dma);
1681 /* Don't pin and map if container doesn't contain IOMMU capable domain*/
1682 if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu))
1685 ret = vfio_pin_map_dma(iommu, dma, size);
1687 if (!ret && iommu->dirty_page_tracking) {
1688 ret = vfio_dma_bitmap_alloc(dma, pgsize);
1690 vfio_remove_dma(iommu, dma);
1694 mutex_unlock(&iommu->lock);
1698 static int vfio_bus_type(struct device *dev, void *data)
1700 struct bus_type **bus = data;
1702 if (*bus && *bus != dev->bus)
1710 static int vfio_iommu_replay(struct vfio_iommu *iommu,
1711 struct vfio_domain *domain)
1713 struct vfio_batch batch;
1714 struct vfio_domain *d = NULL;
1716 unsigned long limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
1719 ret = vfio_wait_all_valid(iommu);
1723 /* Arbitrarily pick the first domain in the list for lookups */
1724 if (!list_empty(&iommu->domain_list))
1725 d = list_first_entry(&iommu->domain_list,
1726 struct vfio_domain, next);
1728 vfio_batch_init(&batch);
1730 n = rb_first(&iommu->dma_list);
1732 for (; n; n = rb_next(n)) {
1733 struct vfio_dma *dma;
1736 dma = rb_entry(n, struct vfio_dma, node);
1739 while (iova < dma->iova + dma->size) {
1743 if (dma->iommu_mapped) {
1747 if (WARN_ON(!d)) { /* mapped w/o a domain?! */
1752 phys = iommu_iova_to_phys(d->domain, iova);
1754 if (WARN_ON(!phys)) {
1762 while (i < dma->iova + dma->size &&
1763 p == iommu_iova_to_phys(d->domain, i)) {
1770 unsigned long vaddr = dma->vaddr +
1772 size_t n = dma->iova + dma->size - iova;
1775 npage = vfio_pin_pages_remote(dma, vaddr,
1785 phys = pfn << PAGE_SHIFT;
1786 size = npage << PAGE_SHIFT;
1789 ret = iommu_map(domain->domain, iova, phys,
1790 size, dma->prot | domain->prot);
1792 if (!dma->iommu_mapped) {
1793 vfio_unpin_pages_remote(dma, iova,
1797 vfio_batch_unpin(&batch, dma);
1806 /* All dmas are now mapped, defer to second tree walk for unwind */
1807 for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
1808 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
1810 dma->iommu_mapped = true;
1813 vfio_batch_fini(&batch);
1817 for (; n; n = rb_prev(n)) {
1818 struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
1821 if (dma->iommu_mapped) {
1822 iommu_unmap(domain->domain, dma->iova, dma->size);
1827 while (iova < dma->iova + dma->size) {
1828 phys_addr_t phys, p;
1832 phys = iommu_iova_to_phys(domain->domain, iova);
1841 while (i < dma->iova + dma->size &&
1842 p == iommu_iova_to_phys(domain->domain, i)) {
1848 iommu_unmap(domain->domain, iova, size);
1849 vfio_unpin_pages_remote(dma, iova, phys >> PAGE_SHIFT,
1850 size >> PAGE_SHIFT, true);
1854 vfio_batch_fini(&batch);
1859 * We change our unmap behavior slightly depending on whether the IOMMU
1860 * supports fine-grained superpages. IOMMUs like AMD-Vi will use a superpage
1861 * for practically any contiguous power-of-two mapping we give it. This means
1862 * we don't need to look for contiguous chunks ourselves to make unmapping
1863 * more efficient. On IOMMUs with coarse-grained super pages, like Intel VT-d
1864 * with discrete 2M/1G/512G/1T superpages, identifying contiguous chunks
1865 * significantly boosts non-hugetlbfs mappings and doesn't seem to hurt when
1866 * hugetlbfs is in use.
1868 static void vfio_test_domain_fgsp(struct vfio_domain *domain)
1871 int ret, order = get_order(PAGE_SIZE * 2);
1873 pages = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
1877 ret = iommu_map(domain->domain, 0, page_to_phys(pages), PAGE_SIZE * 2,
1878 IOMMU_READ | IOMMU_WRITE | domain->prot);
1880 size_t unmapped = iommu_unmap(domain->domain, 0, PAGE_SIZE);
1882 if (unmapped == PAGE_SIZE)
1883 iommu_unmap(domain->domain, PAGE_SIZE, PAGE_SIZE);
1885 domain->fgsp = true;
1888 __free_pages(pages, order);
1891 static struct vfio_iommu_group *find_iommu_group(struct vfio_domain *domain,
1892 struct iommu_group *iommu_group)
1894 struct vfio_iommu_group *g;
1896 list_for_each_entry(g, &domain->group_list, next) {
1897 if (g->iommu_group == iommu_group)
1904 static struct vfio_iommu_group*
1905 vfio_iommu_find_iommu_group(struct vfio_iommu *iommu,
1906 struct iommu_group *iommu_group)
1908 struct vfio_domain *domain;
1909 struct vfio_iommu_group *group = NULL;
1911 list_for_each_entry(domain, &iommu->domain_list, next) {
1912 group = find_iommu_group(domain, iommu_group);
1917 if (iommu->external_domain)
1918 group = find_iommu_group(iommu->external_domain, iommu_group);
1923 static bool vfio_iommu_has_sw_msi(struct list_head *group_resv_regions,
1926 struct iommu_resv_region *region;
1929 list_for_each_entry(region, group_resv_regions, list) {
1931 * The presence of any 'real' MSI regions should take
1932 * precedence over the software-managed one if the
1933 * IOMMU driver happens to advertise both types.
