2 * tools/testing/selftests/kvm/lib/kvm_util.c
4 * Copyright (C) 2018, Google LLC.
6 * This work is licensed under the terms of the GNU GPL, version 2.
11 #include "kvm_util_internal.h"
15 #include <sys/types.h>
18 #define KVM_DEV_PATH "/dev/kvm"
20 #define KVM_UTIL_PGS_PER_HUGEPG 512
21 #define KVM_UTIL_MIN_PADDR 0x2000
23 /* Aligns x up to the next multiple of size. Size must be a power of 2. */
24 static void *align(void *x, size_t size)
26 size_t mask = size - 1;
27 TEST_ASSERT(size != 0 && !(size & (size - 1)),
28 "size not a power of 2: %lu", size);
29 return (void *) (((size_t) x + mask) & ~mask);
40 * On success, the Value corresponding to the capability (KVM_CAP_*)
41 * specified by the value of cap. On failure a TEST_ASSERT failure
44 * Looks up and returns the value corresponding to the capability
45 * (KVM_CAP_*) given by cap.
47 int kvm_check_cap(long cap)
52 kvm_fd = open(KVM_DEV_PATH, O_RDONLY);
56 ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, cap);
57 TEST_ASSERT(ret != -1, "KVM_CHECK_EXTENSION IOCTL failed,\n"
58 " rc: %i errno: %i", ret, errno);
68 * mode - VM Mode (e.g. VM_MODE_FLAT48PG)
69 * phy_pages - Physical memory pages
75 * Pointer to opaque structure that describes the created VM.
77 * Creates a VM with the mode specified by mode (e.g. VM_MODE_FLAT48PG).
78 * When phy_pages is non-zero, a memory region of phy_pages physical pages
79 * is created and mapped starting at guest physical address 0. The file
80 * descriptor to control the created VM is created with the permissions
81 * given by perm (e.g. O_RDWR).
83 struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
88 /* Allocate memory. */
89 vm = calloc(1, sizeof(*vm));
90 TEST_ASSERT(vm != NULL, "Insufficent Memory");
93 kvm_fd = open(KVM_DEV_PATH, perm);
98 vm->fd = ioctl(kvm_fd, KVM_CREATE_VM, NULL);
99 TEST_ASSERT(vm->fd >= 0, "KVM_CREATE_VM ioctl failed, "
100 "rc: %i errno: %i", vm->fd, errno);
104 /* Setup mode specific traits. */
106 case VM_MODE_FLAT48PG:
107 vm->page_size = 0x1000;
110 /* Limit to 48-bit canonical virtual addresses. */
111 vm->vpages_valid = sparsebit_alloc();
112 sparsebit_set_num(vm->vpages_valid,
113 0, (1ULL << (48 - 1)) >> vm->page_shift);
114 sparsebit_set_num(vm->vpages_valid,
115 (~((1ULL << (48 - 1)) - 1)) >> vm->page_shift,
116 (1ULL << (48 - 1)) >> vm->page_shift);
118 /* Limit physical addresses to 52-bits. */
119 vm->max_gfn = ((1ULL << 52) >> vm->page_shift) - 1;
123 TEST_ASSERT(false, "Unknown guest mode, mode: 0x%x", mode);
126 /* Allocate and setup memory for guest. */
127 vm->vpages_mapped = sparsebit_alloc();
129 vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS,
135 /* Userspace Memory Region Find
138 * vm - Virtual Machine
139 * start - Starting VM physical address
140 * end - Ending VM physical address, inclusive.
145 * Pointer to overlapping region, NULL if no such region.
147 * Searches for a region with any physical memory that overlaps with
148 * any portion of the guest physical addresses from start to end
149 * inclusive. If multiple overlapping regions exist, a pointer to any
150 * of the regions is returned. Null is returned only when no overlapping
153 static struct userspace_mem_region *userspace_mem_region_find(
154 struct kvm_vm *vm, uint64_t start, uint64_t end)
156 struct userspace_mem_region *region;
158 for (region = vm->userspace_mem_region_head; region;
159 region = region->next) {
160 uint64_t existing_start = region->region.guest_phys_addr;
161 uint64_t existing_end = region->region.guest_phys_addr
162 + region->region.memory_size - 1;
163 if (start <= existing_end && end >= existing_start)
170 /* KVM Userspace Memory Region Find
173 * vm - Virtual Machine
174 * start - Starting VM physical address
175 * end - Ending VM physical address, inclusive.
180 * Pointer to overlapping region, NULL if no such region.
182 * Public interface to userspace_mem_region_find. Allows tests to look up
183 * the memslot datastructure for a given range of guest physical memory.
185 struct kvm_userspace_memory_region *
186 kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
189 struct userspace_mem_region *region;
191 region = userspace_mem_region_find(vm, start, end);
195 return ®ion->region;
201 * vm - Virtual Machine
207 * Pointer to VCPU structure
209 * Locates a vcpu structure that describes the VCPU specified by vcpuid and
210 * returns a pointer to it. Returns NULL if the VM doesn't contain a VCPU
211 * for the specified vcpuid.
