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_PGS_PER_HUGEPG 512
22 #define KVM_UTIL_MIN_PFN 2
24 /* Aligns x up to the next multiple of size. Size must be a power of 2. */
25 static void *align(void *x, size_t size)
27 size_t mask = size - 1;
28 TEST_ASSERT(size != 0 && !(size & (size - 1)),
29 "size not a power of 2: %lu", size);
30 return (void *) (((size_t) x + mask) & ~mask);
42 * On success, the Value corresponding to the capability (KVM_CAP_*)
43 * specified by the value of cap. On failure a TEST_ASSERT failure
46 * Looks up and returns the value corresponding to the capability
47 * (KVM_CAP_*) given by cap.
49 int kvm_check_cap(long cap)
54 kvm_fd = open(KVM_DEV_PATH, O_RDONLY);
58 ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, cap);
59 TEST_ASSERT(ret != -1, "KVM_CHECK_EXTENSION IOCTL failed,\n"
60 " rc: %i errno: %i", ret, errno);
67 /* VM Enable Capability
70 * vm - Virtual Machine
75 * Return: On success, 0. On failure a TEST_ASSERT failure is produced.
77 * Enables a capability (KVM_CAP_*) on the VM.
79 int vm_enable_cap(struct kvm_vm *vm, struct kvm_enable_cap *cap)
83 ret = ioctl(vm->fd, KVM_ENABLE_CAP, cap);
84 TEST_ASSERT(ret == 0, "KVM_ENABLE_CAP IOCTL failed,\n"
85 " rc: %i errno: %i", ret, errno);
90 /* VCPU Enable Capability
93 * vm - Virtual Machine
99 * Return: On success, 0. On failure a TEST_ASSERT failure is produced.
101 * Enables a capability (KVM_CAP_*) on the VCPU.
103 int vcpu_enable_cap(struct kvm_vm *vm, uint32_t vcpu_id,
104 struct kvm_enable_cap *cap)
106 struct vcpu *vcpu = vcpu_find(vm, vcpu_id);
109 TEST_ASSERT(vcpu, "cannot find vcpu %d", vcpu_id);
111 r = ioctl(vcpu->fd, KVM_ENABLE_CAP, cap);
112 TEST_ASSERT(!r, "KVM_ENABLE_CAP vCPU ioctl failed,\n"
113 " rc: %i, errno: %i", r, errno);
118 void vm_enable_dirty_ring(struct kvm_vm *vm, uint32_t ring_size)
120 struct kvm_enable_cap cap = { 0 };
122 cap.cap = KVM_CAP_DIRTY_LOG_RING;
123 cap.args[0] = ring_size;
124 vm_enable_cap(vm, &cap);
125 vm->dirty_ring_size = ring_size;
128 static void vm_open(struct kvm_vm *vm, int perm)
130 vm->kvm_fd = open(KVM_DEV_PATH, perm);
134 if (!kvm_check_cap(KVM_CAP_IMMEDIATE_EXIT)) {
135 print_skip("immediate_exit not available");
139 vm->fd = ioctl(vm->kvm_fd, KVM_CREATE_VM, vm->type);
140 TEST_ASSERT(vm->fd >= 0, "KVM_CREATE_VM ioctl failed, "
141 "rc: %i errno: %i", vm->fd, errno);
144 const char * const vm_guest_mode_string[] = {
145 "PA-bits:52, VA-bits:48, 4K pages",
146 "PA-bits:52, VA-bits:48, 64K pages",
147 "PA-bits:48, VA-bits:48, 4K pages",
148 "PA-bits:48, VA-bits:48, 64K pages",
149 "PA-bits:40, VA-bits:48, 4K pages",
150 "PA-bits:40, VA-bits:48, 64K pages",
151 "PA-bits:ANY, VA-bits:48, 4K pages",
153 _Static_assert(sizeof(vm_guest_mode_string)/sizeof(char *) == NUM_VM_MODES,
154 "Missing new mode strings?");
156 struct vm_guest_mode_params {
157 unsigned int pa_bits;
158 unsigned int va_bits;
159 unsigned int page_size;
160 unsigned int page_shift;
163 static const struct vm_guest_mode_params vm_guest_mode_params[] = {
164 { 52, 48, 0x1000, 12 },
165 { 52, 48, 0x10000, 16 },
166 { 48, 48, 0x1000, 12 },
167 { 48, 48, 0x10000, 16 },
168 { 40, 48, 0x1000, 12 },
169 { 40, 48, 0x10000, 16 },
170 { 0, 0, 0x1000, 12 },
172 _Static_assert(sizeof(vm_guest_mode_params)/sizeof(struct vm_guest_mode_params) == NUM_VM_MODES,
173 "Missing new mode params?");
179 * mode - VM Mode (e.g. VM_MODE_P52V48_4K)
180 * phy_pages - Physical memory pages
186 * Pointer to opaque structure that describes the created VM.
188 * Creates a VM with the mode specified by mode (e.g. VM_MODE_P52V48_4K).
189 * When phy_pages is non-zero, a memory region of phy_pages physical pages
190 * is created and mapped starting at guest physical address 0. The file
191 * descriptor to control the created VM is created with the permissions
192 * given by perm (e.g. O_RDWR).
194 struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
198 pr_debug("%s: mode='%s' pages='%ld' perm='%d'\n", __func__,
199 vm_guest_mode_string(mode), phy_pages, perm);
201 vm = calloc(1, sizeof(*vm));
202 TEST_ASSERT(vm != NULL, "Insufficient Memory");
204 INIT_LIST_HEAD(&vm->vcpus);
205 INIT_LIST_HEAD(&vm->userspace_mem_regions);
210 vm->pa_bits = vm_guest_mode_params[mode].pa_bits;
211 vm->va_bits = vm_guest_mode_params[mode].va_bits;
212 vm->page_size = vm_guest_mode_params[mode].page_size;
213 vm->page_shift = vm_guest_mode_params[mode].page_shift;
215 /* Setup mode specific traits. */
217 case VM_MODE_P52V48_4K:
218 vm->pgtable_levels = 4;
220 case VM_MODE_P52V48_64K:
221 vm->pgtable_levels = 3;
223 case VM_MODE_P48V48_4K:
224 vm->pgtable_levels = 4;
226 case VM_MODE_P48V48_64K:
227 vm->pgtable_levels = 3;
229 case VM_MODE_P40V48_4K:
230 vm->pgtable_levels = 4;
232 case VM_MODE_P40V48_64K:
233 vm->pgtable_levels = 3;
235 case VM_MODE_PXXV48_4K:
237 kvm_get_cpu_address_width(&vm->pa_bits, &vm->va_bits);
239 * Ignore KVM support for 5-level paging (vm->va_bits == 57),
240 * it doesn't take effect unless a CR4.LA57 is set, which it
241 * isn't for this VM_MODE.
