1 // SPDX-License-Identifier: GPL-2.0-only
3 * tools/testing/selftests/kvm/lib/kvm_util.c
5 * Copyright (C) 2018, Google LLC.
8 #define _GNU_SOURCE /* for program_invocation_name */
11 #include "kvm_util_internal.h"
12 #include "processor.h"
16 #include <sys/types.h>
19 #include <linux/kernel.h>
21 #define KVM_UTIL_MIN_PFN 2
23 static int vcpu_mmap_sz(void);
25 /* Aligns x up to the next multiple of size. Size must be a power of 2. */
26 static void *align(void *x, size_t size)
28 size_t mask = size - 1;
29 TEST_ASSERT(size != 0 && !(size & (size - 1)),
30 "size not a power of 2: %lu", size);
31 return (void *) (((size_t) x + mask) & ~mask);
43 * On success, the Value corresponding to the capability (KVM_CAP_*)
44 * specified by the value of cap. On failure a TEST_ASSERT failure
47 * Looks up and returns the value corresponding to the capability
48 * (KVM_CAP_*) given by cap.
50 int kvm_check_cap(long cap)
55 kvm_fd = open(KVM_DEV_PATH, O_RDONLY);
59 ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, cap);
60 TEST_ASSERT(ret != -1, "KVM_CHECK_EXTENSION IOCTL failed,\n"
61 " rc: %i errno: %i", ret, errno);
68 /* VM Enable Capability
71 * vm - Virtual Machine
76 * Return: On success, 0. On failure a TEST_ASSERT failure is produced.
78 * Enables a capability (KVM_CAP_*) on the VM.
80 int vm_enable_cap(struct kvm_vm *vm, struct kvm_enable_cap *cap)
84 ret = ioctl(vm->fd, KVM_ENABLE_CAP, cap);
85 TEST_ASSERT(ret == 0, "KVM_ENABLE_CAP IOCTL failed,\n"
86 " rc: %i errno: %i", ret, errno);
91 /* VCPU Enable Capability
94 * vm - Virtual Machine
100 * Return: On success, 0. On failure a TEST_ASSERT failure is produced.
102 * Enables a capability (KVM_CAP_*) on the VCPU.
104 int vcpu_enable_cap(struct kvm_vm *vm, uint32_t vcpu_id,
105 struct kvm_enable_cap *cap)
107 struct vcpu *vcpu = vcpu_find(vm, vcpu_id);
110 TEST_ASSERT(vcpu, "cannot find vcpu %d", vcpu_id);
112 r = ioctl(vcpu->fd, KVM_ENABLE_CAP, cap);
113 TEST_ASSERT(!r, "KVM_ENABLE_CAP vCPU ioctl failed,\n"
114 " rc: %i, errno: %i", r, errno);
119 void vm_enable_dirty_ring(struct kvm_vm *vm, uint32_t ring_size)
121 struct kvm_enable_cap cap = { 0 };
123 cap.cap = KVM_CAP_DIRTY_LOG_RING;
124 cap.args[0] = ring_size;
125 vm_enable_cap(vm, &cap);
126 vm->dirty_ring_size = ring_size;
129 static void vm_open(struct kvm_vm *vm, int perm)
131 vm->kvm_fd = open(KVM_DEV_PATH, perm);
135 if (!kvm_check_cap(KVM_CAP_IMMEDIATE_EXIT)) {
136 print_skip("immediate_exit not available");
140 vm->fd = ioctl(vm->kvm_fd, KVM_CREATE_VM, vm->type);
141 TEST_ASSERT(vm->fd >= 0, "KVM_CREATE_VM ioctl failed, "
142 "rc: %i errno: %i", vm->fd, errno);
145 const char *vm_guest_mode_string(uint32_t i)
147 static const char * const strings[] = {
148 [VM_MODE_P52V48_4K] = "PA-bits:52, VA-bits:48, 4K pages",
149 [VM_MODE_P52V48_64K] = "PA-bits:52, VA-bits:48, 64K pages",
150 [VM_MODE_P48V48_4K] = "PA-bits:48, VA-bits:48, 4K pages",
151 [VM_MODE_P48V48_64K] = "PA-bits:48, VA-bits:48, 64K pages",
152 [VM_MODE_P40V48_4K] = "PA-bits:40, VA-bits:48, 4K pages",
153 [VM_MODE_P40V48_64K] = "PA-bits:40, VA-bits:48, 64K pages",
154 [VM_MODE_PXXV48_4K] = "PA-bits:ANY, VA-bits:48, 4K pages",
156 _Static_assert(sizeof(strings)/sizeof(char *) == NUM_VM_MODES,
157 "Missing new mode strings?");
159 TEST_ASSERT(i < NUM_VM_MODES, "Guest mode ID %d too big", i);
164 const struct vm_guest_mode_params vm_guest_mode_params[] = {
165 { 52, 48, 0x1000, 12 },
166 { 52, 48, 0x10000, 16 },
167 { 48, 48, 0x1000, 12 },
168 { 48, 48, 0x10000, 16 },
169 { 40, 48, 0x1000, 12 },
170 { 40, 48, 0x10000, 16 },
171 { 0, 0, 0x1000, 12 },
173 _Static_assert(sizeof(vm_guest_mode_params)/sizeof(struct vm_guest_mode_params) == NUM_VM_MODES,
174 "Missing new mode params?");
180 * mode - VM Mode (e.g. VM_MODE_P52V48_4K)
181 * phy_pages - Physical memory pages
187 * Pointer to opaque structure that describes the created VM.
189 * Creates a VM with the mode specified by mode (e.g. VM_MODE_P52V48_4K).
190 * When phy_pages is non-zero, a memory region of phy_pages physical pages
191 * is created and mapped starting at guest physical address 0. The file
192 * descriptor to control the created VM is created with the permissions
193 * given by perm (e.g. O_RDWR).
195 struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
199 pr_debug("%s: mode='%s' pages='%ld' perm='%d'\n", __func__,
200 vm_guest_mode_string(mode), phy_pages, perm);
202 vm = calloc(1, sizeof(*vm));
203 TEST_ASSERT(vm != NULL, "Insufficient Memory");
205 INIT_LIST_HEAD(&vm->vcpus);
206 INIT_LIST_HEAD(&vm->userspace_mem_regions);
211 vm->pa_bits = vm_guest_mode_params[mode].pa_bits;
212 vm->va_bits = vm_guest_mode_params[mode].va_bits;
213 vm->page_size = vm_guest_mode_params[mode].page_size;
214 vm->page_shift = vm_guest_mode_params[mode].page_shift;
216 /* Setup mode specific traits. */
218 case VM_MODE_P52V48_4K:
219 vm->pgtable_levels = 4;
221 case VM_MODE_P52V48_64K:
222 vm->pgtable_levels = 3;
224 case VM_MODE_P48V48_4K:
225 vm->pgtable_levels = 4;
227 case VM_MODE_P48V48_64K:
228 vm->pgtable_levels = 3;
230 case VM_MODE_P40V48_4K:
231 vm->pgtable_levels = 4;
233 case VM_MODE_P40V48_64K:
234 vm->pgtable_levels = 3;
236 case VM_MODE_PXXV48_4K:
238 kvm_get_cpu_address_width(&vm->pa_bits, &vm->va_bits);
240 * Ignore KVM support for 5-level paging (vm->va_bits == 57),
241 * it doesn't take effect unless a CR4.LA57 is set, which it
242 * isn't for this VM_MODE.