213 struct vcpu *vcpu_find(struct kvm_vm *vm,
218 for (vcpup = vm->vcpu_head; vcpup; vcpup = vcpup->next) {
219 if (vcpup->id == vcpuid)
229 * vm - Virtual Machine
234 * Return: None, TEST_ASSERT failures for all error conditions
236 * Within the VM specified by vm, removes the VCPU given by vcpuid.
238 static void vm_vcpu_rm(struct kvm_vm *vm, uint32_t vcpuid)
240 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
243 ret = munmap(vcpu->state, sizeof(*vcpu->state));
244 TEST_ASSERT(ret == 0, "munmap of VCPU fd failed, rc: %i "
245 "errno: %i", ret, errno);
247 TEST_ASSERT(ret == 0, "Close of VCPU fd failed, rc: %i "
248 "errno: %i", ret, errno);
251 vcpu->next->prev = vcpu->prev;
253 vcpu->prev->next = vcpu->next;
255 vm->vcpu_head = vcpu->next;
260 /* Destroys and frees the VM pointed to by vmp.
262 void kvm_vm_free(struct kvm_vm *vmp)
269 /* Free userspace_mem_regions. */
270 while (vmp->userspace_mem_region_head) {
271 struct userspace_mem_region *region
272 = vmp->userspace_mem_region_head;
274 region->region.memory_size = 0;
275 ret = ioctl(vmp->fd, KVM_SET_USER_MEMORY_REGION,
277 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed, "
278 "rc: %i errno: %i", ret, errno);
280 vmp->userspace_mem_region_head = region->next;
281 sparsebit_free(®ion->unused_phy_pages);
282 ret = munmap(region->mmap_start, region->mmap_size);
283 TEST_ASSERT(ret == 0, "munmap failed, rc: %i errno: %i",
290 while (vmp->vcpu_head)
291 vm_vcpu_rm(vmp, vmp->vcpu_head->id);
293 /* Free sparsebit arrays. */
294 sparsebit_free(&vmp->vpages_valid);
295 sparsebit_free(&vmp->vpages_mapped);
297 /* Close file descriptor for the VM. */
298 ret = close(vmp->fd);
299 TEST_ASSERT(ret == 0, "Close of vm fd failed,\n"
300 " vmp->fd: %i rc: %i errno: %i", vmp->fd, ret, errno);
303 TEST_ASSERT(ret == 0, "Close of /dev/kvm fd failed,\n"
304 " vmp->kvm_fd: %i rc: %i errno: %i", vmp->kvm_fd, ret, errno);
306 /* Free the structure describing the VM. */
310 /* Memory Compare, host virtual to guest virtual
313 * hva - Starting host virtual address
314 * vm - Virtual Machine
315 * gva - Starting guest virtual address
316 * len - number of bytes to compare
320 * Input/Output Args: None
323 * Returns 0 if the bytes starting at hva for a length of len
324 * are equal the guest virtual bytes starting at gva. Returns
325 * a value < 0, if bytes at hva are less than those at gva.
326 * Otherwise a value > 0 is returned.
328 * Compares the bytes starting at the host virtual address hva, for
329 * a length of len, to the guest bytes starting at the guest virtual
330 * address given by gva.
332 int kvm_memcmp_hva_gva(void *hva,
333 struct kvm_vm *vm, vm_vaddr_t gva, size_t len)
337 /* Compare a batch of bytes until either a match is found
338 * or all the bytes have been compared.
340 for (uintptr_t offset = 0; offset < len; offset += amt) {
341 uintptr_t ptr1 = (uintptr_t)hva + offset;
343 /* Determine host address for guest virtual address
346 uintptr_t ptr2 = (uintptr_t)addr_gva2hva(vm, gva + offset);
348 /* Determine amount to compare on this pass.
349 * Don't allow the comparsion to cross a page boundary.
352 if ((ptr1 >> vm->page_shift) != ((ptr1 + amt) >> vm->page_shift))
353 amt = vm->page_size - (ptr1 % vm->page_size);
354 if ((ptr2 >> vm->page_shift) != ((ptr2 + amt) >> vm->page_shift))
355 amt = vm->page_size - (ptr2 % vm->page_size);
357 assert((ptr1 >> vm->page_shift) == ((ptr1 + amt - 1) >> vm->page_shift));
358 assert((ptr2 >> vm->page_shift) == ((ptr2 + amt - 1) >> vm->page_shift));
360 /* Perform the comparison. If there is a difference
361 * return that result to the caller, otherwise need
362 * to continue on looking for a mismatch.
364 int ret = memcmp((void *)ptr1, (void *)ptr2, amt);
369 /* No mismatch found. Let the caller know the two memory
375 /* Allocate an instance of struct kvm_cpuid2
381 * Return: A pointer to the allocated struct. The caller is responsible
382 * for freeing this struct.
384 * Since kvm_cpuid2 uses a 0-length array to allow a the size of the
385 * array to be decided at allocation time, allocation is slightly
386 * complicated. This function uses a reasonable default length for
387 * the array and performs the appropriate allocation.
389 static struct kvm_cpuid2 *allocate_kvm_cpuid2(void)
391 struct kvm_cpuid2 *cpuid;
395 size = sizeof(*cpuid);
396 size += nent * sizeof(struct kvm_cpuid_entry2);
397 cpuid = malloc(size);
408 /* KVM Supported CPUID Get
414 * Return: The supported KVM CPUID
416 * Get the guest CPUID supported by KVM.