243 TEST_ASSERT(vm->va_bits == 48 || vm->va_bits == 57,
244 "Linear address width (%d bits) not supported",
246 pr_debug("Guest physical address width detected: %d\n",
248 vm->pgtable_levels = 4;
251 TEST_FAIL("VM_MODE_PXXV48_4K not supported on non-x86 platforms");
255 TEST_FAIL("Unknown guest mode, mode: 0x%x", mode);
259 if (vm->pa_bits != 40)
260 vm->type = KVM_VM_TYPE_ARM_IPA_SIZE(vm->pa_bits);
265 /* Limit to VA-bit canonical virtual addresses. */
266 vm->vpages_valid = sparsebit_alloc();
267 sparsebit_set_num(vm->vpages_valid,
268 0, (1ULL << (vm->va_bits - 1)) >> vm->page_shift);
269 sparsebit_set_num(vm->vpages_valid,
270 (~((1ULL << (vm->va_bits - 1)) - 1)) >> vm->page_shift,
271 (1ULL << (vm->va_bits - 1)) >> vm->page_shift);
273 /* Limit physical addresses to PA-bits. */
274 vm->max_gfn = ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
276 /* Allocate and setup memory for guest. */
277 vm->vpages_mapped = sparsebit_alloc();
279 vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS,
285 struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
286 uint64_t extra_mem_pages, uint32_t num_percpu_pages,
287 void *guest_code, uint32_t vcpuids[])
289 /* The maximum page table size for a memory region will be when the
290 * smallest pages are used. Considering each page contains x page
291 * table descriptors, the total extra size for page tables (for extra
292 * N pages) will be: N/x+N/x^2+N/x^3+... which is definitely smaller
295 uint64_t vcpu_pages = (DEFAULT_STACK_PGS + num_percpu_pages) * nr_vcpus;
296 uint64_t extra_pg_pages = (extra_mem_pages + vcpu_pages) / PTES_PER_MIN_PAGE * 2;
297 uint64_t pages = DEFAULT_GUEST_PHY_PAGES + vcpu_pages + extra_pg_pages;
301 TEST_ASSERT(nr_vcpus <= kvm_check_cap(KVM_CAP_MAX_VCPUS),
302 "nr_vcpus = %d too large for host, max-vcpus = %d",
303 nr_vcpus, kvm_check_cap(KVM_CAP_MAX_VCPUS));
305 pages = vm_adjust_num_guest_pages(mode, pages);
306 vm = vm_create(mode, pages, O_RDWR);
308 kvm_vm_elf_load(vm, program_invocation_name, 0, 0);
311 vm_create_irqchip(vm);
314 for (i = 0; i < nr_vcpus; ++i) {
315 uint32_t vcpuid = vcpuids ? vcpuids[i] : i;
317 vm_vcpu_add_default(vm, vcpuid, guest_code);
320 vcpu_set_cpuid(vm, vcpuid, kvm_get_supported_cpuid());
327 struct kvm_vm *vm_create_default_with_vcpus(uint32_t nr_vcpus, uint64_t extra_mem_pages,
328 uint32_t num_percpu_pages, void *guest_code,
331 return vm_create_with_vcpus(VM_MODE_DEFAULT, nr_vcpus, extra_mem_pages,
332 num_percpu_pages, guest_code, vcpuids);
335 struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_pages,
338 return vm_create_default_with_vcpus(1, extra_mem_pages, 0, guest_code,
339 (uint32_t []){ vcpuid });
346 * vm - VM that has been released before
351 * Reopens the file descriptors associated to the VM and reinstates the
352 * global state, such as the irqchip and the memory regions that are mapped
355 void kvm_vm_restart(struct kvm_vm *vmp, int perm)
357 struct userspace_mem_region *region;
360 if (vmp->has_irqchip)
361 vm_create_irqchip(vmp);
363 list_for_each_entry(region, &vmp->userspace_mem_regions, list) {
364 int ret = ioctl(vmp->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
365 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
366 " rc: %i errno: %i\n"
367 " slot: %u flags: 0x%x\n"
368 " guest_phys_addr: 0x%llx size: 0x%llx",
369 ret, errno, region->region.slot,
370 region->region.flags,
371 region->region.guest_phys_addr,
372 region->region.memory_size);
376 void kvm_vm_get_dirty_log(struct kvm_vm *vm, int slot, void *log)
378 struct kvm_dirty_log args = { .dirty_bitmap = log, .slot = slot };
381 ret = ioctl(vm->fd, KVM_GET_DIRTY_LOG, &args);
382 TEST_ASSERT(ret == 0, "%s: KVM_GET_DIRTY_LOG failed: %s",
383 __func__, strerror(-ret));
386 void kvm_vm_clear_dirty_log(struct kvm_vm *vm, int slot, void *log,
387 uint64_t first_page, uint32_t num_pages)
389 struct kvm_clear_dirty_log args = { .dirty_bitmap = log, .slot = slot,
390 .first_page = first_page,
391 .num_pages = num_pages };
394 ret = ioctl(vm->fd, KVM_CLEAR_DIRTY_LOG, &args);
395 TEST_ASSERT(ret == 0, "%s: KVM_CLEAR_DIRTY_LOG failed: %s",
396 __func__, strerror(-ret));
399 uint32_t kvm_vm_reset_dirty_ring(struct kvm_vm *vm)
401 return ioctl(vm->fd, KVM_RESET_DIRTY_RINGS);
405 * Userspace Memory Region Find
408 * vm - Virtual Machine
409 * start - Starting VM physical address
410 * end - Ending VM physical address, inclusive.
415 * Pointer to overlapping region, NULL if no such region.
417 * Searches for a region with any physical memory that overlaps with
418 * any portion of the guest physical addresses from start to end
419 * inclusive. If multiple overlapping regions exist, a pointer to any
420 * of the regions is returned. Null is returned only when no overlapping
423 static struct userspace_mem_region *
424 userspace_mem_region_find(struct kvm_vm *vm, uint64_t start, uint64_t end)
426 struct userspace_mem_region *region;
428 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
429 uint64_t existing_start = region->region.guest_phys_addr;
430 uint64_t existing_end = region->region.guest_phys_addr
431 + region->region.memory_size - 1;
432 if (start <= existing_end && end >= existing_start)
440 * KVM Userspace Memory Region Find
443 * vm - Virtual Machine
444 * start - Starting VM physical address
445 * end - Ending VM physical address, inclusive.
450 * Pointer to overlapping region, NULL if no such region.
452 * Public interface to userspace_mem_region_find. Allows tests to look up
453 * the memslot datastructure for a given range of guest physical memory.
455 struct kvm_userspace_memory_region *
456 kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
459 struct userspace_mem_region *region;
461 region = userspace_mem_region_find(vm, start, end);
465 return ®ion->region;
472 * vm - Virtual Machine
478 * Pointer to VCPU structure
480 * Locates a vcpu structure that describes the VCPU specified by vcpuid and
481 * returns a pointer to it. Returns NULL if the VM doesn't contain a VCPU
482 * for the specified vcpuid.
484 struct vcpu *vcpu_find(struct kvm_vm *vm, uint32_t vcpuid)
488 list_for_each_entry(vcpu, &vm->vcpus, list) {
489 if (vcpu->id == vcpuid)
500 * vcpu - VCPU to remove
504 * Return: None, TEST_ASSERT failures for all error conditions
506 * Removes a vCPU from a VM and frees its resources.
