xref: /linux/tools/testing/selftests/kvm/include/kvm_util.h (revision 2b7deea3ec7c81a92d4c17751d3bcd780d065ae4)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (C) 2018, Google LLC.
4  */
5 #ifndef SELFTEST_KVM_UTIL_H
6 #define SELFTEST_KVM_UTIL_H
7 
8 #include "test_util.h"
9 
10 #include <linux/compiler.h>
11 #include "linux/hashtable.h"
12 #include "linux/list.h"
13 #include <linux/kernel.h>
14 #include <linux/kvm.h>
15 #include "linux/rbtree.h"
16 #include <linux/types.h>
17 
18 #include <asm/atomic.h>
19 #include <asm/kvm.h>
20 
21 #include <sys/ioctl.h>
22 
23 #include "kvm_util_arch.h"
24 #include "sparsebit.h"
25 
26 /*
27  * Provide a version of static_assert() that is guaranteed to have an optional
28  * message param.  _GNU_SOURCE is defined for all KVM selftests, _GNU_SOURCE
29  * implies _ISOC11_SOURCE, and if _ISOC11_SOURCE is defined, glibc #undefs and
30  * #defines static_assert() as a direct alias to _Static_assert() (see
31  * usr/include/assert.h).  Define a custom macro instead of redefining
32  * static_assert() to avoid creating non-deterministic behavior that is
33  * dependent on include order.
34  */
35 #define __kvm_static_assert(expr, msg, ...) _Static_assert(expr, msg)
36 #define kvm_static_assert(expr, ...) __kvm_static_assert(expr, ##__VA_ARGS__, #expr)
37 
38 #define KVM_DEV_PATH "/dev/kvm"
39 #define KVM_MAX_VCPUS 512
40 
41 #define NSEC_PER_SEC 1000000000L
42 
43 typedef uint64_t vm_paddr_t; /* Virtual Machine (Guest) physical address */
44 typedef uint64_t vm_vaddr_t; /* Virtual Machine (Guest) virtual address */
45 
46 struct userspace_mem_region {
47 	struct kvm_userspace_memory_region2 region;
48 	struct sparsebit *unused_phy_pages;
49 	struct sparsebit *protected_phy_pages;
50 	int fd;
51 	off_t offset;
52 	enum vm_mem_backing_src_type backing_src_type;
53 	void *host_mem;
54 	void *host_alias;
55 	void *mmap_start;
56 	void *mmap_alias;
57 	size_t mmap_size;
58 	struct rb_node gpa_node;
59 	struct rb_node hva_node;
60 	struct hlist_node slot_node;
61 };
62 
63 struct kvm_vcpu {
64 	struct list_head list;
65 	uint32_t id;
66 	int fd;
67 	struct kvm_vm *vm;
68 	struct kvm_run *run;
69 #ifdef __x86_64__
70 	struct kvm_cpuid2 *cpuid;
71 #endif
72 	struct kvm_dirty_gfn *dirty_gfns;
73 	uint32_t fetch_index;
74 	uint32_t dirty_gfns_count;
75 };
76 
77 struct userspace_mem_regions {
78 	struct rb_root gpa_tree;
79 	struct rb_root hva_tree;
80 	DECLARE_HASHTABLE(slot_hash, 9);
81 };
82 
83 enum kvm_mem_region_type {
84 	MEM_REGION_CODE,
85 	MEM_REGION_DATA,
86 	MEM_REGION_PT,
87 	MEM_REGION_TEST_DATA,
88 	NR_MEM_REGIONS,
89 };
90 
91 struct kvm_vm {
92 	int mode;
93 	unsigned long type;
94 	int kvm_fd;
95 	int fd;
96 	unsigned int pgtable_levels;
97 	unsigned int page_size;
98 	unsigned int page_shift;
99 	unsigned int pa_bits;
100 	unsigned int va_bits;
101 	uint64_t max_gfn;
102 	struct list_head vcpus;
103 	struct userspace_mem_regions regions;
104 	struct sparsebit *vpages_valid;
105 	struct sparsebit *vpages_mapped;
106 	bool has_irqchip;
107 	bool pgd_created;
108 	vm_paddr_t ucall_mmio_addr;
109 	vm_paddr_t pgd;
110 	vm_vaddr_t gdt;
111 	vm_vaddr_t tss;
112 	vm_vaddr_t idt;
113 	vm_vaddr_t handlers;
114 	uint32_t dirty_ring_size;
115 	uint64_t gpa_tag_mask;
116 
117 	struct kvm_vm_arch arch;
118 
119 	/* Cache of information for binary stats interface */
120 	int stats_fd;
121 	struct kvm_stats_header stats_header;
122 	struct kvm_stats_desc *stats_desc;
123 
124 	/*
125 	 * KVM region slots. These are the default memslots used by page
126 	 * allocators, e.g., lib/elf uses the memslots[MEM_REGION_CODE]
127 	 * memslot.