1935 if (region->type == IOMMU_RESV_MSI) {
1940 if (region->type == IOMMU_RESV_SW_MSI) {
1941 *base = region->start;
1949 static int vfio_mdev_attach_domain(struct device *dev, void *data)
1951 struct mdev_device *mdev = to_mdev_device(dev);
1952 struct iommu_domain *domain = data;
1953 struct device *iommu_device;
1955 iommu_device = mdev_get_iommu_device(mdev);
1957 if (iommu_dev_feature_enabled(iommu_device, IOMMU_DEV_FEAT_AUX))
1958 return iommu_aux_attach_device(domain, iommu_device);
1960 return iommu_attach_device(domain, iommu_device);
1966 static int vfio_mdev_detach_domain(struct device *dev, void *data)
1968 struct mdev_device *mdev = to_mdev_device(dev);
1969 struct iommu_domain *domain = data;
1970 struct device *iommu_device;
1972 iommu_device = mdev_get_iommu_device(mdev);
1974 if (iommu_dev_feature_enabled(iommu_device, IOMMU_DEV_FEAT_AUX))
1975 iommu_aux_detach_device(domain, iommu_device);
1977 iommu_detach_device(domain, iommu_device);
1983 static int vfio_iommu_attach_group(struct vfio_domain *domain,
1984 struct vfio_iommu_group *group)
1986 if (group->mdev_group)
1987 return iommu_group_for_each_dev(group->iommu_group,
1989 vfio_mdev_attach_domain);
1991 return iommu_attach_group(domain->domain, group->iommu_group);
1994 static void vfio_iommu_detach_group(struct vfio_domain *domain,
1995 struct vfio_iommu_group *group)
1997 if (group->mdev_group)
1998 iommu_group_for_each_dev(group->iommu_group, domain->domain,
1999 vfio_mdev_detach_domain);
2001 iommu_detach_group(domain->domain, group->iommu_group);
2004 static bool vfio_bus_is_mdev(struct bus_type *bus)
2006 struct bus_type *mdev_bus;
2009 mdev_bus = symbol_get(mdev_bus_type);
2011 ret = (bus == mdev_bus);
2012 symbol_put(mdev_bus_type);
2018 static int vfio_mdev_iommu_device(struct device *dev, void *data)
2020 struct mdev_device *mdev = to_mdev_device(dev);
2021 struct device **old = data, *new;
2023 new = mdev_get_iommu_device(mdev);
2024 if (!new || (*old && *old != new))
2033 * This is a helper function to insert an address range to iova list.
2034 * The list is initially created with a single entry corresponding to
2035 * the IOMMU domain geometry to which the device group is attached.
2036 * The list aperture gets modified when a new domain is added to the
2037 * container if the new aperture doesn't conflict with the current one
2038 * or with any existing dma mappings. The list is also modified to
2039 * exclude any reserved regions associated with the device group.
2041 static int vfio_iommu_iova_insert(struct list_head *head,
2042 dma_addr_t start, dma_addr_t end)
2044 struct vfio_iova *region;
2046 region = kmalloc(sizeof(*region), GFP_KERNEL);
2050 INIT_LIST_HEAD(®ion->list);
2051 region->start = start;
2054 list_add_tail(®ion->list, head);
2059 * Check the new iommu aperture conflicts with existing aper or with any
2060 * existing dma mappings.
2062 static bool vfio_iommu_aper_conflict(struct vfio_iommu *iommu,
2063 dma_addr_t start, dma_addr_t end)
2065 struct vfio_iova *first, *last;
2066 struct list_head *iova = &iommu->iova_list;
2068 if (list_empty(iova))
2071 /* Disjoint sets, return conflict */
2072 first = list_first_entry(iova, struct vfio_iova, list);
2073 last = list_last_entry(iova, struct vfio_iova, list);
2074 if (start > last->end || end < first->start)
2077 /* Check for any existing dma mappings below the new start */
2078 if (start > first->start) {
2079 if (vfio_find_dma(iommu, first->start, start - first->start))
2083 /* Check for any existing dma mappings beyond the new end */
2084 if (end < last->end) {
2085 if (vfio_find_dma(iommu, end + 1, last->end - end))
2093 * Resize iommu iova aperture window. This is called only if the new
2094 * aperture has no conflict with existing aperture and dma mappings.
2096 static int vfio_iommu_aper_resize(struct list_head *iova,
2097 dma_addr_t start, dma_addr_t end)
2099 struct vfio_iova *node, *next;
2101 if (list_empty(iova))
2102 return vfio_iommu_iova_insert(iova, start, end);
2104 /* Adjust iova list start */
2105 list_for_each_entry_safe(node, next, iova, list) {
2106 if (start < node->start)
2108 if (start >= node->start && start < node->end) {
2109 node->start = start;
2112 /* Delete nodes before new start */
2113 list_del(&node->list);
2117 /* Adjust iova list end */
2118 list_for_each_entry_safe(node, next, iova, list) {
2119 if (end > node->end)
2121 if (end > node->start && end <= node->end) {
2125 /* Delete nodes after new end */
2126 list_del(&node->list);
2134 * Check reserved region conflicts with existing dma mappings
2136 static bool vfio_iommu_resv_conflict(struct vfio_iommu *iommu,
2137 struct list_head *resv_regions)
2139 struct iommu_resv_region *region;
2141 /* Check for conflict with existing dma mappings */
2142 list_for_each_entry(region, resv_regions, list) {
2143 if (region->type == IOMMU_RESV_DIRECT_RELAXABLE)
2146 if (vfio_find_dma(iommu, region->start, region->length))
2154 * Check iova region overlap with reserved regions and
2155 * exclude them from the iommu iova range
2157 static int vfio_iommu_resv_exclude(struct list_head *iova,
2158 struct list_head *resv_regions)
2160 struct iommu_resv_region *resv;
2161 struct vfio_iova *n, *next;
2163 list_for_each_entry(resv, resv_regions, list) {
2164 phys_addr_t start, end;
2166 if (resv->type == IOMMU_RESV_DIRECT_RELAXABLE)
2169 start = resv->start;
2170 end = resv->start + resv->length - 1;
2172 list_for_each_entry_safe(n, next, iova, list) {
2176 if (start > n->end || end < n->start)
2179 * Insert a new node if current node overlaps with the
2180 * reserve region to exclude that from valid iova range.