508 static void vm_vcpu_rm(struct kvm_vm *vm, struct vcpu *vcpu)
512 if (vcpu->dirty_gfns) {
513 ret = munmap(vcpu->dirty_gfns, vm->dirty_ring_size);
514 TEST_ASSERT(ret == 0, "munmap of VCPU dirty ring failed, "
515 "rc: %i errno: %i", ret, errno);
516 vcpu->dirty_gfns = NULL;
519 ret = munmap(vcpu->state, sizeof(*vcpu->state));
520 TEST_ASSERT(ret == 0, "munmap of VCPU fd failed, rc: %i "
521 "errno: %i", ret, errno);
523 TEST_ASSERT(ret == 0, "Close of VCPU fd failed, rc: %i "
524 "errno: %i", ret, errno);
526 list_del(&vcpu->list);
530 void kvm_vm_release(struct kvm_vm *vmp)
532 struct vcpu *vcpu, *tmp;
535 list_for_each_entry_safe(vcpu, tmp, &vmp->vcpus, list)
536 vm_vcpu_rm(vmp, vcpu);
538 ret = close(vmp->fd);
539 TEST_ASSERT(ret == 0, "Close of vm fd failed,\n"
540 " vmp->fd: %i rc: %i errno: %i", vmp->fd, ret, errno);
543 TEST_ASSERT(ret == 0, "Close of /dev/kvm fd failed,\n"
544 " vmp->kvm_fd: %i rc: %i errno: %i", vmp->kvm_fd, ret, errno);
547 static void __vm_mem_region_delete(struct kvm_vm *vm,
548 struct userspace_mem_region *region)
552 list_del(®ion->list);
554 region->region.memory_size = 0;
555 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
556 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed, "
557 "rc: %i errno: %i", ret, errno);
559 sparsebit_free(®ion->unused_phy_pages);
560 ret = munmap(region->mmap_start, region->mmap_size);
561 TEST_ASSERT(ret == 0, "munmap failed, rc: %i errno: %i", ret, errno);
567 * Destroys and frees the VM pointed to by vmp.
569 void kvm_vm_free(struct kvm_vm *vmp)
571 struct userspace_mem_region *region, *tmp;
576 /* Free userspace_mem_regions. */
577 list_for_each_entry_safe(region, tmp, &vmp->userspace_mem_regions, list)
578 __vm_mem_region_delete(vmp, region);
580 /* Free sparsebit arrays. */
581 sparsebit_free(&vmp->vpages_valid);
582 sparsebit_free(&vmp->vpages_mapped);
586 /* Free the structure describing the VM. */
591 * Memory Compare, host virtual to guest virtual
594 * hva - Starting host virtual address
595 * vm - Virtual Machine
596 * gva - Starting guest virtual address
597 * len - number of bytes to compare
601 * Input/Output Args: None
604 * Returns 0 if the bytes starting at hva for a length of len
605 * are equal the guest virtual bytes starting at gva. Returns
606 * a value < 0, if bytes at hva are less than those at gva.
607 * Otherwise a value > 0 is returned.
609 * Compares the bytes starting at the host virtual address hva, for
610 * a length of len, to the guest bytes starting at the guest virtual
611 * address given by gva.
613 int kvm_memcmp_hva_gva(void *hva, struct kvm_vm *vm, vm_vaddr_t gva, size_t len)
618 * Compare a batch of bytes until either a match is found
619 * or all the bytes have been compared.
621 for (uintptr_t offset = 0; offset < len; offset += amt) {
622 uintptr_t ptr1 = (uintptr_t)hva + offset;
625 * Determine host address for guest virtual address
628 uintptr_t ptr2 = (uintptr_t)addr_gva2hva(vm, gva + offset);
631 * Determine amount to compare on this pass.
632 * Don't allow the comparsion to cross a page boundary.
635 if ((ptr1 >> vm->page_shift) != ((ptr1 + amt) >> vm->page_shift))
636 amt = vm->page_size - (ptr1 % vm->page_size);
637 if ((ptr2 >> vm->page_shift) != ((ptr2 + amt) >> vm->page_shift))
638 amt = vm->page_size - (ptr2 % vm->page_size);
640 assert((ptr1 >> vm->page_shift) == ((ptr1 + amt - 1) >> vm->page_shift));
641 assert((ptr2 >> vm->page_shift) == ((ptr2 + amt - 1) >> vm->page_shift));
644 * Perform the comparison. If there is a difference
645 * return that result to the caller, otherwise need
646 * to continue on looking for a mismatch.
648 int ret = memcmp((void *)ptr1, (void *)ptr2, amt);
654 * No mismatch found. Let the caller know the two memory
661 * VM Userspace Memory Region Add
664 * vm - Virtual Machine
665 * backing_src - Storage source for this region.
666 * NULL to use anonymous memory.
667 * guest_paddr - Starting guest physical address
668 * slot - KVM region slot
669 * npages - Number of physical pages
670 * flags - KVM memory region flags (e.g. KVM_MEM_LOG_DIRTY_PAGES)
676 * Allocates a memory area of the number of pages specified by npages
677 * and maps it to the VM specified by vm, at a starting physical address
678 * given by guest_paddr. The region is created with a KVM region slot
679 * given by slot, which must be unique and < KVM_MEM_SLOTS_NUM. The
680 * region is created with the flags given by flags.