128 	 */
129 	uint32_t memslots[NR_MEM_REGIONS];
130 };
131 
132 struct vcpu_reg_sublist {
133 	const char *name;
134 	long capability;
135 	int feature;
136 	int feature_type;
137 	bool finalize;
138 	__u64 *regs;
139 	__u64 regs_n;
140 	__u64 *rejects_set;
141 	__u64 rejects_set_n;
142 	__u64 *skips_set;
143 	__u64 skips_set_n;
144 };
145 
146 struct vcpu_reg_list {
147 	char *name;
148 	struct vcpu_reg_sublist sublists[];
149 };
150 
151 #define for_each_sublist(c, s)		\
152 	for ((s) = &(c)->sublists[0]; (s)->regs; ++(s))
153 
154 #define kvm_for_each_vcpu(vm, i, vcpu)			\
155 	for ((i) = 0; (i) <= (vm)->last_vcpu_id; (i)++)	\
156 		if (!((vcpu) = vm->vcpus[i]))		\
157 			continue;			\
158 		else
159 
160 struct userspace_mem_region *
161 memslot2region(struct kvm_vm *vm, uint32_t memslot);
162 
163 static inline struct userspace_mem_region *vm_get_mem_region(struct kvm_vm *vm,
164 							     enum kvm_mem_region_type type)
165 {
166 	assert(type < NR_MEM_REGIONS);
167 	return memslot2region(vm, vm->memslots[type]);
168 }
169 
170 /* Minimum allocated guest virtual and physical addresses */
171 #define KVM_UTIL_MIN_VADDR		0x2000
172 #define KVM_GUEST_PAGE_TABLE_MIN_PADDR	0x180000
173 
174 #define DEFAULT_GUEST_STACK_VADDR_MIN	0xab6000
175 #define DEFAULT_STACK_PGS		5
176 
177 enum vm_guest_mode {
178 	VM_MODE_P52V48_4K,
179 	VM_MODE_P52V48_16K,
180 	VM_MODE_P52V48_64K,
181 	VM_MODE_P48V48_4K,
182 	VM_MODE_P48V48_16K,
183 	VM_MODE_P48V48_64K,
184 	VM_MODE_P40V48_4K,
185 	VM_MODE_P40V48_16K,
186 	VM_MODE_P40V48_64K,
187 	VM_MODE_PXXV48_4K,	/* For 48bits VA but ANY bits PA */
188 	VM_MODE_P47V64_4K,
189 	VM_MODE_P44V64_4K,
190 	VM_MODE_P36V48_4K,
191 	VM_MODE_P36V48_16K,
192 	VM_MODE_P36V48_64K,
193 	VM_MODE_P36V47_16K,
194 	NUM_VM_MODES,
195 };
196 
197 struct vm_shape {
198 	uint32_t type;
199 	uint8_t  mode;
200 	uint8_t  pad0;
201 	uint16_t pad1;
202 };
203 
204 kvm_static_assert(sizeof(struct vm_shape) == sizeof(uint64_t));
205 
206 #define VM_TYPE_DEFAULT			0
207 
208 #define VM_SHAPE(__mode)			\
209 ({						\
210 	struct vm_shape shape = {		\
211 		.mode = (__mode),		\
212 		.type = VM_TYPE_DEFAULT		\
213 	};					\
214 						\
215 	shape;					\
216 })
217 
218 #if defined(__aarch64__)
219 
220 extern enum vm_guest_mode vm_mode_default;
221 
222 #define VM_MODE_DEFAULT			vm_mode_default
223 #define MIN_PAGE_SHIFT			12U
224 #define ptes_per_page(page_size)	((page_size) / 8)
225 
226 #elif defined(__x86_64__)
227 
228 #define VM_MODE_DEFAULT			VM_MODE_PXXV48_4K
229 #define MIN_PAGE_SHIFT			12U
230 #define ptes_per_page(page_size)	((page_size) / 8)
231 
232 #elif defined(__s390x__)
233 
234 #define VM_MODE_DEFAULT			VM_MODE_P44V64_4K
235 #define MIN_PAGE_SHIFT			12U
236 #define ptes_per_page(page_size)	((page_size) / 16)
237 
238 #elif defined(__riscv)
239 
240 #if __riscv_xlen == 32
241 #error "RISC-V 32-bit kvm selftests not supported"
242 #endif
243 
244 #define VM_MODE_DEFAULT			VM_MODE_P40V48_4K