418 struct kvm_cpuid2 *kvm_get_supported_cpuid(void)
420 static struct kvm_cpuid2 *cpuid;
427 cpuid = allocate_kvm_cpuid2();
428 kvm_fd = open(KVM_DEV_PATH, O_RDONLY);
432 ret = ioctl(kvm_fd, KVM_GET_SUPPORTED_CPUID, cpuid);
433 TEST_ASSERT(ret == 0, "KVM_GET_SUPPORTED_CPUID failed %d %d\n",
440 /* Locate a cpuid entry.
444 * function: The function of the cpuid entry to find.
448 * Return: A pointer to the cpuid entry. Never returns NULL.
450 struct kvm_cpuid_entry2 *
451 kvm_get_supported_cpuid_index(uint32_t function, uint32_t index)
453 struct kvm_cpuid2 *cpuid;
454 struct kvm_cpuid_entry2 *entry = NULL;
457 cpuid = kvm_get_supported_cpuid();
458 for (i = 0; i < cpuid->nent; i++) {
459 if (cpuid->entries[i].function == function &&
460 cpuid->entries[i].index == index) {
461 entry = &cpuid->entries[i];
466 TEST_ASSERT(entry, "Guest CPUID entry not found: (EAX=%x, ECX=%x).",
471 /* VM Userspace Memory Region Add
474 * vm - Virtual Machine
475 * backing_src - Storage source for this region.
476 * NULL to use anonymous memory.
477 * guest_paddr - Starting guest physical address
478 * slot - KVM region slot
479 * npages - Number of physical pages
480 * flags - KVM memory region flags (e.g. KVM_MEM_LOG_DIRTY_PAGES)
486 * Allocates a memory area of the number of pages specified by npages
487 * and maps it to the VM specified by vm, at a starting physical address
488 * given by guest_paddr. The region is created with a KVM region slot
489 * given by slot, which must be unique and < KVM_MEM_SLOTS_NUM. The
490 * region is created with the flags given by flags.
492 void vm_userspace_mem_region_add(struct kvm_vm *vm,
493 enum vm_mem_backing_src_type src_type,
494 uint64_t guest_paddr, uint32_t slot, uint64_t npages,
498 unsigned long pmem_size = 0;
499 struct userspace_mem_region *region;
500 size_t huge_page_size = KVM_UTIL_PGS_PER_HUGEPG * vm->page_size;
502 TEST_ASSERT((guest_paddr % vm->page_size) == 0, "Guest physical "
503 "address not on a page boundary.\n"
504 " guest_paddr: 0x%lx vm->page_size: 0x%x",
505 guest_paddr, vm->page_size);
506 TEST_ASSERT((((guest_paddr >> vm->page_shift) + npages) - 1)
507 <= vm->max_gfn, "Physical range beyond maximum "
508 "supported physical address,\n"
509 " guest_paddr: 0x%lx npages: 0x%lx\n"
510 " vm->max_gfn: 0x%lx vm->page_size: 0x%x",
511 guest_paddr, npages, vm->max_gfn, vm->page_size);
513 /* Confirm a mem region with an overlapping address doesn't
516 region = (struct userspace_mem_region *) userspace_mem_region_find(
517 vm, guest_paddr, guest_paddr + npages * vm->page_size);
519 TEST_ASSERT(false, "overlapping userspace_mem_region already "
521 " requested guest_paddr: 0x%lx npages: 0x%lx "
523 " existing guest_paddr: 0x%lx size: 0x%lx",
524 guest_paddr, npages, vm->page_size,
525 (uint64_t) region->region.guest_phys_addr,
526 (uint64_t) region->region.memory_size);
528 /* Confirm no region with the requested slot already exists. */
529 for (region = vm->userspace_mem_region_head; region;
530 region = region->next) {
531 if (region->region.slot == slot)
533 if ((guest_paddr <= (region->region.guest_phys_addr
534 + region->region.memory_size))
535 && ((guest_paddr + npages * vm->page_size)
536 >= region->region.guest_phys_addr))
540 TEST_ASSERT(false, "A mem region with the requested slot "
541 "or overlapping physical memory range already exists.\n"
542 " requested slot: %u paddr: 0x%lx npages: 0x%lx\n"
543 " existing slot: %u paddr: 0x%lx size: 0x%lx",
544 slot, guest_paddr, npages,
546 (uint64_t) region->region.guest_phys_addr,
547 (uint64_t) region->region.memory_size);
549 /* Allocate and initialize new mem region structure. */
550 region = calloc(1, sizeof(*region));
551 TEST_ASSERT(region != NULL, "Insufficient Memory");
552 region->mmap_size = npages * vm->page_size;
554 /* Enough memory to align up to a huge page. */
555 if (src_type == VM_MEM_SRC_ANONYMOUS_THP)
556 region->mmap_size += huge_page_size;
557 region->mmap_start = mmap(NULL, region->mmap_size,
558 PROT_READ | PROT_WRITE,
559 MAP_PRIVATE | MAP_ANONYMOUS
560 | (src_type == VM_MEM_SRC_ANONYMOUS_HUGETLB ? MAP_HUGETLB : 0),
562 TEST_ASSERT(region->mmap_start != MAP_FAILED,
563 "test_malloc failed, mmap_start: %p errno: %i",
564 region->mmap_start, errno);
566 /* Align THP allocation up to start of a huge page. */
567 region->host_mem = align(region->mmap_start,
568 src_type == VM_MEM_SRC_ANONYMOUS_THP ? huge_page_size : 1);
570 /* As needed perform madvise */
571 if (src_type == VM_MEM_SRC_ANONYMOUS || src_type == VM_MEM_SRC_ANONYMOUS_THP) {