244 TEST_ASSERT(vm->va_bits == 48 || vm->va_bits == 57,
245 "Linear address width (%d bits) not supported",
247 pr_debug("Guest physical address width detected: %d\n",
249 vm->pgtable_levels = 4;
252 TEST_FAIL("VM_MODE_PXXV48_4K not supported on non-x86 platforms");
256 TEST_FAIL("Unknown guest mode, mode: 0x%x", mode);
260 if (vm->pa_bits != 40)
261 vm->type = KVM_VM_TYPE_ARM_IPA_SIZE(vm->pa_bits);
266 /* Limit to VA-bit canonical virtual addresses. */
267 vm->vpages_valid = sparsebit_alloc();
268 sparsebit_set_num(vm->vpages_valid,
269 0, (1ULL << (vm->va_bits - 1)) >> vm->page_shift);
270 sparsebit_set_num(vm->vpages_valid,
271 (~((1ULL << (vm->va_bits - 1)) - 1)) >> vm->page_shift,
272 (1ULL << (vm->va_bits - 1)) >> vm->page_shift);
274 /* Limit physical addresses to PA-bits. */
275 vm->max_gfn = ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
277 /* Allocate and setup memory for guest. */
278 vm->vpages_mapped = sparsebit_alloc();
280 vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS,
286 struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
287 uint64_t extra_mem_pages, uint32_t num_percpu_pages,
288 void *guest_code, uint32_t vcpuids[])
290 /* The maximum page table size for a memory region will be when the
291 * smallest pages are used. Considering each page contains x page
292 * table descriptors, the total extra size for page tables (for extra
293 * N pages) will be: N/x+N/x^2+N/x^3+... which is definitely smaller
296 uint64_t vcpu_pages = (DEFAULT_STACK_PGS + num_percpu_pages) * nr_vcpus;
297 uint64_t extra_pg_pages = (extra_mem_pages + vcpu_pages) / PTES_PER_MIN_PAGE * 2;
298 uint64_t pages = DEFAULT_GUEST_PHY_PAGES + vcpu_pages + extra_pg_pages;
302 TEST_ASSERT(nr_vcpus <= kvm_check_cap(KVM_CAP_MAX_VCPUS),
303 "nr_vcpus = %d too large for host, max-vcpus = %d",
304 nr_vcpus, kvm_check_cap(KVM_CAP_MAX_VCPUS));
306 pages = vm_adjust_num_guest_pages(mode, pages);
307 vm = vm_create(mode, pages, O_RDWR);
309 kvm_vm_elf_load(vm, program_invocation_name, 0, 0);
312 vm_create_irqchip(vm);
315 for (i = 0; i < nr_vcpus; ++i) {
316 uint32_t vcpuid = vcpuids ? vcpuids[i] : i;
318 vm_vcpu_add_default(vm, vcpuid, guest_code);
321 vcpu_set_cpuid(vm, vcpuid, kvm_get_supported_cpuid());
328 struct kvm_vm *vm_create_default_with_vcpus(uint32_t nr_vcpus, uint64_t extra_mem_pages,
329 uint32_t num_percpu_pages, void *guest_code,
332 return vm_create_with_vcpus(VM_MODE_DEFAULT, nr_vcpus, extra_mem_pages,
333 num_percpu_pages, guest_code, vcpuids);
336 struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_pages,
339 return vm_create_default_with_vcpus(1, extra_mem_pages, 0, guest_code,
340 (uint32_t []){ vcpuid });
347 * vm - VM that has been released before
352 * Reopens the file descriptors associated to the VM and reinstates the
353 * global state, such as the irqchip and the memory regions that are mapped
356 void kvm_vm_restart(struct kvm_vm *vmp, int perm)
358 struct userspace_mem_region *region;
361 if (vmp->has_irqchip)
362 vm_create_irqchip(vmp);
364 list_for_each_entry(region, &vmp->userspace_mem_regions, list) {
365 int ret = ioctl(vmp->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
366 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
367 " rc: %i errno: %i\n"
368 " slot: %u flags: 0x%x\n"
369 " guest_phys_addr: 0x%llx size: 0x%llx",
370 ret, errno, region->region.slot,
371 region->region.flags,
372 region->region.guest_phys_addr,
373 region->region.memory_size);
377 void kvm_vm_get_dirty_log(struct kvm_vm *vm, int slot, void *log)
379 struct kvm_dirty_log args = { .dirty_bitmap = log, .slot = slot };
382 ret = ioctl(vm->fd, KVM_GET_DIRTY_LOG, &args);
383 TEST_ASSERT(ret == 0, "%s: KVM_GET_DIRTY_LOG failed: %s",
384 __func__, strerror(-ret));
387 void kvm_vm_clear_dirty_log(struct kvm_vm *vm, int slot, void *log,
388 uint64_t first_page, uint32_t num_pages)
390 struct kvm_clear_dirty_log args = { .dirty_bitmap = log, .slot = slot,
391 .first_page = first_page,
392 .num_pages = num_pages };
395 ret = ioctl(vm->fd, KVM_CLEAR_DIRTY_LOG, &args);
396 TEST_ASSERT(ret == 0, "%s: KVM_CLEAR_DIRTY_LOG failed: %s",
397 __func__, strerror(-ret));
400 uint32_t kvm_vm_reset_dirty_ring(struct kvm_vm *vm)
402 return ioctl(vm->fd, KVM_RESET_DIRTY_RINGS);
406 * Userspace Memory Region Find
409 * vm - Virtual Machine
410 * start - Starting VM physical address
411 * end - Ending VM physical address, inclusive.
416 * Pointer to overlapping region, NULL if no such region.
418 * Searches for a region with any physical memory that overlaps with
419 * any portion of the guest physical addresses from start to end
420 * inclusive. If multiple overlapping regions exist, a pointer to any
421 * of the regions is returned. Null is returned only when no overlapping
424 static struct userspace_mem_region *
425 userspace_mem_region_find(struct kvm_vm *vm, uint64_t start, uint64_t end)
427 struct userspace_mem_region *region;
429 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
430 uint64_t existing_start = region->region.guest_phys_addr;
431 uint64_t existing_end = region->region.guest_phys_addr
432 + region->region.memory_size - 1;
433 if (start <= existing_end && end >= existing_start)
441 * KVM Userspace Memory Region Find
444 * vm - Virtual Machine
445 * start - Starting VM physical address
446 * end - Ending VM physical address, inclusive.
451 * Pointer to overlapping region, NULL if no such region.
453 * Public interface to userspace_mem_region_find. Allows tests to look up
454 * the memslot datastructure for a given range of guest physical memory.
456 struct kvm_userspace_memory_region *
457 kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
460 struct userspace_mem_region *region;
462 region = userspace_mem_region_find(vm, start, end);
466 return ®ion->region;
473 * vm - Virtual Machine
479 * Pointer to VCPU structure
481 * Locates a vcpu structure that describes the VCPU specified by vcpuid and
482 * returns a pointer to it. Returns NULL if the VM doesn't contain a VCPU
483 * for the specified vcpuid.
485 struct vcpu *vcpu_find(struct kvm_vm *vm, uint32_t vcpuid)
489 list_for_each_entry(vcpu, &vm->vcpus, list) {
490 if (vcpu->id == vcpuid)
501 * vcpu - VCPU to remove
505 * Return: None, TEST_ASSERT failures for all error conditions
507 * Removes a vCPU from a VM and frees its resources.