682 void vm_userspace_mem_region_add(struct kvm_vm *vm,
683 enum vm_mem_backing_src_type src_type,
684 uint64_t guest_paddr, uint32_t slot, uint64_t npages,
688 struct userspace_mem_region *region;
689 size_t huge_page_size = KVM_UTIL_PGS_PER_HUGEPG * vm->page_size;
692 TEST_ASSERT(vm_adjust_num_guest_pages(vm->mode, npages) == npages,
693 "Number of guest pages is not compatible with the host. "
694 "Try npages=%d", vm_adjust_num_guest_pages(vm->mode, npages));
696 TEST_ASSERT((guest_paddr % vm->page_size) == 0, "Guest physical "
697 "address not on a page boundary.\n"
698 " guest_paddr: 0x%lx vm->page_size: 0x%x",
699 guest_paddr, vm->page_size);
700 TEST_ASSERT((((guest_paddr >> vm->page_shift) + npages) - 1)
701 <= vm->max_gfn, "Physical range beyond maximum "
702 "supported physical address,\n"
703 " guest_paddr: 0x%lx npages: 0x%lx\n"
704 " vm->max_gfn: 0x%lx vm->page_size: 0x%x",
705 guest_paddr, npages, vm->max_gfn, vm->page_size);
708 * Confirm a mem region with an overlapping address doesn't
711 region = (struct userspace_mem_region *) userspace_mem_region_find(
712 vm, guest_paddr, (guest_paddr + npages * vm->page_size) - 1);
714 TEST_FAIL("overlapping userspace_mem_region already "
716 " requested guest_paddr: 0x%lx npages: 0x%lx "
718 " existing guest_paddr: 0x%lx size: 0x%lx",
719 guest_paddr, npages, vm->page_size,
720 (uint64_t) region->region.guest_phys_addr,
721 (uint64_t) region->region.memory_size);
723 /* Confirm no region with the requested slot already exists. */
724 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
725 if (region->region.slot != slot)
728 TEST_FAIL("A mem region with the requested slot "
730 " requested slot: %u paddr: 0x%lx npages: 0x%lx\n"
731 " existing slot: %u paddr: 0x%lx size: 0x%lx",
732 slot, guest_paddr, npages,
734 (uint64_t) region->region.guest_phys_addr,
735 (uint64_t) region->region.memory_size);
738 /* Allocate and initialize new mem region structure. */
739 region = calloc(1, sizeof(*region));
740 TEST_ASSERT(region != NULL, "Insufficient Memory");
741 region->mmap_size = npages * vm->page_size;
744 /* On s390x, the host address must be aligned to 1M (due to PGSTEs) */
745 alignment = 0x100000;
750 if (src_type == VM_MEM_SRC_ANONYMOUS_THP)
751 alignment = max(huge_page_size, alignment);
753 /* Add enough memory to align up if necessary */
755 region->mmap_size += alignment;
757 region->mmap_start = mmap(NULL, region->mmap_size,
758 PROT_READ | PROT_WRITE,
759 MAP_PRIVATE | MAP_ANONYMOUS
760 | (src_type == VM_MEM_SRC_ANONYMOUS_HUGETLB ? MAP_HUGETLB : 0),
762 TEST_ASSERT(region->mmap_start != MAP_FAILED,
763 "test_malloc failed, mmap_start: %p errno: %i",
764 region->mmap_start, errno);
766 /* Align host address */
767 region->host_mem = align(region->mmap_start, alignment);
769 /* As needed perform madvise */
770 if (src_type == VM_MEM_SRC_ANONYMOUS || src_type == VM_MEM_SRC_ANONYMOUS_THP) {
773 ret = stat("/sys/kernel/mm/transparent_hugepage", &statbuf);
774 TEST_ASSERT(ret == 0 || (ret == -1 && errno == ENOENT),
775 "stat /sys/kernel/mm/transparent_hugepage");
777 TEST_ASSERT(ret == 0 || src_type != VM_MEM_SRC_ANONYMOUS_THP,
778 "VM_MEM_SRC_ANONYMOUS_THP requires THP to be configured in the host kernel");
781 ret = madvise(region->host_mem, npages * vm->page_size,
782 src_type == VM_MEM_SRC_ANONYMOUS ? MADV_NOHUGEPAGE : MADV_HUGEPAGE);
783 TEST_ASSERT(ret == 0, "madvise failed, addr: %p length: 0x%lx src_type: %x",
784 region->host_mem, npages * vm->page_size, src_type);
788 region->unused_phy_pages = sparsebit_alloc();
789 sparsebit_set_num(region->unused_phy_pages,
790 guest_paddr >> vm->page_shift, npages);
791 region->region.slot = slot;
792 region->region.flags = flags;
793 region->region.guest_phys_addr = guest_paddr;
794 region->region.memory_size = npages * vm->page_size;
795 region->region.userspace_addr = (uintptr_t) region->host_mem;
796 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
797 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
798 " rc: %i errno: %i\n"
799 " slot: %u flags: 0x%x\n"
800 " guest_phys_addr: 0x%lx size: 0x%lx",
801 ret, errno, slot, flags,
802 guest_paddr, (uint64_t) region->region.memory_size);
804 /* Add to linked-list of memory regions. */
805 list_add(®ion->list, &vm->userspace_mem_regions);
812 * vm - Virtual Machine
813 * memslot - KVM memory slot ID
818 * Pointer to memory region structure that describe memory region
819 * using kvm memory slot ID given by memslot. TEST_ASSERT failure
820 * on error (e.g. currently no memory region using memslot as a KVM
823 struct userspace_mem_region *
824 memslot2region(struct kvm_vm *vm, uint32_t memslot)
826 struct userspace_mem_region *region;
828 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
829 if (region->region.slot == memslot)
833 fprintf(stderr, "No mem region with the requested slot found,\n"
834 " requested slot: %u\n", memslot);
835 fputs("---- vm dump ----\n", stderr);
836 vm_dump(stderr, vm, 2);
837 TEST_FAIL("Mem region not found");
842 * VM Memory Region Flags Set
845 * vm - Virtual Machine
846 * flags - Starting guest physical address
852 * Sets the flags of the memory region specified by the value of slot,
853 * to the values given by flags.
855 void vm_mem_region_set_flags(struct kvm_vm *vm, uint32_t slot, uint32_t flags)
858 struct userspace_mem_region *region;
860 region = memslot2region(vm, slot);
862 region->region.flags = flags;
864 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
866 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
867 " rc: %i errno: %i slot: %u flags: 0x%x",
868 ret, errno, slot, flags);
872 * VM Memory Region Move
875 * vm - Virtual Machine
876 * slot - Slot of the memory region to move
877 * new_gpa - Starting guest physical address
883 * Change the gpa of a memory region.
885 void vm_mem_region_move(struct kvm_vm *vm, uint32_t slot, uint64_t new_gpa)
887 struct userspace_mem_region *region;
890 region = memslot2region(vm, slot);
892 region->region.guest_phys_addr = new_gpa;
894 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
896 TEST_ASSERT(!ret, "KVM_SET_USER_MEMORY_REGION failed\n"
897 "ret: %i errno: %i slot: %u new_gpa: 0x%lx",
898 ret, errno, slot, new_gpa);
902 * VM Memory Region Delete
905 * vm - Virtual Machine
906 * slot - Slot of the memory region to delete
912 * Delete a memory region.
914 void vm_mem_region_delete(struct kvm_vm *vm, uint32_t slot)
916 __vm_mem_region_delete(vm, memslot2region(vm, slot));
929 * Returns the size of the structure pointed to by the return value
932 static int vcpu_mmap_sz(void)
936 dev_fd = open(KVM_DEV_PATH, O_RDONLY);
940 ret = ioctl(dev_fd, KVM_GET_VCPU_MMAP_SIZE, NULL);
941 TEST_ASSERT(ret >= sizeof(struct kvm_run),
942 "%s KVM_GET_VCPU_MMAP_SIZE ioctl failed, rc: %i errno: %i",
943 __func__, ret, errno);
954 * vm - Virtual Machine
961 * Adds a virtual CPU to the VM specified by vm with the ID given by vcpuid.
962 * No additional VCPU setup is done.
964 void vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpuid)
968 /* Confirm a vcpu with the specified id doesn't already exist. */
969 vcpu = vcpu_find(vm, vcpuid);
971 TEST_FAIL("vcpu with the specified id "
973 " requested vcpuid: %u\n"
974 " existing vcpuid: %u state: %p",
975 vcpuid, vcpu->id, vcpu->state);
977 /* Allocate and initialize new vcpu structure. */
978 vcpu = calloc(1, sizeof(*vcpu));
979 TEST_ASSERT(vcpu != NULL, "Insufficient Memory");
981 vcpu->fd = ioctl(vm->fd, KVM_CREATE_VCPU, vcpuid);
982 TEST_ASSERT(vcpu->fd >= 0, "KVM_CREATE_VCPU failed, rc: %i errno: %i",
985 TEST_ASSERT(vcpu_mmap_sz() >= sizeof(*vcpu->state), "vcpu mmap size "
986 "smaller than expected, vcpu_mmap_sz: %i expected_min: %zi",
987 vcpu_mmap_sz(), sizeof(*vcpu->state));
988 vcpu->state = (struct kvm_run *) mmap(NULL, sizeof(*vcpu->state),
989 PROT_READ | PROT_WRITE, MAP_SHARED, vcpu->fd, 0);
990 TEST_ASSERT(vcpu->state != MAP_FAILED, "mmap vcpu_state failed, "
991 "vcpu id: %u errno: %i", vcpuid, errno);
993 /* Add to linked-list of VCPUs. */
994 list_add(&vcpu->list, &vm->vcpus);
998 * VM Virtual Address Unused Gap
1001 * vm - Virtual Machine
1003 * vaddr_min - Minimum Virtual Address
1008 * Lowest virtual address at or below vaddr_min, with at least
1009 * sz unused bytes. TEST_ASSERT failure if no area of at least
1010 * size sz is available.