572 ret = madvise(region->host_mem, npages * vm->page_size,
573 src_type == VM_MEM_SRC_ANONYMOUS ? MADV_NOHUGEPAGE : MADV_HUGEPAGE);
574 TEST_ASSERT(ret == 0, "madvise failed,\n"
578 region->host_mem, npages * vm->page_size, src_type);
581 region->unused_phy_pages = sparsebit_alloc();
582 sparsebit_set_num(region->unused_phy_pages,
583 guest_paddr >> vm->page_shift, npages);
584 region->region.slot = slot;
585 region->region.flags = flags;
586 region->region.guest_phys_addr = guest_paddr;
587 region->region.memory_size = npages * vm->page_size;
588 region->region.userspace_addr = (uintptr_t) region->host_mem;
589 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
590 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
591 " rc: %i errno: %i\n"
592 " slot: %u flags: 0x%x\n"
593 " guest_phys_addr: 0x%lx size: 0x%lx",
594 ret, errno, slot, flags,
595 guest_paddr, (uint64_t) region->region.memory_size);
597 /* Add to linked-list of memory regions. */
598 if (vm->userspace_mem_region_head)
599 vm->userspace_mem_region_head->prev = region;
600 region->next = vm->userspace_mem_region_head;
601 vm->userspace_mem_region_head = region;
607 * vm - Virtual Machine
608 * memslot - KVM memory slot ID
613 * Pointer to memory region structure that describe memory region
614 * using kvm memory slot ID given by memslot. TEST_ASSERT failure
615 * on error (e.g. currently no memory region using memslot as a KVM
618 static struct userspace_mem_region *memslot2region(struct kvm_vm *vm,
621 struct userspace_mem_region *region;
623 for (region = vm->userspace_mem_region_head; region;
624 region = region->next) {
625 if (region->region.slot == memslot)
628 if (region == NULL) {
629 fprintf(stderr, "No mem region with the requested slot found,\n"
630 " requested slot: %u\n", memslot);
631 fputs("---- vm dump ----\n", stderr);
632 vm_dump(stderr, vm, 2);
633 TEST_ASSERT(false, "Mem region not found");
639 /* VM Memory Region Flags Set
642 * vm - Virtual Machine
643 * flags - Starting guest physical address
649 * Sets the flags of the memory region specified by the value of slot,
650 * to the values given by flags.
652 void vm_mem_region_set_flags(struct kvm_vm *vm, uint32_t slot, uint32_t flags)
655 struct userspace_mem_region *region;
657 /* Locate memory region. */
658 region = memslot2region(vm, slot);
660 region->region.flags = flags;
662 ret = ioctl(vm->fd, KVM_SET_USER_MEMORY_REGION, ®ion->region);
664 TEST_ASSERT(ret == 0, "KVM_SET_USER_MEMORY_REGION IOCTL failed,\n"
665 " rc: %i errno: %i slot: %u flags: 0x%x",
666 ret, errno, slot, flags);
678 * Returns the size of the structure pointed to by the return value
681 static int vcpu_mmap_sz(void)
685 dev_fd = open(KVM_DEV_PATH, O_RDONLY);
689 ret = ioctl(dev_fd, KVM_GET_VCPU_MMAP_SIZE, NULL);
690 TEST_ASSERT(ret >= sizeof(struct kvm_run),
691 "%s KVM_GET_VCPU_MMAP_SIZE ioctl failed, rc: %i errno: %i",
692 __func__, ret, errno);
702 * vm - Virtual Machine
709 * Creates and adds to the VM specified by vm and virtual CPU with
710 * the ID given by vcpuid.
712 void vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpuid, int pgd_memslot, int gdt_memslot)
716 /* Confirm a vcpu with the specified id doesn't already exist. */
717 vcpu = vcpu_find(vm, vcpuid);
719 TEST_ASSERT(false, "vcpu with the specified id "
721 " requested vcpuid: %u\n"
722 " existing vcpuid: %u state: %p",
723 vcpuid, vcpu->id, vcpu->state);
725 /* Allocate and initialize new vcpu structure. */
726 vcpu = calloc(1, sizeof(*vcpu));
727 TEST_ASSERT(vcpu != NULL, "Insufficient Memory");
729 vcpu->fd = ioctl(vm->fd, KVM_CREATE_VCPU, vcpuid);
730 TEST_ASSERT(vcpu->fd >= 0, "KVM_CREATE_VCPU failed, rc: %i errno: %i",
733 TEST_ASSERT(vcpu_mmap_sz() >= sizeof(*vcpu->state), "vcpu mmap size "
734 "smaller than expected, vcpu_mmap_sz: %i expected_min: %zi",
735 vcpu_mmap_sz(), sizeof(*vcpu->state));
736 vcpu->state = (struct kvm_run *) mmap(NULL, sizeof(*vcpu->state),
737 PROT_READ | PROT_WRITE, MAP_SHARED, vcpu->fd, 0);
738 TEST_ASSERT(vcpu->state != MAP_FAILED, "mmap vcpu_state failed, "
739 "vcpu id: %u errno: %i", vcpuid, errno);
741 /* Add to linked-list of VCPUs. */
743 vm->vcpu_head->prev = vcpu;
744 vcpu->next = vm->vcpu_head;
745 vm->vcpu_head = vcpu;
747 vcpu_setup(vm, vcpuid, pgd_memslot, gdt_memslot);
750 /* VM Virtual Address Unused Gap
753 * vm - Virtual Machine
755 * vaddr_min - Minimum Virtual Address
760 * Lowest virtual address at or below vaddr_min, with at least
761 * sz unused bytes. TEST_ASSERT failure if no area of at least
762 * size sz is available.