509 static void vm_vcpu_rm(struct kvm_vm *vm, struct vcpu *vcpu)
513 if (vcpu->dirty_gfns) {
514 ret = munmap(vcpu->dirty_gfns, vm->dirty_ring_size);
515 TEST_ASSERT(ret == 0, "munmap of VCPU dirty ring failed, "
516 "rc: %i errno: %i", ret, errno);
517 vcpu->dirty_gfns = NULL;
520 ret = munmap(vcpu->state, vcpu_mmap_sz());
521 TEST_ASSERT(ret == 0, "munmap of VCPU fd failed, rc: %i "
522 "errno: %i", ret, errno);
523 ret = close(vcpu->fd);
524 TEST_ASSERT(ret == 0, "Close of VCPU fd failed, rc: %i "
525 "errno: %i", ret, errno);
527 list_del(&vcpu->list);
531 void kvm_vm_release(struct kvm_vm *vmp)
533 struct vcpu *vcpu, *tmp;
536 list_for_each_entry_safe(vcpu, tmp, &vmp->vcpus, list)
537 vm_vcpu_rm(vmp, vcpu);
539 ret = close(vmp->fd);
540 TEST_ASSERT(ret == 0, "Close of vm fd failed,\n"
541 " vmp->fd: %i rc: %i errno: %i", vmp->fd, ret, errno);
543 ret = close(vmp->kvm_fd);
544 TEST_ASSERT(ret == 0, "Close of /dev/kvm fd failed,\n"
545 " vmp->kvm_fd: %i rc: %i errno: %i", vmp->kvm_fd, ret, errno);
548 static void __vm_mem_region_delete(struct kvm_vm *vm,
549 struct userspace_mem_region *region)
553 list_del(®ion->list);
555 region->region.memory_size = 0;
556 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
557 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed, "
558 "rc: %i errno: %i", ret, errno);
560 sparsebit_free(®ion->unused_phy_pages);
561 ret = munmap(region->mmap_start, region->mmap_size);
562 TEST_ASSERT(ret == 0, "munmap failed, rc: %i errno: %i", ret, errno);
568 * Destroys and frees the VM pointed to by vmp.
570 void kvm_vm_free(struct kvm_vm *vmp)
572 struct userspace_mem_region *region, *tmp;
577 /* Free userspace_mem_regions. */
578 list_for_each_entry_safe(region, tmp, &vmp->userspace_mem_regions, list)
579 __vm_mem_region_delete(vmp, region);
581 /* Free sparsebit arrays. */
582 sparsebit_free(&vmp->vpages_valid);
583 sparsebit_free(&vmp->vpages_mapped);
587 /* Free the structure describing the VM. */
592 * Memory Compare, host virtual to guest virtual
595 * hva - Starting host virtual address
596 * vm - Virtual Machine
597 * gva - Starting guest virtual address
598 * len - number of bytes to compare
602 * Input/Output Args: None
605 * Returns 0 if the bytes starting at hva for a length of len
606 * are equal the guest virtual bytes starting at gva. Returns
607 * a value < 0, if bytes at hva are less than those at gva.
608 * Otherwise a value > 0 is returned.
610 * Compares the bytes starting at the host virtual address hva, for
611 * a length of len, to the guest bytes starting at the guest virtual
612 * address given by gva.
614 int kvm_memcmp_hva_gva(void *hva, struct kvm_vm *vm, vm_vaddr_t gva, size_t len)
619 * Compare a batch of bytes until either a match is found
620 * or all the bytes have been compared.
622 for (uintptr_t offset = 0; offset < len; offset += amt) {
623 uintptr_t ptr1 = (uintptr_t)hva + offset;
626 * Determine host address for guest virtual address
629 uintptr_t ptr2 = (uintptr_t)addr_gva2hva(vm, gva + offset);
632 * Determine amount to compare on this pass.
633 * Don't allow the comparsion to cross a page boundary.
636 if ((ptr1 >> vm->page_shift) != ((ptr1 + amt) >> vm->page_shift))
637 amt = vm->page_size - (ptr1 % vm->page_size);
638 if ((ptr2 >> vm->page_shift) != ((ptr2 + amt) >> vm->page_shift))
639 amt = vm->page_size - (ptr2 % vm->page_size);
641 assert((ptr1 >> vm->page_shift) == ((ptr1 + amt - 1) >> vm->page_shift));
642 assert((ptr2 >> vm->page_shift) == ((ptr2 + amt - 1) >> vm->page_shift));
645 * Perform the comparison. If there is a difference
646 * return that result to the caller, otherwise need
647 * to continue on looking for a mismatch.
649 int ret = memcmp((void *)ptr1, (void *)ptr2, amt);
655 * No mismatch found. Let the caller know the two memory
662 * VM Userspace Memory Region Add
665 * vm - Virtual Machine
666 * backing_src - Storage source for this region.
667 * NULL to use anonymous memory.
668 * guest_paddr - Starting guest physical address
669 * slot - KVM region slot
670 * npages - Number of physical pages
671 * flags - KVM memory region flags (e.g. KVM_MEM_LOG_DIRTY_PAGES)
677 * Allocates a memory area of the number of pages specified by npages
678 * and maps it to the VM specified by vm, at a starting physical address
679 * given by guest_paddr. The region is created with a KVM region slot
680 * given by slot, which must be unique and < KVM_MEM_SLOTS_NUM. The
681 * region is created with the flags given by flags.