2181 * Note that, new node is inserted before the current
2182 * node and finally the current node is deleted keeping
2183 * the list updated and sorted.
2185 if (start > n->start)
2186 ret = vfio_iommu_iova_insert(&n->list, n->start,
2188 if (!ret && end < n->end)
2189 ret = vfio_iommu_iova_insert(&n->list, end + 1,
2199 if (list_empty(iova))
2205 static void vfio_iommu_resv_free(struct list_head *resv_regions)
2207 struct iommu_resv_region *n, *next;
2209 list_for_each_entry_safe(n, next, resv_regions, list) {
2215 static void vfio_iommu_iova_free(struct list_head *iova)
2217 struct vfio_iova *n, *next;
2219 list_for_each_entry_safe(n, next, iova, list) {
2225 static int vfio_iommu_iova_get_copy(struct vfio_iommu *iommu,
2226 struct list_head *iova_copy)
2228 struct list_head *iova = &iommu->iova_list;
2229 struct vfio_iova *n;
2232 list_for_each_entry(n, iova, list) {
2233 ret = vfio_iommu_iova_insert(iova_copy, n->start, n->end);
2241 vfio_iommu_iova_free(iova_copy);
2245 static void vfio_iommu_iova_insert_copy(struct vfio_iommu *iommu,
2246 struct list_head *iova_copy)
2248 struct list_head *iova = &iommu->iova_list;
2250 vfio_iommu_iova_free(iova);
2252 list_splice_tail(iova_copy, iova);
2255 static int vfio_iommu_type1_attach_group(void *iommu_data,
2256 struct iommu_group *iommu_group)
2258 struct vfio_iommu *iommu = iommu_data;
2259 struct vfio_iommu_group *group;
2260 struct vfio_domain *domain, *d;
2261 struct bus_type *bus = NULL;
2263 bool resv_msi, msi_remap;
2264 phys_addr_t resv_msi_base = 0;
2265 struct iommu_domain_geometry *geo;
2266 LIST_HEAD(iova_copy);
2267 LIST_HEAD(group_resv_regions);
2269 mutex_lock(&iommu->lock);
2271 /* Check for duplicates */
2272 if (vfio_iommu_find_iommu_group(iommu, iommu_group)) {
2273 mutex_unlock(&iommu->lock);
2277 group = kzalloc(sizeof(*group), GFP_KERNEL);
2278 domain = kzalloc(sizeof(*domain), GFP_KERNEL);
2279 if (!group || !domain) {
2284 group->iommu_group = iommu_group;
2286 /* Determine bus_type in order to allocate a domain */
2287 ret = iommu_group_for_each_dev(iommu_group, &bus, vfio_bus_type);
2291 if (vfio_bus_is_mdev(bus)) {
2292 struct device *iommu_device = NULL;
2294 group->mdev_group = true;
2296 /* Determine the isolation type */
2297 ret = iommu_group_for_each_dev(iommu_group, &iommu_device,
2298 vfio_mdev_iommu_device);
2299 if (ret || !iommu_device) {
2300 if (!iommu->external_domain) {
2301 INIT_LIST_HEAD(&domain->group_list);
2302 iommu->external_domain = domain;
2303 vfio_update_pgsize_bitmap(iommu);
2308 list_add(&group->next,
2309 &iommu->external_domain->group_list);
2311 * Non-iommu backed group cannot dirty memory directly,
2312 * it can only use interfaces that provide dirty
2314 * The iommu scope can only be promoted with the
2315 * addition of a dirty tracking group.
2317 group->pinned_page_dirty_scope = true;
2318 mutex_unlock(&iommu->lock);
2323 bus = iommu_device->bus;
2326 domain->domain = iommu_domain_alloc(bus);
2327 if (!domain->domain) {
2332 if (iommu->nesting) {
2333 ret = iommu_enable_nesting(domain->domain);
2338 ret = vfio_iommu_attach_group(domain, group);
2342 /* Get aperture info */
2343 geo = &domain->domain->geometry;
2344 if (vfio_iommu_aper_conflict(iommu, geo->aperture_start,
2345 geo->aperture_end)) {
2350 ret = iommu_get_group_resv_regions(iommu_group, &group_resv_regions);
2354 if (vfio_iommu_resv_conflict(iommu, &group_resv_regions)) {
2360 * We don't want to work on the original iova list as the list
2361 * gets modified and in case of failure we have to retain the
2362 * original list. Get a copy here.