764 * Within the VM specified by vm, locates the lowest starting virtual
765 * address >= vaddr_min, that has at least sz unallocated bytes. A
766 * TEST_ASSERT failure occurs for invalid input or no area of at least
767 * sz unallocated bytes >= vaddr_min is available.
769 static vm_vaddr_t vm_vaddr_unused_gap(struct kvm_vm *vm, size_t sz,
770 vm_vaddr_t vaddr_min)
772 uint64_t pages = (sz + vm->page_size - 1) >> vm->page_shift;
774 /* Determine lowest permitted virtual page index. */
775 uint64_t pgidx_start = (vaddr_min + vm->page_size - 1) >> vm->page_shift;
776 if ((pgidx_start * vm->page_size) < vaddr_min)
779 /* Loop over section with enough valid virtual page indexes. */
780 if (!sparsebit_is_set_num(vm->vpages_valid,
782 pgidx_start = sparsebit_next_set_num(vm->vpages_valid,
786 * Are there enough unused virtual pages available at
787 * the currently proposed starting virtual page index.
788 * If not, adjust proposed starting index to next
791 if (sparsebit_is_clear_num(vm->vpages_mapped,
794 pgidx_start = sparsebit_next_clear_num(vm->vpages_mapped,
796 if (pgidx_start == 0)
800 * If needed, adjust proposed starting virtual address,
801 * to next range of valid virtual addresses.
803 if (!sparsebit_is_set_num(vm->vpages_valid,
804 pgidx_start, pages)) {
805 pgidx_start = sparsebit_next_set_num(
806 vm->vpages_valid, pgidx_start, pages);
807 if (pgidx_start == 0)
810 } while (pgidx_start != 0);
813 TEST_ASSERT(false, "No vaddr of specified pages available, "
814 "pages: 0x%lx", pages);
820 TEST_ASSERT(sparsebit_is_set_num(vm->vpages_valid,
822 "Unexpected, invalid virtual page index range,\n"
823 " pgidx_start: 0x%lx\n"
826 TEST_ASSERT(sparsebit_is_clear_num(vm->vpages_mapped,
828 "Unexpected, pages already mapped,\n"
829 " pgidx_start: 0x%lx\n"
833 return pgidx_start * vm->page_size;
836 /* VM Virtual Address Allocate
839 * vm - Virtual Machine
841 * vaddr_min - Minimum starting virtual address
842 * data_memslot - Memory region slot for data pages
843 * pgd_memslot - Memory region slot for new virtual translation tables
848 * Starting guest virtual address
850 * Allocates at least sz bytes within the virtual address space of the vm
851 * given by vm. The allocated bytes are mapped to a virtual address >=
852 * the address given by vaddr_min. Note that each allocation uses a
853 * a unique set of pages, with the minimum real allocation being at least
856 vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
857 uint32_t data_memslot, uint32_t pgd_memslot)
859 uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
861 virt_pgd_alloc(vm, pgd_memslot);
863 /* Find an unused range of virtual page addresses of at least
866 vm_vaddr_t vaddr_start = vm_vaddr_unused_gap(vm, sz, vaddr_min);
868 /* Map the virtual pages. */
869 for (vm_vaddr_t vaddr = vaddr_start; pages > 0;
870 pages--, vaddr += vm->page_size) {
873 paddr = vm_phy_page_alloc(vm, KVM_UTIL_MIN_PADDR, data_memslot);
875 virt_pg_map(vm, vaddr, paddr, pgd_memslot);
877 sparsebit_set(vm->vpages_mapped,
878 vaddr >> vm->page_shift);
884 /* Address VM Physical to Host Virtual
887 * vm - Virtual Machine
888 * gpa - VM physical address
893 * Equivalent host virtual address
895 * Locates the memory region containing the VM physical address given
896 * by gpa, within the VM given by vm. When found, the host virtual
897 * address providing the memory to the vm physical address is returned.
898 * A TEST_ASSERT failure occurs if no region containing gpa exists.