683 void vm_userspace_mem_region_add(struct kvm_vm *vm,
684 enum vm_mem_backing_src_type src_type,
685 uint64_t guest_paddr, uint32_t slot, uint64_t npages,
689 struct userspace_mem_region *region;
690 size_t backing_src_pagesz = get_backing_src_pagesz(src_type);
693 TEST_ASSERT(vm_adjust_num_guest_pages(vm->mode, npages) == npages,
694 "Number of guest pages is not compatible with the host. "
695 "Try npages=%d", vm_adjust_num_guest_pages(vm->mode, npages));
697 TEST_ASSERT((guest_paddr % vm->page_size) == 0, "Guest physical "
698 "address not on a page boundary.\n"
699 " guest_paddr: 0x%lx vm->page_size: 0x%x",
700 guest_paddr, vm->page_size);
701 TEST_ASSERT((((guest_paddr >> vm->page_shift) + npages) - 1)
702 <= vm->max_gfn, "Physical range beyond maximum "
703 "supported physical address,\n"
704 " guest_paddr: 0x%lx npages: 0x%lx\n"
705 " vm->max_gfn: 0x%lx vm->page_size: 0x%x",
706 guest_paddr, npages, vm->max_gfn, vm->page_size);
709 * Confirm a mem region with an overlapping address doesn't
712 region = (struct userspace_mem_region *) userspace_mem_region_find(
713 vm, guest_paddr, (guest_paddr + npages * vm->page_size) - 1);
715 TEST_FAIL("overlapping userspace_mem_region already "
717 " requested guest_paddr: 0x%lx npages: 0x%lx "
719 " existing guest_paddr: 0x%lx size: 0x%lx",
720 guest_paddr, npages, vm->page_size,
721 (uint64_t) region->region.guest_phys_addr,
722 (uint64_t) region->region.memory_size);
724 /* Confirm no region with the requested slot already exists. */
725 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
726 if (region->region.slot != slot)
729 TEST_FAIL("A mem region with the requested slot "
731 " requested slot: %u paddr: 0x%lx npages: 0x%lx\n"
732 " existing slot: %u paddr: 0x%lx size: 0x%lx",
733 slot, guest_paddr, npages,
735 (uint64_t) region->region.guest_phys_addr,
736 (uint64_t) region->region.memory_size);
739 /* Allocate and initialize new mem region structure. */
740 region = calloc(1, sizeof(*region));
741 TEST_ASSERT(region != NULL, "Insufficient Memory");
742 region->mmap_size = npages * vm->page_size;
745 /* On s390x, the host address must be aligned to 1M (due to PGSTEs) */
746 alignment = 0x100000;
751 if (src_type == VM_MEM_SRC_ANONYMOUS_THP)
752 alignment = max(backing_src_pagesz, alignment);
754 /* Add enough memory to align up if necessary */
756 region->mmap_size += alignment;
758 region->mmap_start = mmap(NULL, region->mmap_size,
759 PROT_READ | PROT_WRITE,
760 MAP_PRIVATE | MAP_ANONYMOUS
761 | vm_mem_backing_src_alias(src_type)->flag,
763 TEST_ASSERT(region->mmap_start != MAP_FAILED,
764 "test_malloc failed, mmap_start: %p errno: %i",
765 region->mmap_start, errno);
767 /* Align host address */
768 region->host_mem = align(region->mmap_start, alignment);
770 /* As needed perform madvise */
771 if ((src_type == VM_MEM_SRC_ANONYMOUS ||
772 src_type == VM_MEM_SRC_ANONYMOUS_THP) && thp_configured()) {
773 ret = madvise(region->host_mem, npages * vm->page_size,
774 src_type == VM_MEM_SRC_ANONYMOUS ? MADV_NOHUGEPAGE : MADV_HUGEPAGE);
775 TEST_ASSERT(ret == 0, "madvise failed, addr: %p length: 0x%lx src_type: %s",
776 region->host_mem, npages * vm->page_size,
777 vm_mem_backing_src_alias(src_type)->name);
780 region->unused_phy_pages = sparsebit_alloc();
781 sparsebit_set_num(region->unused_phy_pages,
782 guest_paddr >> vm->page_shift, npages);
783 region->region.slot = slot;
784 region->region.flags = flags;
785 region->region.guest_phys_addr = guest_paddr;
786 region->region.memory_size = npages * vm->page_size;
787 region->region.userspace_addr = (uintptr_t) region->host_mem;
788 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
789 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
790 " rc: %i errno: %i\n"
791 " slot: %u flags: 0x%x\n"
792 " guest_phys_addr: 0x%lx size: 0x%lx",
793 ret, errno, slot, flags,
794 guest_paddr, (uint64_t) region->region.memory_size);
796 /* Add to linked-list of memory regions. */
797 list_add(®ion->list, &vm->userspace_mem_regions);
804 * vm - Virtual Machine
805 * memslot - KVM memory slot ID
810 * Pointer to memory region structure that describe memory region
811 * using kvm memory slot ID given by memslot. TEST_ASSERT failure
812 * on error (e.g. currently no memory region using memslot as a KVM
815 struct userspace_mem_region *
816 memslot2region(struct kvm_vm *vm, uint32_t memslot)
818 struct userspace_mem_region *region;
820 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
821 if (region->region.slot == memslot)
825 fprintf(stderr, "No mem region with the requested slot found,\n"
826 " requested slot: %u\n", memslot);
827 fputs("---- vm dump ----\n", stderr);
828 vm_dump(stderr, vm, 2);
829 TEST_FAIL("Mem region not found");
834 * VM Memory Region Flags Set
837 * vm - Virtual Machine
838 * flags - Starting guest physical address
844 * Sets the flags of the memory region specified by the value of slot,
845 * to the values given by flags.
847 void vm_mem_region_set_flags(struct kvm_vm *vm, uint32_t slot, uint32_t flags)
850 struct userspace_mem_region *region;
852 region = memslot2region(vm, slot);
854 region->region.flags = flags;
856 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
858 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
859 " rc: %i errno: %i slot: %u flags: 0x%x",
860 ret, errno, slot, flags);
864 * VM Memory Region Move
867 * vm - Virtual Machine
868 * slot - Slot of the memory region to move
869 * new_gpa - Starting guest physical address
875 * Change the gpa of a memory region.
877 void vm_mem_region_move(struct kvm_vm *vm, uint32_t slot, uint64_t new_gpa)
879 struct userspace_mem_region *region;
882 region = memslot2region(vm, slot);
884 region->region.guest_phys_addr = new_gpa;
886 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
888 TEST_ASSERT(!ret, "KVM_SET_USER_MEMORY_REGION failed\n"
889 "ret: %i errno: %i slot: %u new_gpa: 0x%lx",
890 ret, errno, slot, new_gpa);
894 * VM Memory Region Delete
897 * vm - Virtual Machine
898 * slot - Slot of the memory region to delete
904 * Delete a memory region.
906 void vm_mem_region_delete(struct kvm_vm *vm, uint32_t slot)
908 __vm_mem_region_delete(vm, memslot2region(vm, slot));
921 * Returns the size of the structure pointed to by the return value
924 static int vcpu_mmap_sz(void)
928 dev_fd = open(KVM_DEV_PATH, O_RDONLY);
932 ret = ioctl(dev_fd, KVM_GET_VCPU_MMAP_SIZE, NULL);
933 TEST_ASSERT(ret >= sizeof(struct kvm_run),
934 "%s KVM_GET_VCPU_MMAP_SIZE ioctl failed, rc: %i errno: %i",
935 __func__, ret, errno);
946 * vm - Virtual Machine
953 * Adds a virtual CPU to the VM specified by vm with the ID given by vcpuid.
954 * No additional VCPU setup is done.
956 void vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpuid)
960 /* Confirm a vcpu with the specified id doesn't already exist. */
961 vcpu = vcpu_find(vm, vcpuid);
963 TEST_FAIL("vcpu with the specified id "
965 " requested vcpuid: %u\n"
966 " existing vcpuid: %u state: %p",
967 vcpuid, vcpu->id, vcpu->state);
969 /* Allocate and initialize new vcpu structure. */
970 vcpu = calloc(1, sizeof(*vcpu));
971 TEST_ASSERT(vcpu != NULL, "Insufficient Memory");
973 vcpu->fd = ioctl(vm->fd, KVM_CREATE_VCPU, vcpuid);
974 TEST_ASSERT(vcpu->fd >= 0, "KVM_CREATE_VCPU failed, rc: %i errno: %i",
977 TEST_ASSERT(vcpu_mmap_sz() >= sizeof(*vcpu->state), "vcpu mmap size "
978 "smaller than expected, vcpu_mmap_sz: %i expected_min: %zi",
979 vcpu_mmap_sz(), sizeof(*vcpu->state));
980 vcpu->state = (struct kvm_run *) mmap(NULL, vcpu_mmap_sz(),
981 PROT_READ | PROT_WRITE, MAP_SHARED, vcpu->fd, 0);
982 TEST_ASSERT(vcpu->state != MAP_FAILED, "mmap vcpu_state failed, "
983 "vcpu id: %u errno: %i", vcpuid, errno);
985 /* Add to linked-list of VCPUs. */
986 list_add(&vcpu->list, &vm->vcpus);
990 * VM Virtual Address Unused Gap
993 * vm - Virtual Machine
995 * vaddr_min - Minimum Virtual Address
1000 * Lowest virtual address at or below vaddr_min, with at least
1001 * sz unused bytes. TEST_ASSERT failure if no area of at least
1002 * size sz is available.