2364 ret = vfio_iommu_iova_get_copy(iommu, &iova_copy);
2368 ret = vfio_iommu_aper_resize(&iova_copy, geo->aperture_start,
2373 ret = vfio_iommu_resv_exclude(&iova_copy, &group_resv_regions);
2377 resv_msi = vfio_iommu_has_sw_msi(&group_resv_regions, &resv_msi_base);
2379 INIT_LIST_HEAD(&domain->group_list);
2380 list_add(&group->next, &domain->group_list);
2382 msi_remap = irq_domain_check_msi_remap() ||
2383 iommu_capable(bus, IOMMU_CAP_INTR_REMAP);
2385 if (!allow_unsafe_interrupts && !msi_remap) {
2386 pr_warn("%s: No interrupt remapping support. Use the module param \"allow_unsafe_interrupts\" to enable VFIO IOMMU support on this platform\n",
2392 if (iommu_capable(bus, IOMMU_CAP_CACHE_COHERENCY))
2393 domain->prot |= IOMMU_CACHE;
2396 * Try to match an existing compatible domain. We don't want to
2397 * preclude an IOMMU driver supporting multiple bus_types and being
2398 * able to include different bus_types in the same IOMMU domain, so
2399 * we test whether the domains use the same iommu_ops rather than
2400 * testing if they're on the same bus_type.
2402 list_for_each_entry(d, &iommu->domain_list, next) {
2403 if (d->domain->ops == domain->domain->ops &&
2404 d->prot == domain->prot) {
2405 vfio_iommu_detach_group(domain, group);
2406 if (!vfio_iommu_attach_group(d, group)) {
2407 list_add(&group->next, &d->group_list);
2408 iommu_domain_free(domain->domain);
2413 ret = vfio_iommu_attach_group(domain, group);
2419 vfio_test_domain_fgsp(domain);
2421 /* replay mappings on new domains */
2422 ret = vfio_iommu_replay(iommu, domain);
2427 ret = iommu_get_msi_cookie(domain->domain, resv_msi_base);
2428 if (ret && ret != -ENODEV)
2432 list_add(&domain->next, &iommu->domain_list);
2433 vfio_update_pgsize_bitmap(iommu);
2435 /* Delete the old one and insert new iova list */
2436 vfio_iommu_iova_insert_copy(iommu, &iova_copy);
2439 * An iommu backed group can dirty memory directly and therefore
2440 * demotes the iommu scope until it declares itself dirty tracking
2441 * capable via the page pinning interface.
2443 iommu->num_non_pinned_groups++;
2444 mutex_unlock(&iommu->lock);
2445 vfio_iommu_resv_free(&group_resv_regions);
2450 vfio_iommu_detach_group(domain, group);
2452 iommu_domain_free(domain->domain);
2453 vfio_iommu_iova_free(&iova_copy);
2454 vfio_iommu_resv_free(&group_resv_regions);
2458 mutex_unlock(&iommu->lock);
2462 static void vfio_iommu_unmap_unpin_all(struct vfio_iommu *iommu)
2464 struct rb_node *node;
2466 while ((node = rb_first(&iommu->dma_list)))
2467 vfio_remove_dma(iommu, rb_entry(node, struct vfio_dma, node));
2470 static void vfio_iommu_unmap_unpin_reaccount(struct vfio_iommu *iommu)
2472 struct rb_node *n, *p;
2474 n = rb_first(&iommu->dma_list);
2475 for (; n; n = rb_next(n)) {
2476 struct vfio_dma *dma;
2477 long locked = 0, unlocked = 0;
2479 dma = rb_entry(n, struct vfio_dma, node);
2480 unlocked += vfio_unmap_unpin(iommu, dma, false);
2481 p = rb_first(&dma->pfn_list);
2482 for (; p; p = rb_next(p)) {
2483 struct vfio_pfn *vpfn = rb_entry(p, struct vfio_pfn,
2486 if (!is_invalid_reserved_pfn(vpfn->pfn))
2489 vfio_lock_acct(dma, locked - unlocked, true);
2494 * Called when a domain is removed in detach. It is possible that
2495 * the removed domain decided the iova aperture window. Modify the
2496 * iova aperture with the smallest window among existing domains.