900 void *addr_gpa2hva(struct kvm_vm *vm, vm_paddr_t gpa)
902 struct userspace_mem_region *region;
903 for (region = vm->userspace_mem_region_head; region;
904 region = region->next) {
905 if ((gpa >= region->region.guest_phys_addr)
906 && (gpa <= (region->region.guest_phys_addr
907 + region->region.memory_size - 1)))
908 return (void *) ((uintptr_t) region->host_mem
909 + (gpa - region->region.guest_phys_addr));
912 TEST_ASSERT(false, "No vm physical memory at 0x%lx", gpa);
916 /* Address Host Virtual to VM Physical
919 * vm - Virtual Machine
920 * hva - Host virtual address
925 * Equivalent VM physical address
927 * Locates the memory region containing the host virtual address given
928 * by hva, within the VM given by vm. When found, the equivalent
929 * VM physical address is returned. A TEST_ASSERT failure occurs if no
930 * region containing hva exists.
932 vm_paddr_t addr_hva2gpa(struct kvm_vm *vm, void *hva)
934 struct userspace_mem_region *region;
935 for (region = vm->userspace_mem_region_head; region;
936 region = region->next) {
937 if ((hva >= region->host_mem)
938 && (hva <= (region->host_mem
939 + region->region.memory_size - 1)))
940 return (vm_paddr_t) ((uintptr_t)
941 region->region.guest_phys_addr
942 + (hva - (uintptr_t) region->host_mem));
945 TEST_ASSERT(false, "No mapping to a guest physical address, "
950 /* VM Create IRQ Chip
953 * vm - Virtual Machine
959 * Creates an interrupt controller chip for the VM specified by vm.
961 void vm_create_irqchip(struct kvm_vm *vm)
965 ret = ioctl(vm->fd, KVM_CREATE_IRQCHIP, 0);
966 TEST_ASSERT(ret == 0, "KVM_CREATE_IRQCHIP IOCTL failed, "
967 "rc: %i errno: %i", ret, errno);
973 * vm - Virtual Machine
979 * Pointer to structure that describes the state of the VCPU.
981 * Locates and returns a pointer to a structure that describes the
982 * state of the VCPU with the given vcpuid.
984 struct kvm_run *vcpu_state(struct kvm_vm *vm, uint32_t vcpuid)
986 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
987 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
995 * vm - Virtual Machine
1002 * Switch to executing the code for the VCPU given by vcpuid, within the VM
1005 void vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1007 int ret = _vcpu_run(vm, vcpuid);
1008 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1009 "rc: %i errno: %i", ret, errno);
1012 int _vcpu_run(struct kvm_vm *vm, uint32_t vcpuid)
1014 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1017 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1019 rc = ioctl(vcpu->fd, KVM_RUN, NULL);
1020 } while (rc == -1 && errno == EINTR);
1024 /* VM VCPU Set MP State
1027 * vm - Virtual Machine
1029 * mp_state - mp_state to be set
1035 * Sets the MP state of the VCPU given by vcpuid, to the state given
1038 void vcpu_set_mp_state(struct kvm_vm *vm, uint32_t vcpuid,
1039 struct kvm_mp_state *mp_state)
1041 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1044 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1046 ret = ioctl(vcpu->fd, KVM_SET_MP_STATE, mp_state);
1047 TEST_ASSERT(ret == 0, "KVM_SET_MP_STATE IOCTL failed, "
1048 "rc: %i errno: %i", ret, errno);
1054 * vm - Virtual Machine
1058 * regs - current state of VCPU regs
1062 * Obtains the current register state for the VCPU specified by vcpuid
1063 * and stores it at the location given by regs.
1065 void vcpu_regs_get(struct kvm_vm *vm,
1066 uint32_t vcpuid, struct kvm_regs *regs)
1068 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1071 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1074 ret = ioctl(vcpu->fd, KVM_GET_REGS, regs);
1075 TEST_ASSERT(ret == 0, "KVM_GET_REGS failed, rc: %i errno: %i",
1082 * vm - Virtual Machine
1084 * regs - Values to set VCPU regs to
1090 * Sets the regs of the VCPU specified by vcpuid to the values
1093 void vcpu_regs_set(struct kvm_vm *vm,
1094 uint32_t vcpuid, struct kvm_regs *regs)
1096 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1099 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1102 ret = ioctl(vcpu->fd, KVM_SET_REGS, regs);
1103 TEST_ASSERT(ret == 0, "KVM_SET_REGS failed, rc: %i errno: %i",
1107 void vcpu_events_get(struct kvm_vm *vm, uint32_t vcpuid,
1108 struct kvm_vcpu_events *events)
1110 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1113 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1116 ret = ioctl(vcpu->fd, KVM_GET_VCPU_EVENTS, events);
1117 TEST_ASSERT(ret == 0, "KVM_GET_VCPU_EVENTS, failed, rc: %i errno: %i",
1121 void vcpu_events_set(struct kvm_vm *vm, uint32_t vcpuid,
1122 struct kvm_vcpu_events *events)
1124 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1127 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1130 ret = ioctl(vcpu->fd, KVM_SET_VCPU_EVENTS, events);
1131 TEST_ASSERT(ret == 0, "KVM_SET_VCPU_EVENTS, failed, rc: %i errno: %i",
1138 * vm - Virtual Machine
1140 * num - number of arguments
1141 * ... - arguments, each of type uint64_t
1147 * Sets the first num function input arguments to the values
1148 * given as variable args. Each of the variable args is expected to
1149 * be of type uint64_t.