1004 * Within the VM specified by vm, locates the lowest starting virtual
1005 * address >= vaddr_min, that has at least sz unallocated bytes. A
1006 * TEST_ASSERT failure occurs for invalid input or no area of at least
1007 * sz unallocated bytes >= vaddr_min is available.
1009 static vm_vaddr_t vm_vaddr_unused_gap(struct kvm_vm *vm, size_t sz,
1010 vm_vaddr_t vaddr_min)
1012 uint64_t pages = (sz + vm->page_size - 1) >> vm->page_shift;
1014 /* Determine lowest permitted virtual page index. */
1015 uint64_t pgidx_start = (vaddr_min + vm->page_size - 1) >> vm->page_shift;
1016 if ((pgidx_start * vm->page_size) < vaddr_min)
1019 /* Loop over section with enough valid virtual page indexes. */
1020 if (!sparsebit_is_set_num(vm->vpages_valid,
1021 pgidx_start, pages))
1022 pgidx_start = sparsebit_next_set_num(vm->vpages_valid,
1023 pgidx_start, pages);
1026 * Are there enough unused virtual pages available at
1027 * the currently proposed starting virtual page index.
1028 * If not, adjust proposed starting index to next
1031 if (sparsebit_is_clear_num(vm->vpages_mapped,
1032 pgidx_start, pages))
1034 pgidx_start = sparsebit_next_clear_num(vm->vpages_mapped,
1035 pgidx_start, pages);
1036 if (pgidx_start == 0)
1040 * If needed, adjust proposed starting virtual address,
1041 * to next range of valid virtual addresses.
1043 if (!sparsebit_is_set_num(vm->vpages_valid,
1044 pgidx_start, pages)) {
1045 pgidx_start = sparsebit_next_set_num(
1046 vm->vpages_valid, pgidx_start, pages);
1047 if (pgidx_start == 0)
1050 } while (pgidx_start != 0);
1053 TEST_FAIL("No vaddr of specified pages available, pages: 0x%lx", pages);
1059 TEST_ASSERT(sparsebit_is_set_num(vm->vpages_valid,
1060 pgidx_start, pages),
1061 "Unexpected, invalid virtual page index range,\n"
1062 " pgidx_start: 0x%lx\n"
1064 pgidx_start, pages);
1065 TEST_ASSERT(sparsebit_is_clear_num(vm->vpages_mapped,
1066 pgidx_start, pages),
1067 "Unexpected, pages already mapped,\n"
1068 " pgidx_start: 0x%lx\n"
1070 pgidx_start, pages);
1072 return pgidx_start * vm->page_size;
1076 * VM Virtual Address Allocate
1079 * vm - Virtual Machine
1080 * sz - Size in bytes
1081 * vaddr_min - Minimum starting virtual address
1082 * data_memslot - Memory region slot for data pages
1083 * pgd_memslot - Memory region slot for new virtual translation tables
1088 * Starting guest virtual address
1090 * Allocates at least sz bytes within the virtual address space of the vm
1091 * given by vm. The allocated bytes are mapped to a virtual address >=
1092 * the address given by vaddr_min. Note that each allocation uses a
1093 * a unique set of pages, with the minimum real allocation being at least
1096 vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
1097 uint32_t data_memslot, uint32_t pgd_memslot)
1099 uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
1101 virt_pgd_alloc(vm, pgd_memslot);
1104 * Find an unused range of virtual page addresses of at least
1107 vm_vaddr_t vaddr_start = vm_vaddr_unused_gap(vm, sz, vaddr_min);
1109 /* Map the virtual pages. */
1110 for (vm_vaddr_t vaddr = vaddr_start; pages > 0;
1111 pages--, vaddr += vm->page_size) {
1114 paddr = vm_phy_page_alloc(vm,
1115 KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
1117 virt_pg_map(vm, vaddr, paddr, pgd_memslot);
1119 sparsebit_set(vm->vpages_mapped,
1120 vaddr >> vm->page_shift);
1127 * Map a range of VM virtual address to the VM's physical address
1130 * vm - Virtual Machine
1131 * vaddr - Virtuall address to map
1132 * paddr - VM Physical Address
1133 * npages - The number of pages to map
1134 * pgd_memslot - Memory region slot for new virtual translation tables
1140 * Within the VM given by @vm, creates a virtual translation for
1141 * @npages starting at @vaddr to the page range starting at @paddr.
1143 void virt_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr,
1144 unsigned int npages, uint32_t pgd_memslot)
1146 size_t page_size = vm->page_size;
1147 size_t size = npages * page_size;
1149 TEST_ASSERT(vaddr + size > vaddr, "Vaddr overflow");
1150 TEST_ASSERT(paddr + size > paddr, "Paddr overflow");
1153 virt_pg_map(vm, vaddr, paddr, pgd_memslot);
1160 * Address VM Physical to Host Virtual
1163 * vm - Virtual Machine
1164 * gpa - VM physical address
1169 * Equivalent host virtual address
1171 * Locates the memory region containing the VM physical address given
1172 * by gpa, within the VM given by vm. When found, the host virtual
1173 * address providing the memory to the vm physical address is returned.
1174 * A TEST_ASSERT failure occurs if no region containing gpa exists.
1176 void *addr_gpa2hva(struct kvm_vm *vm, vm_paddr_t gpa)
1178 struct userspace_mem_region *region;
1180 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1181 if ((gpa >= region->region.guest_phys_addr)
1182 && (gpa <= (region->region.guest_phys_addr
1183 + region->region.memory_size - 1)))
1184 return (void *) ((uintptr_t) region->host_mem
1185 + (gpa - region->region.guest_phys_addr));
1188 TEST_FAIL("No vm physical memory at 0x%lx", gpa);
1193 * Address Host Virtual to VM Physical
1196 * vm - Virtual Machine
1197 * hva - Host virtual address
1202 * Equivalent VM physical address
1204 * Locates the memory region containing the host virtual address given
1205 * by hva, within the VM given by vm. When found, the equivalent
1206 * VM physical address is returned. A TEST_ASSERT failure occurs if no
1207 * region containing hva exists.
1209 vm_paddr_t addr_hva2gpa(struct kvm_vm *vm, void *hva)
1211 struct userspace_mem_region *region;
1213 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1214 if ((hva >= region->host_mem)
1215 && (hva <= (region->host_mem
1216 + region->region.memory_size - 1)))
1217 return (vm_paddr_t) ((uintptr_t)
1218 region->region.guest_phys_addr
1219 + (hva - (uintptr_t) region->host_mem));
1222 TEST_FAIL("No mapping to a guest physical address, hva: %p", hva);
1227 * VM Create IRQ Chip
1230 * vm - Virtual Machine
1236 * Creates an interrupt controller chip for the VM specified by vm.
1238 void vm_create_irqchip(struct kvm_vm *vm)
1242 ret = ioctl(vm->fd, KVM_CREATE_IRQCHIP, 0);
1243 TEST_ASSERT(ret == 0, "KVM_CREATE_IRQCHIP IOCTL failed, "
1244 "rc: %i errno: %i", ret, errno);
1246 vm->has_irqchip = true;
1253 * vm - Virtual Machine
1259 * Pointer to structure that describes the state of the VCPU.