2498 static void vfio_iommu_aper_expand(struct vfio_iommu *iommu,
2499 struct list_head *iova_copy)
2501 struct vfio_domain *domain;
2502 struct vfio_iova *node;
2503 dma_addr_t start = 0;
2504 dma_addr_t end = (dma_addr_t)~0;
2506 if (list_empty(iova_copy))
2509 list_for_each_entry(domain, &iommu->domain_list, next) {
2510 struct iommu_domain_geometry *geo = &domain->domain->geometry;
2512 if (geo->aperture_start > start)
2513 start = geo->aperture_start;
2514 if (geo->aperture_end < end)
2515 end = geo->aperture_end;
2518 /* Modify aperture limits. The new aper is either same or bigger */
2519 node = list_first_entry(iova_copy, struct vfio_iova, list);
2520 node->start = start;
2521 node = list_last_entry(iova_copy, struct vfio_iova, list);
2526 * Called when a group is detached. The reserved regions for that
2527 * group can be part of valid iova now. But since reserved regions
2528 * may be duplicated among groups, populate the iova valid regions
2531 static int vfio_iommu_resv_refresh(struct vfio_iommu *iommu,
2532 struct list_head *iova_copy)
2534 struct vfio_domain *d;
2535 struct vfio_iommu_group *g;
2536 struct vfio_iova *node;
2537 dma_addr_t start, end;
2538 LIST_HEAD(resv_regions);
2541 if (list_empty(iova_copy))
2544 list_for_each_entry(d, &iommu->domain_list, next) {
2545 list_for_each_entry(g, &d->group_list, next) {
2546 ret = iommu_get_group_resv_regions(g->iommu_group,
2553 node = list_first_entry(iova_copy, struct vfio_iova, list);
2554 start = node->start;
2555 node = list_last_entry(iova_copy, struct vfio_iova, list);
2558 /* purge the iova list and create new one */
2559 vfio_iommu_iova_free(iova_copy);
2561 ret = vfio_iommu_aper_resize(iova_copy, start, end);
2565 /* Exclude current reserved regions from iova ranges */
2566 ret = vfio_iommu_resv_exclude(iova_copy, &resv_regions);
2568 vfio_iommu_resv_free(&resv_regions);
2572 static void vfio_iommu_type1_detach_group(void *iommu_data,
2573 struct iommu_group *iommu_group)
2575 struct vfio_iommu *iommu = iommu_data;
2576 struct vfio_domain *domain;
2577 struct vfio_iommu_group *group;
2578 bool update_dirty_scope = false;
2579 LIST_HEAD(iova_copy);
2581 mutex_lock(&iommu->lock);
2583 if (iommu->external_domain) {
2584 group = find_iommu_group(iommu->external_domain, iommu_group);
2586 update_dirty_scope = !group->pinned_page_dirty_scope;
2587 list_del(&group->next);
2590 if (list_empty(&iommu->external_domain->group_list)) {
2591 if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu)) {
2592 WARN_ON(iommu->notifier.head);
2593 vfio_iommu_unmap_unpin_all(iommu);
2596 kfree(iommu->external_domain);
2597 iommu->external_domain = NULL;
2599 goto detach_group_done;
2604 * Get a copy of iova list. This will be used to update
2605 * and to replace the current one later. Please note that
2606 * we will leave the original list as it is if update fails.
2608 vfio_iommu_iova_get_copy(iommu, &iova_copy);
2610 list_for_each_entry(domain, &iommu->domain_list, next) {
2611 group = find_iommu_group(domain, iommu_group);
2615 vfio_iommu_detach_group(domain, group);
2616 update_dirty_scope = !group->pinned_page_dirty_scope;
2617 list_del(&group->next);
2620 * Group ownership provides privilege, if the group list is
2621 * empty, the domain goes away. If it's the last domain with
2622 * iommu and external domain doesn't exist, then all the
2623 * mappings go away too. If it's the last domain with iommu and
2624 * external domain exist, update accounting
2626 if (list_empty(&domain->group_list)) {
2627 if (list_is_singular(&iommu->domain_list)) {
2628 if (!iommu->external_domain) {
2629 WARN_ON(iommu->notifier.head);
2630 vfio_iommu_unmap_unpin_all(iommu);
2632 vfio_iommu_unmap_unpin_reaccount(iommu);
2635 iommu_domain_free(domain->domain);
2636 list_del(&domain->next);
2638 vfio_iommu_aper_expand(iommu, &iova_copy);
2639 vfio_update_pgsize_bitmap(iommu);
2644 if (!vfio_iommu_resv_refresh(iommu, &iova_copy))
2645 vfio_iommu_iova_insert_copy(iommu, &iova_copy);
2647 vfio_iommu_iova_free(&iova_copy);
2651 * Removal of a group without dirty tracking may allow the iommu scope
2654 if (update_dirty_scope) {
2655 iommu->num_non_pinned_groups--;
2656 if (iommu->dirty_page_tracking)
2657 vfio_iommu_populate_bitmap_full(iommu);
2659 mutex_unlock(&iommu->lock);
2662 static void *vfio_iommu_type1_open(unsigned long arg)
2664 struct vfio_iommu *iommu;
2666 iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
2668 return ERR_PTR(-ENOMEM);
2671 case VFIO_TYPE1_IOMMU:
2673 case VFIO_TYPE1_NESTING_IOMMU:
2674 iommu->nesting = true;
2676 case VFIO_TYPE1v2_IOMMU:
2681 return ERR_PTR(-EINVAL);
2684 INIT_LIST_HEAD(&iommu->domain_list);
2685 INIT_LIST_HEAD(&iommu->iova_list);
2686 iommu->dma_list = RB_ROOT;
2687 iommu->dma_avail = dma_entry_limit;
2688 iommu->container_open = true;
2689 mutex_init(&iommu->lock);
2690 BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier);
2691 init_waitqueue_head(&iommu->vaddr_wait);