1151 void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)
1154 struct kvm_regs regs;
1156 TEST_ASSERT(num >= 1 && num <= 6, "Unsupported number of args,\n"
1161 vcpu_regs_get(vm, vcpuid, ®s);
1164 regs.rdi = va_arg(ap, uint64_t);
1167 regs.rsi = va_arg(ap, uint64_t);
1170 regs.rdx = va_arg(ap, uint64_t);
1173 regs.rcx = va_arg(ap, uint64_t);
1176 regs.r8 = va_arg(ap, uint64_t);
1179 regs.r9 = va_arg(ap, uint64_t);
1181 vcpu_regs_set(vm, vcpuid, ®s);
1185 /* VM VCPU System Regs Get
1188 * vm - Virtual Machine
1192 * sregs - current state of VCPU system regs
1196 * Obtains the current system register state for the VCPU specified by
1197 * vcpuid and stores it at the location given by sregs.
1199 void vcpu_sregs_get(struct kvm_vm *vm,
1200 uint32_t vcpuid, struct kvm_sregs *sregs)
1202 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1205 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1209 ret = ioctl(vcpu->fd, KVM_GET_SREGS, sregs);
1210 TEST_ASSERT(ret == 0, "KVM_GET_SREGS failed, rc: %i errno: %i",
1214 /* VM VCPU System Regs Set
1217 * vm - Virtual Machine
1219 * sregs - Values to set VCPU system regs to
1225 * Sets the system regs of the VCPU specified by vcpuid to the values
1228 void vcpu_sregs_set(struct kvm_vm *vm,
1229 uint32_t vcpuid, struct kvm_sregs *sregs)
1231 int ret = _vcpu_sregs_set(vm, vcpuid, sregs);
1232 TEST_ASSERT(ret == 0, "KVM_RUN IOCTL failed, "
1233 "rc: %i errno: %i", ret, errno);
1236 int _vcpu_sregs_set(struct kvm_vm *vm,
1237 uint32_t vcpuid, struct kvm_sregs *sregs)
1239 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1242 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1245 return ioctl(vcpu->fd, KVM_SET_SREGS, sregs);
1251 * vm - Virtual Machine
1253 * cmd - Ioctl number
1254 * arg - Argument to pass to the ioctl
1258 * Issues an arbitrary ioctl on a VCPU fd.
1260 void vcpu_ioctl(struct kvm_vm *vm,
1261 uint32_t vcpuid, unsigned long cmd, void *arg)
1263 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
1266 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
1268 ret = ioctl(vcpu->fd, cmd, arg);
1269 TEST_ASSERT(ret == 0, "vcpu ioctl %lu failed, rc: %i errno: %i (%s)",
1270 cmd, ret, errno, strerror(errno));
1276 * vm - Virtual Machine
1277 * cmd - Ioctl number
1278 * arg - Argument to pass to the ioctl
1282 * Issues an arbitrary ioctl on a VM fd.
1284 void vm_ioctl(struct kvm_vm *vm, unsigned long cmd, void *arg)
1288 ret = ioctl(vm->fd, cmd, arg);
1289 TEST_ASSERT(ret == 0, "vm ioctl %lu failed, rc: %i errno: %i (%s)",
1290 cmd, ret, errno, strerror(errno));
1296 * vm - Virtual Machine
1297 * indent - Left margin indent amount
1300 * stream - Output FILE stream
1304 * Dumps the current state of the VM given by vm, to the FILE stream
1307 void vm_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
1309 struct userspace_mem_region *region;
1312 fprintf(stream, "%*smode: 0x%x\n", indent, "", vm->mode);
1313 fprintf(stream, "%*sfd: %i\n", indent, "", vm->fd);
1314 fprintf(stream, "%*spage_size: 0x%x\n", indent, "", vm->page_size);
1315 fprintf(stream, "%*sMem Regions:\n", indent, "");
1316 for (region = vm->userspace_mem_region_head; region;
1317 region = region->next) {
1318 fprintf(stream, "%*sguest_phys: 0x%lx size: 0x%lx "
1319 "host_virt: %p\n", indent + 2, "",
1320 (uint64_t) region->region.guest_phys_addr,
1321 (uint64_t) region->region.memory_size,
1323 fprintf(stream, "%*sunused_phy_pages: ", indent + 2, "");
1324 sparsebit_dump(stream, region->unused_phy_pages, 0);
1326 fprintf(stream, "%*sMapped Virtual Pages:\n", indent, "");
1327 sparsebit_dump(stream, vm->vpages_mapped, indent + 2);
1328 fprintf(stream, "%*spgd_created: %u\n", indent, "",
1330 if (vm->pgd_created) {
1331 fprintf(stream, "%*sVirtual Translation Tables:\n",
1333 virt_dump(stream, vm, indent + 4);
1335 fprintf(stream, "%*sVCPUs:\n", indent, "");
1336 for (vcpu = vm->vcpu_head; vcpu; vcpu = vcpu->next)
1337 vcpu_dump(stream, vm, vcpu->id, indent + 2);
1343 * vm - Virtual Machine
1345 * indent - Left margin indent amount
1348 * stream - Output FILE stream
1352 * Dumps the current state of the VCPU specified by vcpuid, within the VM
1353 * given by vm, to the FILE stream given by stream.