1261 * Locates and returns a pointer to a structure that describes the
1262 * state of the VCPU with the given vcpuid.
1264 struct kvm_run *vcpu_state(struct kvm_vm *vm, uint32_t vcpuid)
1266 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1267 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1276 * vm - Virtual Machine
1283 * Switch to executing the code for the VCPU given by vcpuid, within the VM
1286 void vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1288 int ret = _vcpu_run(vm, vcpuid);
1289 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1290 "rc: %i errno: %i", ret, errno);
1293 int _vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1295 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1298 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1300 rc = ioctl(vcpu->fd, KVM_RUN, NULL);
1301 } while (rc == -1 && errno == EINTR);
1303 assert_on_unhandled_exception(vm, vcpuid);
1308 int vcpu_get_fd(struct kvm_vm *vm, uint32_t vcpuid)
1310 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1312 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1317 void vcpu_run_complete_io(struct kvm_vm *vm, uint32_t vcpuid)
1319 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1322 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1324 vcpu->state->immediate_exit = 1;
1325 ret = ioctl(vcpu->fd, KVM_RUN, NULL);
1326 vcpu->state->immediate_exit = 0;
1328 TEST_ASSERT(ret == -1 && errno == EINTR,
1329 "KVM_RUN IOCTL didn't exit immediately, rc: %i, errno: %i",
1333 void vcpu_set_guest_debug(struct kvm_vm *vm, uint32_t vcpuid,
1334 struct kvm_guest_debug *debug)
1336 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1337 int ret = ioctl(vcpu->fd, KVM_SET_GUEST_DEBUG, debug);
1339 TEST_ASSERT(ret == 0, "KVM_SET_GUEST_DEBUG failed: %d", ret);
1343 * VM VCPU Set MP State
1346 * vm - Virtual Machine
1348 * mp_state - mp_state to be set
1354 * Sets the MP state of the VCPU given by vcpuid, to the state given
1357 void vcpu_set_mp_state(struct kvm_vm *vm, uint32_t vcpuid,
1358 struct kvm_mp_state *mp_state)
1360 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1363 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1365 ret = ioctl(vcpu->fd, KVM_SET_MP_STATE, mp_state);
1366 TEST_ASSERT(ret == 0, "KVM_SET_MP_STATE IOCTL failed, "
1367 "rc: %i errno: %i", ret, errno);
1371 * VM VCPU Get Reg List
1374 * vm - Virtual Machine
1381 * A pointer to an allocated struct kvm_reg_list
1383 * Get the list of guest registers which are supported for
1384 * KVM_GET_ONE_REG/KVM_SET_ONE_REG calls
1386 struct kvm_reg_list *vcpu_get_reg_list(struct kvm_vm *vm, uint32_t vcpuid)
1388 struct kvm_reg_list reg_list_n = { .n = 0 }, *reg_list;
1391 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_REG_LIST, ®_list_n);
1392 TEST_ASSERT(ret == -1 && errno == E2BIG, "KVM_GET_REG_LIST n=0");
1393 reg_list = calloc(1, sizeof(*reg_list) + reg_list_n.n * sizeof(__u64));
1394 reg_list->n = reg_list_n.n;
1395 vcpu_ioctl(vm, vcpuid, KVM_GET_REG_LIST, reg_list);
1403 * vm - Virtual Machine
1407 * regs - current state of VCPU regs
1411 * Obtains the current register state for the VCPU specified by vcpuid
1412 * and stores it at the location given by regs.
1414 void vcpu_regs_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_regs *regs)
1416 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1419 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1421 ret = ioctl(vcpu->fd, KVM_GET_REGS, regs);
1422 TEST_ASSERT(ret == 0, "KVM_GET_REGS failed, rc: %i errno: %i",
1430 * vm - Virtual Machine
1432 * regs - Values to set VCPU regs to
1438 * Sets the regs of the VCPU specified by vcpuid to the values
1441 void vcpu_regs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_regs *regs)
1443 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1446 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1448 ret = ioctl(vcpu->fd, KVM_SET_REGS, regs);
1449 TEST_ASSERT(ret == 0, "KVM_SET_REGS failed, rc: %i errno: %i",
1453 #ifdef __KVM_HAVE_VCPU_EVENTS
1454 void vcpu_events_get(struct kvm_vm *vm, uint32_t vcpuid,
1455 struct kvm_vcpu_events *events)
1457 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1460 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1462 ret = ioctl(vcpu->fd, KVM_GET_VCPU_EVENTS, events);
1463 TEST_ASSERT(ret == 0, "KVM_GET_VCPU_EVENTS, failed, rc: %i errno: %i",
1467 void vcpu_events_set(struct kvm_vm *vm, uint32_t vcpuid,
1468 struct kvm_vcpu_events *events)
1470 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1473 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1475 ret = ioctl(vcpu->fd, KVM_SET_VCPU_EVENTS, events);
1476 TEST_ASSERT(ret == 0, "KVM_SET_VCPU_EVENTS, failed, rc: %i errno: %i",
1482 void vcpu_nested_state_get(struct kvm_vm *vm, uint32_t vcpuid,
1483 struct kvm_nested_state *state)
1485 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1488 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1490 ret = ioctl(vcpu->fd, KVM_GET_NESTED_STATE, state);
1491 TEST_ASSERT(ret == 0,
1492 "KVM_SET_NESTED_STATE failed, ret: %i errno: %i",
1496 int vcpu_nested_state_set(struct kvm_vm *vm, uint32_t vcpuid,
1497 struct kvm_nested_state *state, bool ignore_error)
1499 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1502 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1504 ret = ioctl(vcpu->fd, KVM_SET_NESTED_STATE, state);
1505 if (!ignore_error) {
1506 TEST_ASSERT(ret == 0,
1507 "KVM_SET_NESTED_STATE failed, ret: %i errno: %i",
1516 * VM VCPU System Regs Get
1519 * vm - Virtual Machine
1523 * sregs - current state of VCPU system regs
1527 * Obtains the current system register state for the VCPU specified by
1528 * vcpuid and stores it at the location given by sregs.
1530 void vcpu_sregs_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1532 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1535 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1537 ret = ioctl(vcpu->fd, KVM_GET_SREGS, sregs);
1538 TEST_ASSERT(ret == 0, "KVM_GET_SREGS failed, rc: %i errno: %i",
1543 * VM VCPU System Regs Set
1546 * vm - Virtual Machine
1548 * sregs - Values to set VCPU system regs to
1554 * Sets the system regs of the VCPU specified by vcpuid to the values
1557 void vcpu_sregs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1559 int ret = _vcpu_sregs_set(vm, vcpuid, sregs);
1560 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1561 "rc: %i errno: %i", ret, errno);
1564 int _vcpu_sregs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1566 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1568 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1570 return ioctl(vcpu->fd, KVM_SET_SREGS, sregs);
1573 void vcpu_fpu_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu)
1577 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_FPU, fpu);
1578 TEST_ASSERT(ret == 0, "KVM_GET_FPU failed, rc: %i errno: %i (%s)",
1579 ret, errno, strerror(errno));
1582 void vcpu_fpu_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu)
1586 ret = _vcpu_ioctl(vm, vcpuid, KVM_SET_FPU, fpu);
1587 TEST_ASSERT(ret == 0, "KVM_SET_FPU failed, rc: %i errno: %i (%s)",
1588 ret, errno, strerror(errno));
1591 void vcpu_get_reg(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_one_reg *reg)
1595 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_ONE_REG, reg);
1596 TEST_ASSERT(ret == 0, "KVM_GET_ONE_REG failed, rc: %i errno: %i (%s)",
1597 ret, errno, strerror(errno));
1600 void vcpu_set_reg(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_one_reg *reg)
1604 ret = _vcpu_ioctl(vm, vcpuid, KVM_SET_ONE_REG, reg);
1605 TEST_ASSERT(ret == 0, "KVM_SET_ONE_REG failed, rc: %i errno: %i (%s)",
1606 ret, errno, strerror(errno));
1613 * vm - Virtual Machine
1615 * cmd - Ioctl number
1616 * arg - Argument to pass to the ioctl
1620 * Issues an arbitrary ioctl on a VCPU fd.