2696 static void vfio_release_domain(struct vfio_domain *domain, bool external)
2698 struct vfio_iommu_group *group, *group_tmp;
2700 list_for_each_entry_safe(group, group_tmp,
2701 &domain->group_list, next) {
2703 vfio_iommu_detach_group(domain, group);
2704 list_del(&group->next);
2709 iommu_domain_free(domain->domain);
2712 static void vfio_iommu_type1_release(void *iommu_data)
2714 struct vfio_iommu *iommu = iommu_data;
2715 struct vfio_domain *domain, *domain_tmp;
2717 if (iommu->external_domain) {
2718 vfio_release_domain(iommu->external_domain, true);
2719 kfree(iommu->external_domain);
2722 vfio_iommu_unmap_unpin_all(iommu);
2724 list_for_each_entry_safe(domain, domain_tmp,
2725 &iommu->domain_list, next) {
2726 vfio_release_domain(domain, false);
2727 list_del(&domain->next);
2731 vfio_iommu_iova_free(&iommu->iova_list);
2736 static int vfio_domains_have_iommu_cache(struct vfio_iommu *iommu)
2738 struct vfio_domain *domain;
2741 mutex_lock(&iommu->lock);
2742 list_for_each_entry(domain, &iommu->domain_list, next) {
2743 if (!(domain->prot & IOMMU_CACHE)) {
2748 mutex_unlock(&iommu->lock);
2753 static int vfio_iommu_type1_check_extension(struct vfio_iommu *iommu,
2757 case VFIO_TYPE1_IOMMU:
2758 case VFIO_TYPE1v2_IOMMU:
2759 case VFIO_TYPE1_NESTING_IOMMU:
2760 case VFIO_UNMAP_ALL:
2761 case VFIO_UPDATE_VADDR:
2763 case VFIO_DMA_CC_IOMMU:
2766 return vfio_domains_have_iommu_cache(iommu);
2772 static int vfio_iommu_iova_add_cap(struct vfio_info_cap *caps,
2773 struct vfio_iommu_type1_info_cap_iova_range *cap_iovas,
2776 struct vfio_info_cap_header *header;
2777 struct vfio_iommu_type1_info_cap_iova_range *iova_cap;
2779 header = vfio_info_cap_add(caps, size,
2780 VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE, 1);
2782 return PTR_ERR(header);
2784 iova_cap = container_of(header,
2785 struct vfio_iommu_type1_info_cap_iova_range,
2787 iova_cap->nr_iovas = cap_iovas->nr_iovas;
2788 memcpy(iova_cap->iova_ranges, cap_iovas->iova_ranges,
2789 cap_iovas->nr_iovas * sizeof(*cap_iovas->iova_ranges));
2793 static int vfio_iommu_iova_build_caps(struct vfio_iommu *iommu,
2794 struct vfio_info_cap *caps)
2796 struct vfio_iommu_type1_info_cap_iova_range *cap_iovas;
2797 struct vfio_iova *iova;
2799 int iovas = 0, i = 0, ret;
2801 list_for_each_entry(iova, &iommu->iova_list, list)
2806 * Return 0 as a container with a single mdev device
2807 * will have an empty list
2812 size = struct_size(cap_iovas, iova_ranges, iovas);
2814 cap_iovas = kzalloc(size, GFP_KERNEL);
2818 cap_iovas->nr_iovas = iovas;
2820 list_for_each_entry(iova, &iommu->iova_list, list) {
2821 cap_iovas->iova_ranges[i].start = iova->start;
2822 cap_iovas->iova_ranges[i].end = iova->end;
2826 ret = vfio_iommu_iova_add_cap(caps, cap_iovas, size);
2832 static int vfio_iommu_migration_build_caps(struct vfio_iommu *iommu,
2833 struct vfio_info_cap *caps)
2835 struct vfio_iommu_type1_info_cap_migration cap_mig;
2837 cap_mig.header.id = VFIO_IOMMU_TYPE1_INFO_CAP_MIGRATION;
2838 cap_mig.header.version = 1;
2841 /* support minimum pgsize */
2842 cap_mig.pgsize_bitmap = (size_t)1 << __ffs(iommu->pgsize_bitmap);
2843 cap_mig.max_dirty_bitmap_size = DIRTY_BITMAP_SIZE_MAX;
2845 return vfio_info_add_capability(caps, &cap_mig.header, sizeof(cap_mig));
2848 static int vfio_iommu_dma_avail_build_caps(struct vfio_iommu *iommu,
2849 struct vfio_info_cap *caps)
2851 struct vfio_iommu_type1_info_dma_avail cap_dma_avail;
2853 cap_dma_avail.header.id = VFIO_IOMMU_TYPE1_INFO_DMA_AVAIL;
2854 cap_dma_avail.header.version = 1;
2856 cap_dma_avail.avail = iommu->dma_avail;
2858 return vfio_info_add_capability(caps, &cap_dma_avail.header,
2859 sizeof(cap_dma_avail));
2862 static int vfio_iommu_type1_get_info(struct vfio_iommu *iommu,
2865 struct vfio_iommu_type1_info info;
2866 unsigned long minsz;
2867 struct vfio_info_cap caps = { .buf = NULL, .size = 0 };
2868 unsigned long capsz;
2871 minsz = offsetofend(struct vfio_iommu_type1_info, iova_pgsizes);
2873 /* For backward compatibility, cannot require this */
2874 capsz = offsetofend(struct vfio_iommu_type1_info, cap_offset);
2876 if (copy_from_user(&info, (void __user *)arg, minsz))
2879 if (info.argsz < minsz)
2882 if (info.argsz >= capsz) {
2884 info.cap_offset = 0; /* output, no-recopy necessary */
2887 mutex_lock(&iommu->lock);
2888 info.flags = VFIO_IOMMU_INFO_PGSIZES;
2890 info.iova_pgsizes = iommu->pgsize_bitmap;
2892 ret = vfio_iommu_migration_build_caps(iommu, &caps);
2895 ret = vfio_iommu_dma_avail_build_caps(iommu, &caps);
2898 ret = vfio_iommu_iova_build_caps(iommu, &caps);
2900 mutex_unlock(&iommu->lock);
2906 info.flags |= VFIO_IOMMU_INFO_CAPS;
2908 if (info.argsz < sizeof(info) + caps.size) {
2909 info.argsz = sizeof(info) + caps.size;
2911 vfio_info_cap_shift(&caps, sizeof(info));
2912 if (copy_to_user((void __user *)arg +
2913 sizeof(info), caps.buf,
2918 info.cap_offset = sizeof(info);
2924 return copy_to_user((void __user *)arg, &info, minsz) ?