245 #define MIN_PAGE_SHIFT			12U
246 #define ptes_per_page(page_size)	((page_size) / 8)
247 
248 #endif
249 
250 #define VM_SHAPE_DEFAULT	VM_SHAPE(VM_MODE_DEFAULT)
251 
252 #define MIN_PAGE_SIZE		(1U << MIN_PAGE_SHIFT)
253 #define PTES_PER_MIN_PAGE	ptes_per_page(MIN_PAGE_SIZE)
254 
255 struct vm_guest_mode_params {
256 	unsigned int pa_bits;
257 	unsigned int va_bits;
258 	unsigned int page_size;
259 	unsigned int page_shift;
260 };
261 extern const struct vm_guest_mode_params vm_guest_mode_params[];
262 
263 int open_path_or_exit(const char *path, int flags);
264 int open_kvm_dev_path_or_exit(void);
265 
266 bool get_kvm_param_bool(const char *param);
267 bool get_kvm_intel_param_bool(const char *param);
268 bool get_kvm_amd_param_bool(const char *param);
269 
270 int get_kvm_param_integer(const char *param);
271 int get_kvm_intel_param_integer(const char *param);
272 int get_kvm_amd_param_integer(const char *param);
273 
274 unsigned int kvm_check_cap(long cap);
275 
276 static inline bool kvm_has_cap(long cap)
277 {
278 	return kvm_check_cap(cap);
279 }
280 
281 #define __KVM_SYSCALL_ERROR(_name, _ret) \
282 	"%s failed, rc: %i errno: %i (%s)", (_name), (_ret), errno, strerror(errno)
283 
284 /*
285  * Use the "inner", double-underscore macro when reporting errors from within
286  * other macros so that the name of ioctl() and not its literal numeric value
287  * is printed on error.  The "outer" macro is strongly preferred when reporting
288  * errors "directly", i.e. without an additional layer of macros, as it reduces
289  * the probability of passing in the wrong string.
290  */
291 #define __KVM_IOCTL_ERROR(_name, _ret)	__KVM_SYSCALL_ERROR(_name, _ret)
292 #define KVM_IOCTL_ERROR(_ioctl, _ret) __KVM_IOCTL_ERROR(#_ioctl, _ret)
293 
294 #define kvm_do_ioctl(fd, cmd, arg)						\
295 ({										\
296 	kvm_static_assert(!_IOC_SIZE(cmd) || sizeof(*arg) == _IOC_SIZE(cmd));	\
297 	ioctl(fd, cmd, arg);							\
298 })
299 
300 #define __kvm_ioctl(kvm_fd, cmd, arg)				\
301 	kvm_do_ioctl(kvm_fd, cmd, arg)
302 
303 #define kvm_ioctl(kvm_fd, cmd, arg)				\
304 ({								\
305 	int ret = __kvm_ioctl(kvm_fd, cmd, arg);		\
306 								\
307 	TEST_ASSERT(!ret, __KVM_IOCTL_ERROR(#cmd, ret));	\
308 })
309 
310 static __always_inline void static_assert_is_vm(struct kvm_vm *vm) { }
311 
312 #define __vm_ioctl(vm, cmd, arg)				\
313 ({								\
314 	static_assert_is_vm(vm);				\
315 	kvm_do_ioctl((vm)->fd, cmd, arg);			\
316 })
317 
318 /*
319  * Assert that a VM or vCPU ioctl() succeeded, with extra magic to detect if
320  * the ioctl() failed because KVM killed/bugged the VM.  To detect a dead VM,
321  * probe KVM_CAP_USER_MEMORY, which (a) has been supported by KVM since before
322  * selftests existed and (b) should never outright fail, i.e. is supposed to
323  * return 0 or 1.  If KVM kills a VM, KVM returns -EIO for all ioctl()s for the
324  * VM and its vCPUs, including KVM_CHECK_EXTENSION.