1622 void vcpu_ioctl(struct kvm_vm *vm, uint32_t vcpuid,
1623 unsigned long cmd, void *arg)
1627 ret = _vcpu_ioctl(vm, vcpuid, cmd, arg);
1628 TEST_ASSERT(ret == 0, "vcpu ioctl %lu failed, rc: %i errno: %i (%s)",
1629 cmd, ret, errno, strerror(errno));
1632 int _vcpu_ioctl(struct kvm_vm *vm, uint32_t vcpuid,
1633 unsigned long cmd, void *arg)
1635 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1638 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1640 ret = ioctl(vcpu->fd, cmd, arg);
1645 void *vcpu_map_dirty_ring(struct kvm_vm *vm, uint32_t vcpuid)
1648 uint32_t size = vm->dirty_ring_size;
1650 TEST_ASSERT(size > 0, "Should enable dirty ring first");
1652 vcpu = vcpu_find(vm, vcpuid);
1654 TEST_ASSERT(vcpu, "Cannot find vcpu %u", vcpuid);
1656 if (!vcpu->dirty_gfns) {
1659 addr = mmap(NULL, size, PROT_READ,
1660 MAP_PRIVATE, vcpu->fd,
1661 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1662 TEST_ASSERT(addr == MAP_FAILED, "Dirty ring mapped private");
1664 addr = mmap(NULL, size, PROT_READ | PROT_EXEC,
1665 MAP_PRIVATE, vcpu->fd,
1666 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1667 TEST_ASSERT(addr == MAP_FAILED, "Dirty ring mapped exec");
1669 addr = mmap(NULL, size, PROT_READ | PROT_WRITE,
1670 MAP_SHARED, vcpu->fd,
1671 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1672 TEST_ASSERT(addr != MAP_FAILED, "Dirty ring map failed");
1674 vcpu->dirty_gfns = addr;
1675 vcpu->dirty_gfns_count = size / sizeof(struct kvm_dirty_gfn);
1678 return vcpu->dirty_gfns;
1685 * vm - Virtual Machine
1686 * cmd - Ioctl number
1687 * arg - Argument to pass to the ioctl
1691 * Issues an arbitrary ioctl on a VM fd.
1693 void vm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1697 ret = _vm_ioctl(vm, cmd, arg);
1698 TEST_ASSERT(ret == 0, "vm ioctl %lu failed, rc: %i errno: %i (%s)",
1699 cmd, ret, errno, strerror(errno));
1702 int _vm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1704 return ioctl(vm->fd, cmd, arg);
1711 * vm - Virtual Machine
1712 * cmd - Ioctl number
1713 * arg - Argument to pass to the ioctl
1717 * Issues an arbitrary ioctl on a KVM fd.
1719 void kvm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1723 ret = ioctl(vm->kvm_fd, cmd, arg);
1724 TEST_ASSERT(ret == 0, "KVM ioctl %lu failed, rc: %i errno: %i (%s)",
1725 cmd, ret, errno, strerror(errno));
1728 int _kvm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1730 return ioctl(vm->kvm_fd, cmd, arg);
1737 int _kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr)
1739 struct kvm_device_attr attribute = {
1745 return ioctl(dev_fd, KVM_HAS_DEVICE_ATTR, &attribute);
1748 int kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr)
1750 int ret = _kvm_device_check_attr(dev_fd, group, attr);
1752 TEST_ASSERT(ret >= 0, "KVM_HAS_DEVICE_ATTR failed, rc: %i errno: %i", ret, errno);
1756 int _kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test, int *fd)
1758 struct kvm_create_device create_dev;
1761 create_dev.type = type;
1763 create_dev.flags = test ? KVM_CREATE_DEVICE_TEST : 0;
1764 ret = ioctl(vm_get_fd(vm), KVM_CREATE_DEVICE, &create_dev);
1765 *fd = create_dev.fd;
1769 int kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test)
1773 ret = _kvm_create_device(vm, type, test, &fd);
1776 TEST_ASSERT(ret >= 0,
1777 "KVM_CREATE_DEVICE IOCTL failed, rc: %i errno: %i", ret, errno);
1783 int _kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
1784 void *val, bool write)
1786 struct kvm_device_attr kvmattr = {
1790 .addr = (uintptr_t)val,
1794 ret = ioctl(dev_fd, write ? KVM_SET_DEVICE_ATTR : KVM_GET_DEVICE_ATTR,
1799 int kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,
1800 void *val, bool write)
1802 int ret = _kvm_device_access(dev_fd, group, attr, val, write);
1804 TEST_ASSERT(ret >= 0, "KVM_SET|GET_DEVICE_ATTR IOCTL failed, rc: %i errno: %i", ret, errno);
1812 * vm - Virtual Machine
1813 * indent - Left margin indent amount
1816 * stream - Output FILE stream
1820 * Dumps the current state of the VM given by vm, to the FILE stream
1823 void vm_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
1825 struct userspace_mem_region *region;
1828 fprintf(stream, "%*smode: 0x%x\n", indent, "", vm->mode);
1829 fprintf(stream, "%*sfd: %i\n", indent, "", vm->fd);
1830 fprintf(stream, "%*spage_size: 0x%x\n", indent, "", vm->page_size);
1831 fprintf(stream, "%*sMem Regions:\n", indent, "");
1832 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1833 fprintf(stream, "%*sguest_phys: 0x%lx size: 0x%lx "
1834 "host_virt: %p\n", indent + 2, "",
1835 (uint64_t) region->region.guest_phys_addr,
1836 (uint64_t) region->region.memory_size,
1838 fprintf(stream, "%*sunused_phy_pages: ", indent + 2, "");
1839 sparsebit_dump(stream, region->unused_phy_pages, 0);
1841 fprintf(stream, "%*sMapped Virtual Pages:\n", indent, "");
1842 sparsebit_dump(stream, vm->vpages_mapped, indent + 2);
1843 fprintf(stream, "%*spgd_created: %u\n", indent, "",
1845 if (vm->pgd_created) {
1846 fprintf(stream, "%*sVirtual Translation Tables:\n",
1848 virt_dump(stream, vm, indent + 4);
1850 fprintf(stream, "%*sVCPUs:\n", indent, "");
1851 list_for_each_entry(vcpu, &vm->vcpus, list)
1852 vcpu_dump(stream, vm, vcpu->id, indent + 2);
1855 /* Known KVM exit reasons */
1856 static struct exit_reason {
1857 unsigned int reason;
1859 } exit_reasons_known[] = {
1860 {KVM_EXIT_UNKNOWN, "UNKNOWN"},
1861 {KVM_EXIT_EXCEPTION, "EXCEPTION"},
1862 {KVM_EXIT_IO, "IO"},
1863 {KVM_EXIT_HYPERCALL, "HYPERCALL"},
1864 {KVM_EXIT_DEBUG, "DEBUG"},
1865 {KVM_EXIT_HLT, "HLT"},
1866 {KVM_EXIT_MMIO, "MMIO"},
1867 {KVM_EXIT_IRQ_WINDOW_OPEN, "IRQ_WINDOW_OPEN"},
1868 {KVM_EXIT_SHUTDOWN, "SHUTDOWN"},
1869 {KVM_EXIT_FAIL_ENTRY, "FAIL_ENTRY"},
1870 {KVM_EXIT_INTR, "INTR"},
1871 {KVM_EXIT_SET_TPR, "SET_TPR"},
1872 {KVM_EXIT_TPR_ACCESS, "TPR_ACCESS"},
1873 {KVM_EXIT_S390_SIEIC, "S390_SIEIC"},
1874 {KVM_EXIT_S390_RESET, "S390_RESET"},
1875 {KVM_EXIT_DCR, "DCR"},
1876 {KVM_EXIT_NMI, "NMI"},
1877 {KVM_EXIT_INTERNAL_ERROR, "INTERNAL_ERROR"},
1878 {KVM_EXIT_OSI, "OSI"},
1879 {KVM_EXIT_PAPR_HCALL, "PAPR_HCALL"},
1880 {KVM_EXIT_DIRTY_RING_FULL, "DIRTY_RING_FULL"},
1881 {KVM_EXIT_X86_RDMSR, "RDMSR"},
1882 {KVM_EXIT_X86_WRMSR, "WRMSR"},
1883 {KVM_EXIT_XEN, "XEN"},
1884 #ifdef KVM_EXIT_MEMORY_NOT_PRESENT
1885 {KVM_EXIT_MEMORY_NOT_PRESENT, "MEMORY_NOT_PRESENT"},
1890 * Exit Reason String
1893 * exit_reason - Exit reason
1898 * Constant string pointer describing the exit reason.