2928 static int vfio_iommu_type1_map_dma(struct vfio_iommu *iommu,
2931 struct vfio_iommu_type1_dma_map map;
2932 unsigned long minsz;
2933 uint32_t mask = VFIO_DMA_MAP_FLAG_READ | VFIO_DMA_MAP_FLAG_WRITE |
2934 VFIO_DMA_MAP_FLAG_VADDR;
2936 minsz = offsetofend(struct vfio_iommu_type1_dma_map, size);
2938 if (copy_from_user(&map, (void __user *)arg, minsz))
2941 if (map.argsz < minsz || map.flags & ~mask)
2944 return vfio_dma_do_map(iommu, &map);
2947 static int vfio_iommu_type1_unmap_dma(struct vfio_iommu *iommu,
2950 struct vfio_iommu_type1_dma_unmap unmap;
2951 struct vfio_bitmap bitmap = { 0 };
2952 uint32_t mask = VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP |
2953 VFIO_DMA_UNMAP_FLAG_VADDR |
2954 VFIO_DMA_UNMAP_FLAG_ALL;
2955 unsigned long minsz;
2958 minsz = offsetofend(struct vfio_iommu_type1_dma_unmap, size);
2960 if (copy_from_user(&unmap, (void __user *)arg, minsz))
2963 if (unmap.argsz < minsz || unmap.flags & ~mask)
2966 if ((unmap.flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) &&
2967 (unmap.flags & (VFIO_DMA_UNMAP_FLAG_ALL |
2968 VFIO_DMA_UNMAP_FLAG_VADDR)))
2971 if (unmap.flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) {
2972 unsigned long pgshift;
2974 if (unmap.argsz < (minsz + sizeof(bitmap)))
2977 if (copy_from_user(&bitmap,
2978 (void __user *)(arg + minsz),
2982 if (!access_ok((void __user *)bitmap.data, bitmap.size))
2985 pgshift = __ffs(bitmap.pgsize);
2986 ret = verify_bitmap_size(unmap.size >> pgshift,
2992 ret = vfio_dma_do_unmap(iommu, &unmap, &bitmap);
2996 return copy_to_user((void __user *)arg, &unmap, minsz) ?
3000 static int vfio_iommu_type1_dirty_pages(struct vfio_iommu *iommu,
3003 struct vfio_iommu_type1_dirty_bitmap dirty;
3004 uint32_t mask = VFIO_IOMMU_DIRTY_PAGES_FLAG_START |
3005 VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP |
3006 VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP;
3007 unsigned long minsz;
3013 minsz = offsetofend(struct vfio_iommu_type1_dirty_bitmap, flags);
3015 if (copy_from_user(&dirty, (void __user *)arg, minsz))
3018 if (dirty.argsz < minsz || dirty.flags & ~mask)
3021 /* only one flag should be set at a time */
3022 if (__ffs(dirty.flags) != __fls(dirty.flags))
3025 if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_START) {
3028 mutex_lock(&iommu->lock);
3029 pgsize = 1 << __ffs(iommu->pgsize_bitmap);
3030 if (!iommu->dirty_page_tracking) {
3031 ret = vfio_dma_bitmap_alloc_all(iommu, pgsize);
3033 iommu->dirty_page_tracking = true;
3035 mutex_unlock(&iommu->lock);
3037 } else if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP) {
3038 mutex_lock(&iommu->lock);
3039 if (iommu->dirty_page_tracking) {
3040 iommu->dirty_page_tracking = false;
3041 vfio_dma_bitmap_free_all(iommu);
3043 mutex_unlock(&iommu->lock);
3045 } else if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP) {
3046 struct vfio_iommu_type1_dirty_bitmap_get range;
3047 unsigned long pgshift;
3048 size_t data_size = dirty.argsz - minsz;
3049 size_t iommu_pgsize;
3051 if (!data_size || data_size < sizeof(range))
3054 if (copy_from_user(&range, (void __user *)(arg + minsz),
3058 if (range.iova + range.size < range.iova)
3060 if (!access_ok((void __user *)range.bitmap.data,
3064 pgshift = __ffs(range.bitmap.pgsize);
3065 ret = verify_bitmap_size(range.size >> pgshift,
3070 mutex_lock(&iommu->lock);
3072 iommu_pgsize = (size_t)1 << __ffs(iommu->pgsize_bitmap);
3074 /* allow only smallest supported pgsize */
3075 if (range.bitmap.pgsize != iommu_pgsize) {
3079 if (range.iova & (iommu_pgsize - 1)) {
3083 if (!range.size || range.size & (iommu_pgsize - 1)) {
3088 if (iommu->dirty_page_tracking)
3089 ret = vfio_iova_dirty_bitmap(range.bitmap.data,
3092 range.bitmap.pgsize);
3096 mutex_unlock(&iommu->lock);
3104 static long vfio_iommu_type1_ioctl(void *iommu_data,
3105 unsigned int cmd, unsigned long arg)
3107 struct vfio_iommu *iommu = iommu_data;
3110 case VFIO_CHECK_EXTENSION:
3111 return vfio_iommu_type1_check_extension(iommu, arg);
3112 case VFIO_IOMMU_GET_INFO:
3113 return vfio_iommu_type1_get_info(iommu, arg);
3114 case VFIO_IOMMU_MAP_DMA:
3115 return vfio_iommu_type1_map_dma(iommu, arg);