1012 * Within the VM specified by vm, locates the lowest starting virtual
1013 * address >= vaddr_min, that has at least sz unallocated bytes. A
1014 * TEST_ASSERT failure occurs for invalid input or no area of at least
1015 * sz unallocated bytes >= vaddr_min is available.
1017 static vm_vaddr_t vm_vaddr_unused_gap(struct kvm_vm *vm, size_t sz,
1018 vm_vaddr_t vaddr_min)
1020 uint64_t pages = (sz + vm->page_size - 1) >> vm->page_shift;
1022 /* Determine lowest permitted virtual page index. */
1023 uint64_t pgidx_start = (vaddr_min + vm->page_size - 1) >> vm->page_shift;
1024 if ((pgidx_start * vm->page_size) < vaddr_min)
1027 /* Loop over section with enough valid virtual page indexes. */
1028 if (!sparsebit_is_set_num(vm->vpages_valid,
1029 pgidx_start, pages))
1030 pgidx_start = sparsebit_next_set_num(vm->vpages_valid,
1031 pgidx_start, pages);
1034 * Are there enough unused virtual pages available at
1035 * the currently proposed starting virtual page index.
1036 * If not, adjust proposed starting index to next
1039 if (sparsebit_is_clear_num(vm->vpages_mapped,
1040 pgidx_start, pages))
1042 pgidx_start = sparsebit_next_clear_num(vm->vpages_mapped,
1043 pgidx_start, pages);
1044 if (pgidx_start == 0)
1048 * If needed, adjust proposed starting virtual address,
1049 * to next range of valid virtual addresses.
1051 if (!sparsebit_is_set_num(vm->vpages_valid,
1052 pgidx_start, pages)) {
1053 pgidx_start = sparsebit_next_set_num(
1054 vm->vpages_valid, pgidx_start, pages);
1055 if (pgidx_start == 0)
1058 } while (pgidx_start != 0);
1061 TEST_FAIL("No vaddr of specified pages available, pages: 0x%lx", pages);
1067 TEST_ASSERT(sparsebit_is_set_num(vm->vpages_valid,
1068 pgidx_start, pages),
1069 "Unexpected, invalid virtual page index range,\n"
1070 " pgidx_start: 0x%lx\n"
1072 pgidx_start, pages);
1073 TEST_ASSERT(sparsebit_is_clear_num(vm->vpages_mapped,
1074 pgidx_start, pages),
1075 "Unexpected, pages already mapped,\n"
1076 " pgidx_start: 0x%lx\n"
1078 pgidx_start, pages);
1080 return pgidx_start * vm->page_size;
1084 * VM Virtual Address Allocate
1087 * vm - Virtual Machine
1088 * sz - Size in bytes
1089 * vaddr_min - Minimum starting virtual address
1090 * data_memslot - Memory region slot for data pages
1091 * pgd_memslot - Memory region slot for new virtual translation tables
1096 * Starting guest virtual address
1098 * Allocates at least sz bytes within the virtual address space of the vm
1099 * given by vm. The allocated bytes are mapped to a virtual address >=
1100 * the address given by vaddr_min. Note that each allocation uses a
1101 * a unique set of pages, with the minimum real allocation being at least
1104 vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
1105 uint32_t data_memslot, uint32_t pgd_memslot)
1107 uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
1109 virt_pgd_alloc(vm, pgd_memslot);
1112 * Find an unused range of virtual page addresses of at least
1115 vm_vaddr_t vaddr_start = vm_vaddr_unused_gap(vm, sz, vaddr_min);
1117 /* Map the virtual pages. */
1118 for (vm_vaddr_t vaddr = vaddr_start; pages > 0;
1119 pages--, vaddr += vm->page_size) {
1122 paddr = vm_phy_page_alloc(vm,
1123 KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
1125 virt_pg_map(vm, vaddr, paddr, pgd_memslot);
1127 sparsebit_set(vm->vpages_mapped,
1128 vaddr >> vm->page_shift);
1135 * Map a range of VM virtual address to the VM's physical address
1138 * vm - Virtual Machine
1139 * vaddr - Virtuall address to map
1140 * paddr - VM Physical Address
1141 * npages - The number of pages to map
1142 * pgd_memslot - Memory region slot for new virtual translation tables
1148 * Within the VM given by @vm, creates a virtual translation for
1149 * @npages starting at @vaddr to the page range starting at @paddr.
1151 void virt_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr,
1152 unsigned int npages, uint32_t pgd_memslot)
1154 size_t page_size = vm->page_size;
1155 size_t size = npages * page_size;
1157 TEST_ASSERT(vaddr + size > vaddr, "Vaddr overflow");
1158 TEST_ASSERT(paddr + size > paddr, "Paddr overflow");
1161 virt_pg_map(vm, vaddr, paddr, pgd_memslot);
1168 * Address VM Physical to Host Virtual
1171 * vm - Virtual Machine
1172 * gpa - VM physical address
1177 * Equivalent host virtual address
1179 * Locates the memory region containing the VM physical address given
1180 * by gpa, within the VM given by vm. When found, the host virtual
1181 * address providing the memory to the vm physical address is returned.
1182 * A TEST_ASSERT failure occurs if no region containing gpa exists.
1184 void *addr_gpa2hva(struct kvm_vm *vm, vm_paddr_t gpa)
1186 struct userspace_mem_region *region;
1188 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1189 if ((gpa >= region->region.guest_phys_addr)
1190 && (gpa <= (region->region.guest_phys_addr
1191 + region->region.memory_size - 1)))
1192 return (void *) ((uintptr_t) region->host_mem
1193 + (gpa - region->region.guest_phys_addr));
1196 TEST_FAIL("No vm physical memory at 0x%lx", gpa);
1201 * Address Host Virtual to VM Physical
1204 * vm - Virtual Machine
1205 * hva - Host virtual address
1210 * Equivalent VM physical address
1212 * Locates the memory region containing the host virtual address given
1213 * by hva, within the VM given by vm. When found, the equivalent
1214 * VM physical address is returned. A TEST_ASSERT failure occurs if no
1215 * region containing hva exists.