1355 void vcpu_dump(FILE *stream, struct kvm_vm *vm,
1356 uint32_t vcpuid, uint8_t indent)
1358 struct kvm_regs regs;
1359 struct kvm_sregs sregs;
1361 fprintf(stream, "%*scpuid: %u\n", indent, "", vcpuid);
1363 fprintf(stream, "%*sregs:\n", indent + 2, "");
1364 vcpu_regs_get(vm, vcpuid, ®s);
1365 regs_dump(stream, ®s, indent + 4);
1367 fprintf(stream, "%*ssregs:\n", indent + 2, "");
1368 vcpu_sregs_get(vm, vcpuid, &sregs);
1369 sregs_dump(stream, &sregs, indent + 4);
1372 /* Known KVM exit reasons */
1373 static struct exit_reason {
1374 unsigned int reason;
1376 } exit_reasons_known[] = {
1377 {KVM_EXIT_UNKNOWN, "UNKNOWN"},
1378 {KVM_EXIT_EXCEPTION, "EXCEPTION"},
1379 {KVM_EXIT_IO, "IO"},
1380 {KVM_EXIT_HYPERCALL, "HYPERCALL"},
1381 {KVM_EXIT_DEBUG, "DEBUG"},
1382 {KVM_EXIT_HLT, "HLT"},
1383 {KVM_EXIT_MMIO, "MMIO"},
1384 {KVM_EXIT_IRQ_WINDOW_OPEN, "IRQ_WINDOW_OPEN"},
1385 {KVM_EXIT_SHUTDOWN, "SHUTDOWN"},
1386 {KVM_EXIT_FAIL_ENTRY, "FAIL_ENTRY"},
1387 {KVM_EXIT_INTR, "INTR"},
1388 {KVM_EXIT_SET_TPR, "SET_TPR"},
1389 {KVM_EXIT_TPR_ACCESS, "TPR_ACCESS"},
1390 {KVM_EXIT_S390_SIEIC, "S390_SIEIC"},
1391 {KVM_EXIT_S390_RESET, "S390_RESET"},
1392 {KVM_EXIT_DCR, "DCR"},
1393 {KVM_EXIT_NMI, "NMI"},
1394 {KVM_EXIT_INTERNAL_ERROR, "INTERNAL_ERROR"},
1395 {KVM_EXIT_OSI, "OSI"},
1396 {KVM_EXIT_PAPR_HCALL, "PAPR_HCALL"},
1397 #ifdef KVM_EXIT_MEMORY_NOT_PRESENT
1398 {KVM_EXIT_MEMORY_NOT_PRESENT, "MEMORY_NOT_PRESENT"},
1402 /* Exit Reason String
1405 * exit_reason - Exit reason
1410 * Constant string pointer describing the exit reason.
1412 * Locates and returns a constant string that describes the KVM exit
1413 * reason given by exit_reason. If no such string is found, a constant
1414 * string of "Unknown" is returned.
1416 const char *exit_reason_str(unsigned int exit_reason)
1420 for (n1 = 0; n1 < ARRAY_SIZE(exit_reasons_known); n1++) {
1421 if (exit_reason == exit_reasons_known[n1].reason)
1422 return exit_reasons_known[n1].name;
1428 /* Physical Page Allocate
1431 * vm - Virtual Machine
1432 * paddr_min - Physical address minimum
1433 * memslot - Memory region to allocate page from
1438 * Starting physical address
1440 * Within the VM specified by vm, locates an available physical page
1441 * at or above paddr_min. If found, the page is marked as in use
1442 * and its address is returned. A TEST_ASSERT failure occurs if no
1443 * page is available at or above paddr_min.
1445 vm_paddr_t vm_phy_page_alloc(struct kvm_vm *vm,
1446 vm_paddr_t paddr_min, uint32_t memslot)
1448 struct userspace_mem_region *region;
1451 TEST_ASSERT((paddr_min % vm->page_size) == 0, "Min physical address "
1452 "not divisible by page size.\n"
1453 " paddr_min: 0x%lx page_size: 0x%x",
1454 paddr_min, vm->page_size);
1456 /* Locate memory region. */
1457 region = memslot2region(vm, memslot);
1459 /* Locate next available physical page at or above paddr_min. */
1460 pg = paddr_min >> vm->page_shift;
1462 if (!sparsebit_is_set(region->unused_phy_pages, pg)) {
1463 pg = sparsebit_next_set(region->unused_phy_pages, pg);
1465 fprintf(stderr, "No guest physical page available, "
1466 "paddr_min: 0x%lx page_size: 0x%x memslot: %u",
1467 paddr_min, vm->page_size, memslot);
1468 fputs("---- vm dump ----\n", stderr);
1469 vm_dump(stderr, vm, 2);
1474 /* Specify page as in use and return its address. */
1475 sparsebit_clear(region->unused_phy_pages, pg);
1477 return pg * vm->page_size;
1480 /* Address Guest Virtual to Host Virtual
1483 * vm - Virtual Machine
1484 * gva - VM virtual address
1489 * Equivalent host virtual address
1491 void *addr_gva2hva(struct kvm_vm *vm, vm_vaddr_t gva)
1493 return addr_gpa2hva(vm, addr_gva2gpa(vm, gva));