3116 case VFIO_IOMMU_UNMAP_DMA:
3117 return vfio_iommu_type1_unmap_dma(iommu, arg);
3118 case VFIO_IOMMU_DIRTY_PAGES:
3119 return vfio_iommu_type1_dirty_pages(iommu, arg);
3125 static int vfio_iommu_type1_register_notifier(void *iommu_data,
3126 unsigned long *events,
3127 struct notifier_block *nb)
3129 struct vfio_iommu *iommu = iommu_data;
3131 /* clear known events */
3132 *events &= ~VFIO_IOMMU_NOTIFY_DMA_UNMAP;
3134 /* refuse to register if still events remaining */
3138 return blocking_notifier_chain_register(&iommu->notifier, nb);
3141 static int vfio_iommu_type1_unregister_notifier(void *iommu_data,
3142 struct notifier_block *nb)
3144 struct vfio_iommu *iommu = iommu_data;
3146 return blocking_notifier_chain_unregister(&iommu->notifier, nb);
3149 static int vfio_iommu_type1_dma_rw_chunk(struct vfio_iommu *iommu,
3150 dma_addr_t user_iova, void *data,
3151 size_t count, bool write,
3154 struct mm_struct *mm;
3155 unsigned long vaddr;
3156 struct vfio_dma *dma;
3157 bool kthread = current->mm == NULL;
3163 ret = vfio_find_dma_valid(iommu, user_iova, 1, &dma);
3167 if ((write && !(dma->prot & IOMMU_WRITE)) ||
3168 !(dma->prot & IOMMU_READ))
3171 mm = get_task_mm(dma->task);
3178 else if (current->mm != mm)
3181 offset = user_iova - dma->iova;
3183 if (count > dma->size - offset)
3184 count = dma->size - offset;
3186 vaddr = dma->vaddr + offset;
3189 *copied = copy_to_user((void __user *)vaddr, data,
3191 if (*copied && iommu->dirty_page_tracking) {
3192 unsigned long pgshift = __ffs(iommu->pgsize_bitmap);
3194 * Bitmap populated with the smallest supported page
3197 bitmap_set(dma->bitmap, offset >> pgshift,
3198 ((offset + *copied - 1) >> pgshift) -
3199 (offset >> pgshift) + 1);
3202 *copied = copy_from_user(data, (void __user *)vaddr,
3205 kthread_unuse_mm(mm);
3208 return *copied ? 0 : -EFAULT;
3211 static int vfio_iommu_type1_dma_rw(void *iommu_data, dma_addr_t user_iova,
3212 void *data, size_t count, bool write)
3214 struct vfio_iommu *iommu = iommu_data;
3218 mutex_lock(&iommu->lock);
3220 ret = vfio_iommu_type1_dma_rw_chunk(iommu, user_iova, data,
3221 count, write, &done);
3230 mutex_unlock(&iommu->lock);
3234 static struct iommu_domain *
3235 vfio_iommu_type1_group_iommu_domain(void *iommu_data,
3236 struct iommu_group *iommu_group)
3238 struct iommu_domain *domain = ERR_PTR(-ENODEV);
3239 struct vfio_iommu *iommu = iommu_data;
3240 struct vfio_domain *d;
3242 if (!iommu || !iommu_group)
3243 return ERR_PTR(-EINVAL);
3245 mutex_lock(&iommu->lock);
3246 list_for_each_entry(d, &iommu->domain_list, next) {
3247 if (find_iommu_group(d, iommu_group)) {
3252 mutex_unlock(&iommu->lock);
3257 static void vfio_iommu_type1_notify(void *iommu_data,
3258 enum vfio_iommu_notify_type event)
3260 struct vfio_iommu *iommu = iommu_data;
3262 if (event != VFIO_IOMMU_CONTAINER_CLOSE)
3264 mutex_lock(&iommu->lock);
3265 iommu->container_open = false;
3266 mutex_unlock(&iommu->lock);
3267 wake_up_all(&iommu->vaddr_wait);
3270 static const struct vfio_iommu_driver_ops vfio_iommu_driver_ops_type1 = {
3271 .name = "vfio-iommu-type1",
3272 .owner = THIS_MODULE,
3273 .open = vfio_iommu_type1_open,
3274 .release = vfio_iommu_type1_release,
3275 .ioctl = vfio_iommu_type1_ioctl,
3276 .attach_group = vfio_iommu_type1_attach_group,
3277 .detach_group = vfio_iommu_type1_detach_group,
3278 .pin_pages = vfio_iommu_type1_pin_pages,
3279 .unpin_pages = vfio_iommu_type1_unpin_pages,
3280 .register_notifier = vfio_iommu_type1_register_notifier,
3281 .unregister_notifier = vfio_iommu_type1_unregister_notifier,
3282 .dma_rw = vfio_iommu_type1_dma_rw,
3283 .group_iommu_domain = vfio_iommu_type1_group_iommu_domain,
3284 .notify = vfio_iommu_type1_notify,
3287 static int __init vfio_iommu_type1_init(void)
3289 return vfio_register_iommu_driver(&vfio_iommu_driver_ops_type1);
3292 static void __exit vfio_iommu_type1_cleanup(void)
3294 vfio_unregister_iommu_driver(&vfio_iommu_driver_ops_type1);
3297 module_init(vfio_iommu_type1_init);
3298 module_exit(vfio_iommu_type1_cleanup);
3300 MODULE_VERSION(DRIVER_VERSION);
3301 MODULE_LICENSE("GPL v2");
3302 MODULE_AUTHOR(DRIVER_AUTHOR);
3303 MODULE_DESCRIPTION(DRIVER_DESC);