1217 vm_paddr_t addr_hva2gpa(struct kvm_vm *vm, void *hva)
1219 struct userspace_mem_region *region;
1221 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1222 if ((hva >= region->host_mem)
1223 && (hva <= (region->host_mem
1224 + region->region.memory_size - 1)))
1225 return (vm_paddr_t) ((uintptr_t)
1226 region->region.guest_phys_addr
1227 + (hva - (uintptr_t) region->host_mem));
1230 TEST_FAIL("No mapping to a guest physical address, hva: %p", hva);
1235 * VM Create IRQ Chip
1238 * vm - Virtual Machine
1244 * Creates an interrupt controller chip for the VM specified by vm.
1246 void vm_create_irqchip(struct kvm_vm *vm)
1250 ret = ioctl(vm->fd, KVM_CREATE_IRQCHIP, 0);
1251 TEST_ASSERT(ret == 0, "KVM_CREATE_IRQCHIP IOCTL failed, "
1252 "rc: %i errno: %i", ret, errno);
1254 vm->has_irqchip = true;
1261 * vm - Virtual Machine
1267 * Pointer to structure that describes the state of the VCPU.
1269 * Locates and returns a pointer to a structure that describes the
1270 * state of the VCPU with the given vcpuid.
1272 struct kvm_run *vcpu_state(struct kvm_vm *vm, uint32_t vcpuid)
1274 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1275 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1284 * vm - Virtual Machine
1291 * Switch to executing the code for the VCPU given by vcpuid, within the VM
1294 void vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1296 int ret = _vcpu_run(vm, vcpuid);
1297 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1298 "rc: %i errno: %i", ret, errno);
1301 int _vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1303 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1306 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1308 rc = ioctl(vcpu->fd, KVM_RUN, NULL);
1309 } while (rc == -1 && errno == EINTR);
1311 assert_on_unhandled_exception(vm, vcpuid);
1316 int vcpu_get_fd(struct kvm_vm *vm, uint32_t vcpuid)
1318 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1320 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1325 void vcpu_run_complete_io(struct kvm_vm *vm, uint32_t vcpuid)
1327 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1330 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1332 vcpu->state->immediate_exit = 1;
1333 ret = ioctl(vcpu->fd, KVM_RUN, NULL);
1334 vcpu->state->immediate_exit = 0;
1336 TEST_ASSERT(ret == -1 && errno == EINTR,
1337 "KVM_RUN IOCTL didn't exit immediately, rc: %i, errno: %i",
1341 void vcpu_set_guest_debug(struct kvm_vm *vm, uint32_t vcpuid,
1342 struct kvm_guest_debug *debug)
1344 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1345 int ret = ioctl(vcpu->fd, KVM_SET_GUEST_DEBUG, debug);
1347 TEST_ASSERT(ret == 0, "KVM_SET_GUEST_DEBUG failed: %d", ret);
1351 * VM VCPU Set MP State
1354 * vm - Virtual Machine
1356 * mp_state - mp_state to be set
1362 * Sets the MP state of the VCPU given by vcpuid, to the state given
1365 void vcpu_set_mp_state(struct kvm_vm *vm, uint32_t vcpuid,
1366 struct kvm_mp_state *mp_state)
1368 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1371 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1373 ret = ioctl(vcpu->fd, KVM_SET_MP_STATE, mp_state);
1374 TEST_ASSERT(ret == 0, "KVM_SET_MP_STATE IOCTL failed, "
1375 "rc: %i errno: %i", ret, errno);
1379 * VM VCPU Get Reg List
1382 * vm - Virtual Machine
1389 * A pointer to an allocated struct kvm_reg_list
1391 * Get the list of guest registers which are supported for
1392 * KVM_GET_ONE_REG/KVM_SET_ONE_REG calls
1394 struct kvm_reg_list *vcpu_get_reg_list(struct kvm_vm *vm, uint32_t vcpuid)
1396 struct kvm_reg_list reg_list_n = { .n = 0 }, *reg_list;
1399 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_REG_LIST, ®_list_n);
1400 TEST_ASSERT(ret == -1 && errno == E2BIG, "KVM_GET_REG_LIST n=0");
1401 reg_list = calloc(1, sizeof(*reg_list) + reg_list_n.n * sizeof(__u64));
1402 reg_list->n = reg_list_n.n;
1403 vcpu_ioctl(vm, vcpuid, KVM_GET_REG_LIST, reg_list);
1411 * vm - Virtual Machine
1415 * regs - current state of VCPU regs
1419 * Obtains the current register state for the VCPU specified by vcpuid
1420 * and stores it at the location given by regs.
1422 void vcpu_regs_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_regs *regs)
1424 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1427 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1429 ret = ioctl(vcpu->fd, KVM_GET_REGS, regs);
1430 TEST_ASSERT(ret == 0, "KVM_GET_REGS failed, rc: %i errno: %i",
1438 * vm - Virtual Machine
1440 * regs - Values to set VCPU regs to
1446 * Sets the regs of the VCPU specified by vcpuid to the values
1449 void vcpu_regs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_regs *regs)
1451 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1454 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1456 ret = ioctl(vcpu->fd, KVM_SET_REGS, regs);
1457 TEST_ASSERT(ret == 0, "KVM_SET_REGS failed, rc: %i errno: %i",
1461 #ifdef __KVM_HAVE_VCPU_EVENTS
1462 void vcpu_events_get(struct kvm_vm *vm, uint32_t vcpuid,
1463 struct kvm_vcpu_events *events)
1465 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1468 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1470 ret = ioctl(vcpu->fd, KVM_GET_VCPU_EVENTS, events);
1471 TEST_ASSERT(ret == 0, "KVM_GET_VCPU_EVENTS, failed, rc: %i errno: %i",
1475 void vcpu_events_set(struct kvm_vm *vm, uint32_t vcpuid,
1476 struct kvm_vcpu_events *events)
1478 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1481 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1483 ret = ioctl(vcpu->fd, KVM_SET_VCPU_EVENTS, events);
1484 TEST_ASSERT(ret == 0, "KVM_SET_VCPU_EVENTS, failed, rc: %i errno: %i",
1490 void vcpu_nested_state_get(struct kvm_vm *vm, uint32_t vcpuid,
1491 struct kvm_nested_state *state)
1493 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1496 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1498 ret = ioctl(vcpu->fd, KVM_GET_NESTED_STATE, state);
1499 TEST_ASSERT(ret == 0,
1500 "KVM_SET_NESTED_STATE failed, ret: %i errno: %i",
1504 int vcpu_nested_state_set(struct kvm_vm *vm, uint32_t vcpuid,
1505 struct kvm_nested_state *state, bool ignore_error)
1507 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1510 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1512 ret = ioctl(vcpu->fd, KVM_SET_NESTED_STATE, state);
1513 if (!ignore_error) {
1514 TEST_ASSERT(ret == 0,
1515 "KVM_SET_NESTED_STATE failed, ret: %i errno: %i",
1524 * VM VCPU System Regs Get
1527 * vm - Virtual Machine
1531 * sregs - current state of VCPU system regs
1535 * Obtains the current system register state for the VCPU specified by
1536 * vcpuid and stores it at the location given by sregs.