1900 * Locates and returns a constant string that describes the KVM exit
1901 * reason given by exit_reason. If no such string is found, a constant
1902 * string of "Unknown" is returned.
1904 const char *exit_reason_str(unsigned int exit_reason)
1908 for (n1 = 0; n1 < ARRAY_SIZE(exit_reasons_known); n1++) {
1909 if (exit_reason == exit_reasons_known[n1].reason)
1910 return exit_reasons_known[n1].name;
1917 * Physical Contiguous Page Allocator
1920 * vm - Virtual Machine
1921 * num - number of pages
1922 * paddr_min - Physical address minimum
1923 * memslot - Memory region to allocate page from
1928 * Starting physical address
1930 * Within the VM specified by vm, locates a range of available physical
1931 * pages at or above paddr_min. If found, the pages are marked as in use
1932 * and their base address is returned. A TEST_ASSERT failure occurs if
1933 * not enough pages are available at or above paddr_min.
1935 vm_paddr_t vm_phy_pages_alloc(struct kvm_vm *vm, size_t num,
1936 vm_paddr_t paddr_min, uint32_t memslot)
1938 struct userspace_mem_region *region;
1939 sparsebit_idx_t pg, base;
1941 TEST_ASSERT(num > 0, "Must allocate at least one page");
1943 TEST_ASSERT((paddr_min % vm->page_size) == 0, "Min physical address "
1944 "not divisible by page size.\n"
1945 " paddr_min: 0x%lx page_size: 0x%x",
1946 paddr_min, vm->page_size);
1948 region = memslot2region(vm, memslot);
1949 base = pg = paddr_min >> vm->page_shift;
1952 for (; pg < base + num; ++pg) {
1953 if (!sparsebit_is_set(region->unused_phy_pages, pg)) {
1954 base = pg = sparsebit_next_set(region->unused_phy_pages, pg);
1958 } while (pg && pg != base + num);
1961 fprintf(stderr, "No guest physical page available, "
1962 "paddr_min: 0x%lx page_size: 0x%x memslot: %u\n",
1963 paddr_min, vm->page_size, memslot);
1964 fputs("---- vm dump ----\n", stderr);
1965 vm_dump(stderr, vm, 2);
1969 for (pg = base; pg < base + num; ++pg)
1970 sparsebit_clear(region->unused_phy_pages, pg);
1972 return base * vm->page_size;
1975 vm_paddr_t vm_phy_page_alloc(struct kvm_vm *vm, vm_paddr_t paddr_min,
1978 return vm_phy_pages_alloc(vm, 1, paddr_min, memslot);
1982 * Address Guest Virtual to Host Virtual
1985 * vm - Virtual Machine
1986 * gva - VM virtual address
1991 * Equivalent host virtual address
1993 void *addr_gva2hva(struct kvm_vm *vm, vm_vaddr_t gva)
1995 return addr_gpa2hva(vm, addr_gva2gpa(vm, gva));
1999 * Is Unrestricted Guest
2002 * vm - Virtual Machine
2006 * Return: True if the unrestricted guest is set to 'Y', otherwise return false.
2008 * Check if the unrestricted guest flag is enabled.
2010 bool vm_is_unrestricted_guest(struct kvm_vm *vm)
2017 /* Ensure that the KVM vendor-specific module is loaded. */
2018 f = fopen(KVM_DEV_PATH, "r");
2019 TEST_ASSERT(f != NULL, "Error in opening KVM dev file: %d",
2024 f = fopen("/sys/module/kvm_intel/parameters/unrestricted_guest", "r");
2026 count = fread(&val, sizeof(char), 1, f);
2027 TEST_ASSERT(count == 1, "Unable to read from param file.");
2034 unsigned int vm_get_page_size(struct kvm_vm *vm)
2036 return vm->page_size;
2039 unsigned int vm_get_page_shift(struct kvm_vm *vm)
2041 return vm->page_shift;
2044 unsigned int vm_get_max_gfn(struct kvm_vm *vm)
2049 int vm_get_fd(struct kvm_vm *vm)
2054 static unsigned int vm_calc_num_pages(unsigned int num_pages,
2055 unsigned int page_shift,
2056 unsigned int new_page_shift,
2059 unsigned int n = 1 << (new_page_shift - page_shift);
2061 if (page_shift >= new_page_shift)
2062 return num_pages * (1 << (page_shift - new_page_shift));
2064 return num_pages / n + !!(ceil && num_pages % n);
2067 static inline int getpageshift(void)
2069 return __builtin_ffs(getpagesize()) - 1;
2073 vm_num_host_pages(enum vm_guest_mode mode, unsigned int num_guest_pages)
2075 return vm_calc_num_pages(num_guest_pages,
2076 vm_guest_mode_params[mode].page_shift,
2077 getpageshift(), true);
2081 vm_num_guest_pages(enum vm_guest_mode mode, unsigned int num_host_pages)
2083 return vm_calc_num_pages(num_host_pages, getpageshift(),
2084 vm_guest_mode_params[mode].page_shift, false);
2087 unsigned int vm_calc_num_guest_pages(enum vm_guest_mode mode, size_t size)
2090 n = DIV_ROUND_UP(size, vm_guest_mode_params[mode].page_size);
2091 return vm_adjust_num_guest_pages(mode, n);