1538 void vcpu_sregs_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1540 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1543 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1545 ret = ioctl(vcpu->fd, KVM_GET_SREGS, sregs);
1546 TEST_ASSERT(ret == 0, "KVM_GET_SREGS failed, rc: %i errno: %i",
1551 * VM VCPU System Regs Set
1554 * vm - Virtual Machine
1556 * sregs - Values to set VCPU system regs to
1562 * Sets the system regs of the VCPU specified by vcpuid to the values
1565 void vcpu_sregs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1567 int ret = _vcpu_sregs_set(vm, vcpuid, sregs);
1568 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1569 "rc: %i errno: %i", ret, errno);
1572 int _vcpu_sregs_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_sregs *sregs)
1574 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1576 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1578 return ioctl(vcpu->fd, KVM_SET_SREGS, sregs);
1581 void vcpu_fpu_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu)
1585 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_FPU, fpu);
1586 TEST_ASSERT(ret == 0, "KVM_GET_FPU failed, rc: %i errno: %i (%s)",
1587 ret, errno, strerror(errno));
1590 void vcpu_fpu_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu)
1594 ret = _vcpu_ioctl(vm, vcpuid, KVM_SET_FPU, fpu);
1595 TEST_ASSERT(ret == 0, "KVM_SET_FPU failed, rc: %i errno: %i (%s)",
1596 ret, errno, strerror(errno));
1599 void vcpu_get_reg(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_one_reg *reg)
1603 ret = _vcpu_ioctl(vm, vcpuid, KVM_GET_ONE_REG, reg);
1604 TEST_ASSERT(ret == 0, "KVM_GET_ONE_REG failed, rc: %i errno: %i (%s)",
1605 ret, errno, strerror(errno));
1608 void vcpu_set_reg(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_one_reg *reg)
1612 ret = _vcpu_ioctl(vm, vcpuid, KVM_SET_ONE_REG, reg);
1613 TEST_ASSERT(ret == 0, "KVM_SET_ONE_REG failed, rc: %i errno: %i (%s)",
1614 ret, errno, strerror(errno));
1621 * vm - Virtual Machine
1623 * cmd - Ioctl number
1624 * arg - Argument to pass to the ioctl
1628 * Issues an arbitrary ioctl on a VCPU fd.
1630 void vcpu_ioctl(struct kvm_vm *vm, uint32_t vcpuid,
1631 unsigned long cmd, void *arg)
1635 ret = _vcpu_ioctl(vm, vcpuid, cmd, arg);
1636 TEST_ASSERT(ret == 0, "vcpu ioctl %lu failed, rc: %i errno: %i (%s)",
1637 cmd, ret, errno, strerror(errno));
1640 int _vcpu_ioctl(struct kvm_vm *vm, uint32_t vcpuid,
1641 unsigned long cmd, void *arg)
1643 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1646 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1648 ret = ioctl(vcpu->fd, cmd, arg);
1653 void *vcpu_map_dirty_ring(struct kvm_vm *vm, uint32_t vcpuid)
1656 uint32_t size = vm->dirty_ring_size;
1658 TEST_ASSERT(size > 0, "Should enable dirty ring first");
1660 vcpu = vcpu_find(vm, vcpuid);
1662 TEST_ASSERT(vcpu, "Cannot find vcpu %u", vcpuid);
1664 if (!vcpu->dirty_gfns) {
1667 addr = mmap(NULL, size, PROT_READ,
1668 MAP_PRIVATE, vcpu->fd,
1669 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1670 TEST_ASSERT(addr == MAP_FAILED, "Dirty ring mapped private");
1672 addr = mmap(NULL, size, PROT_READ | PROT_EXEC,
1673 MAP_PRIVATE, vcpu->fd,
1674 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1675 TEST_ASSERT(addr == MAP_FAILED, "Dirty ring mapped exec");
1677 addr = mmap(NULL, size, PROT_READ | PROT_WRITE,
1678 MAP_SHARED, vcpu->fd,
1679 vm->page_size * KVM_DIRTY_LOG_PAGE_OFFSET);
1680 TEST_ASSERT(addr != MAP_FAILED, "Dirty ring map failed");
1682 vcpu->dirty_gfns = addr;
1683 vcpu->dirty_gfns_count = size / sizeof(struct kvm_dirty_gfn);
1686 return vcpu->dirty_gfns;
1693 * vm - Virtual Machine
1694 * cmd - Ioctl number
1695 * arg - Argument to pass to the ioctl
1699 * Issues an arbitrary ioctl on a VM fd.
1701 void vm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1705 ret = ioctl(vm->fd, cmd, arg);
1706 TEST_ASSERT(ret == 0, "vm ioctl %lu failed, rc: %i errno: %i (%s)",
1707 cmd, ret, errno, strerror(errno));
1714 * vm - Virtual Machine
1715 * cmd - Ioctl number
1716 * arg - Argument to pass to the ioctl
1720 * Issues an arbitrary ioctl on a KVM fd.
1722 void kvm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1726 ret = ioctl(vm->kvm_fd, cmd, arg);
1727 TEST_ASSERT(ret == 0, "KVM ioctl %lu failed, rc: %i errno: %i (%s)",
1728 cmd, ret, errno, strerror(errno));
1731 int _kvm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1733 return ioctl(vm->kvm_fd, cmd, arg);
1740 * vm - Virtual Machine
1741 * indent - Left margin indent amount
1744 * stream - Output FILE stream
1748 * Dumps the current state of the VM given by vm, to the FILE stream
1751 void vm_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
1753 struct userspace_mem_region *region;
1756 fprintf(stream, "%*smode: 0x%x\n", indent, "", vm->mode);
1757 fprintf(stream, "%*sfd: %i\n", indent, "", vm->fd);
1758 fprintf(stream, "%*spage_size: 0x%x\n", indent, "", vm->page_size);
1759 fprintf(stream, "%*sMem Regions:\n", indent, "");
1760 list_for_each_entry(region, &vm->userspace_mem_regions, list) {
1761 fprintf(stream, "%*sguest_phys: 0x%lx size: 0x%lx "
1762 "host_virt: %p\n", indent + 2, "",
1763 (uint64_t) region->region.guest_phys_addr,
1764 (uint64_t) region->region.memory_size,
1766 fprintf(stream, "%*sunused_phy_pages: ", indent + 2, "");
1767 sparsebit_dump(stream, region->unused_phy_pages, 0);
1769 fprintf(stream, "%*sMapped Virtual Pages:\n", indent, "");
1770 sparsebit_dump(stream, vm->vpages_mapped, indent + 2);
1771 fprintf(stream, "%*spgd_created: %u\n", indent, "",
1773 if (vm->pgd_created) {
1774 fprintf(stream, "%*sVirtual Translation Tables:\n",
1776 virt_dump(stream, vm, indent + 4);
1778 fprintf(stream, "%*sVCPUs:\n", indent, "");
1779 list_for_each_entry(vcpu, &vm->vcpus, list)
1780 vcpu_dump(stream, vm, vcpu->id, indent + 2);
1783 /* Known KVM exit reasons */
1784 static struct exit_reason {
1785 unsigned int reason;
1787 } exit_reasons_known[] = {
1788 {KVM_EXIT_UNKNOWN, "UNKNOWN"},
1789 {KVM_EXIT_EXCEPTION, "EXCEPTION"},
1790 {KVM_EXIT_IO, "IO"},
1791 {KVM_EXIT_HYPERCALL, "HYPERCALL"},
1792 {KVM_EXIT_DEBUG, "DEBUG"},
1793 {KVM_EXIT_HLT, "HLT"},
1794 {KVM_EXIT_MMIO, "MMIO"},
1795 {KVM_EXIT_IRQ_WINDOW_OPEN, "IRQ_WINDOW_OPEN"},
1796 {KVM_EXIT_SHUTDOWN, "SHUTDOWN"},
1797 {KVM_EXIT_FAIL_ENTRY, "FAIL_ENTRY"},
1798 {KVM_EXIT_INTR, "INTR"},
1799 {KVM_EXIT_SET_TPR, "SET_TPR"},
1800 {KVM_EXIT_TPR_ACCESS, "TPR_ACCESS"},
1801 {KVM_EXIT_S390_SIEIC, "S390_SIEIC"},
1802 {KVM_EXIT_S390_RESET, "S390_RESET"},
1803 {KVM_EXIT_DCR, "DCR"},
1804 {KVM_EXIT_NMI, "NMI"},
1805 {KVM_EXIT_INTERNAL_ERROR, "INTERNAL_ERROR"},
1806 {KVM_EXIT_OSI, "OSI"},
1807 {KVM_EXIT_PAPR_HCALL, "PAPR_HCALL"},
1808 {KVM_EXIT_DIRTY_RING_FULL, "DIRTY_RING_FULL"},
1809 {KVM_EXIT_X86_RDMSR, "RDMSR"},
1810 {KVM_EXIT_X86_WRMSR, "WRMSR"},
1811 #ifdef KVM_EXIT_MEMORY_NOT_PRESENT
1812 {KVM_EXIT_MEMORY_NOT_PRESENT, "MEMORY_NOT_PRESENT"},
1817 * Exit Reason String
1820 * exit_reason - Exit reason
1825 * Constant string pointer describing the exit reason.
1827 * Locates and returns a constant string that describes the KVM exit
1828 * reason given by exit_reason. If no such string is found, a constant
1829 * string of "Unknown" is returned.
1831 const char *exit_reason_str(unsigned int exit_reason)
1835 for (n1 = 0; n1 < ARRAY_SIZE(exit_reasons_known); n1++) {
1836 if (exit_reason == exit_reasons_known[n1].reason)
1837 return exit_reasons_known[n1].name;
1844 * Physical Contiguous Page Allocator
1847 * vm - Virtual Machine
1848 * num - number of pages
1849 * paddr_min - Physical address minimum
1850 * memslot - Memory region to allocate page from
1855 * Starting physical address
1857 * Within the VM specified by vm, locates a range of available physical
1858 * pages at or above paddr_min. If found, the pages are marked as in use
1859 * and their base address is returned. A TEST_ASSERT failure occurs if
1860 * not enough pages are available at or above paddr_min.
1862 vm_paddr_t vm_phy_pages_alloc(struct kvm_vm *vm, size_t num,
1863 vm_paddr_t paddr_min, uint32_t memslot)
1865 struct userspace_mem_region *region;
1866 sparsebit_idx_t pg, base;
1868 TEST_ASSERT(num > 0, "Must allocate at least one page");
1870 TEST_ASSERT((paddr_min % vm->page_size) == 0, "Min physical address "
1871 "not divisible by page size.\n"
1872 " paddr_min: 0x%lx page_size: 0x%x",
1873 paddr_min, vm->page_size);
1875 region = memslot2region(vm, memslot);
1876 base = pg = paddr_min >> vm->page_shift;
1879 for (; pg < base + num; ++pg) {
1880 if (!sparsebit_is_set(region->unused_phy_pages, pg)) {
1881 base = pg = sparsebit_next_set(region->unused_phy_pages, pg);
1885 } while (pg && pg != base + num);
1888 fprintf(stderr, "No guest physical page available, "
1889 "paddr_min: 0x%lx page_size: 0x%x memslot: %u\n",
1890 paddr_min, vm->page_size, memslot);
1891 fputs("---- vm dump ----\n", stderr);
1892 vm_dump(stderr, vm, 2);
1896 for (pg = base; pg < base + num; ++pg)
1897 sparsebit_clear(region->unused_phy_pages, pg);
1899 return base * vm->page_size;
1902 vm_paddr_t vm_phy_page_alloc(struct kvm_vm *vm, vm_paddr_t paddr_min,
1905 return vm_phy_pages_alloc(vm, 1, paddr_min, memslot);
1909 * Address Guest Virtual to Host Virtual
1912 * vm - Virtual Machine
1913 * gva - VM virtual address
1918 * Equivalent host virtual address
1920 void *addr_gva2hva(struct kvm_vm *vm, vm_vaddr_t gva)
1922 return addr_gpa2hva(vm, addr_gva2gpa(vm, gva));
1926 * Is Unrestricted Guest
1929 * vm - Virtual Machine
1933 * Return: True if the unrestricted guest is set to 'Y', otherwise return false.
1935 * Check if the unrestricted guest flag is enabled.
1937 bool vm_is_unrestricted_guest(struct kvm_vm *vm)
1944 /* Ensure that the KVM vendor-specific module is loaded. */
1945 f = fopen(KVM_DEV_PATH, "r");
1946 TEST_ASSERT(f != NULL, "Error in opening KVM dev file: %d",
1951 f = fopen("/sys/module/kvm_intel/parameters/unrestricted_guest", "r");
1953 count = fread(&val, sizeof(char), 1, f);
1954 TEST_ASSERT(count == 1, "Unable to read from param file.");
1961 unsigned int vm_get_page_size(struct kvm_vm *vm)
1963 return vm->page_size;
1966 unsigned int vm_get_page_shift(struct kvm_vm *vm)
1968 return vm->page_shift;
1971 unsigned int vm_get_max_gfn(struct kvm_vm *vm)
1976 int vm_get_fd(struct kvm_vm *vm)
1981 static unsigned int vm_calc_num_pages(unsigned int num_pages,
1982 unsigned int page_shift,
1983 unsigned int new_page_shift,
1986 unsigned int n = 1 << (new_page_shift - page_shift);
1988 if (page_shift >= new_page_shift)
1989 return num_pages * (1 << (page_shift - new_page_shift));
1991 return num_pages / n + !!(ceil && num_pages % n);
1994 static inline int getpageshift(void)
1996 return __builtin_ffs(getpagesize()) - 1;
2000 vm_num_host_pages(enum vm_guest_mode mode, unsigned int num_guest_pages)
2002 return vm_calc_num_pages(num_guest_pages,
2003 vm_guest_mode_params[mode].page_shift,
2004 getpageshift(), true);
2008 vm_num_guest_pages(enum vm_guest_mode mode, unsigned int num_host_pages)
2010 return vm_calc_num_pages(num_host_pages, getpageshift(),
2011 vm_guest_mode_params[mode].page_shift, false);
2014 unsigned int vm_calc_num_guest_pages(enum vm_guest_mode mode, size_t size)
2017 n = DIV_ROUND_UP(size, vm_guest_mode_params[mode].page_size);
2018 return vm_adjust_num_guest_pages(mode, n);