325  */
326 #define __TEST_ASSERT_VM_VCPU_IOCTL(cond, name, ret, vm)				\
327 do {											\
328 	int __errno = errno;								\
329 											\
330 	static_assert_is_vm(vm);							\
331 											\
332 	if (cond)									\
333 		break;									\
334 											\
335 	if (errno == EIO &&								\
336 	    __vm_ioctl(vm, KVM_CHECK_EXTENSION, (void *)KVM_CAP_USER_MEMORY) < 0) {	\
337 		TEST_ASSERT(errno == EIO, "KVM killed the VM, should return -EIO");	\
338 		TEST_FAIL("KVM killed/bugged the VM, check the kernel log for clues");	\
339 	}										\
340 	errno = __errno;								\
341 	TEST_ASSERT(cond, __KVM_IOCTL_ERROR(name, ret));				\
342 } while (0)
343 
344 #define TEST_ASSERT_VM_VCPU_IOCTL(cond, cmd, ret, vm)		\
345 	__TEST_ASSERT_VM_VCPU_IOCTL(cond, #cmd, ret, vm)
346 
347 #define vm_ioctl(vm, cmd, arg)					\
348 ({								\
349 	int ret = __vm_ioctl(vm, cmd, arg);			\
350 								\
351 	__TEST_ASSERT_VM_VCPU_IOCTL(!ret, #cmd, ret, vm);		\
352 })
353 
354 static __always_inline void static_assert_is_vcpu(struct kvm_vcpu *vcpu) { }
355 
356 #define __vcpu_ioctl(vcpu, cmd, arg)				\
357 ({								\
358 	static_assert_is_vcpu(vcpu);				\
359 	kvm_do_ioctl((vcpu)->fd, cmd, arg);			\
360 })
361 
362 #define vcpu_ioctl(vcpu, cmd, arg)				\
363 ({								\
364 	int ret = __vcpu_ioctl(vcpu, cmd, arg);			\
365 								\
366 	__TEST_ASSERT_VM_VCPU_IOCTL(!ret, #cmd, ret, (vcpu)->vm);	\
367 })
368 
369 /*
370  * Looks up and returns the value corresponding to the capability
371  * (KVM_CAP_*) given by cap.
372  */
373 static inline int vm_check_cap(struct kvm_vm *vm, long cap)
374 {
375 	int ret =  __vm_ioctl(vm, KVM_CHECK_EXTENSION, (void *)cap);
376 
377 	TEST_ASSERT_VM_VCPU_IOCTL(ret >= 0, KVM_CHECK_EXTENSION, ret, vm);
378 	return ret;
379 }
380 
381 static inline int __vm_enable_cap(struct kvm_vm *vm, uint32_t cap, uint64_t arg0)
382 {
383 	struct kvm_enable_cap enable_cap = { .cap = cap, .args = { arg0 } };
384 
385 	return __vm_ioctl(vm, KVM_ENABLE_CAP, &enable_cap);
386 }
387 static inline void vm_enable_cap(struct kvm_vm *vm, uint32_t cap, uint64_t arg0)
388 {
389 	struct kvm_enable_cap enable_cap = { .cap = cap, .args = { arg0 } };
390 
391 	vm_ioctl(vm, KVM_ENABLE_CAP, &enable_cap);
392 }
393 
394 static inline void vm_set_memory_attributes(struct kvm_vm *vm, uint64_t gpa,
395 					    uint64_t size, uint64_t attributes)
396 {
397 	struct kvm_memory_attributes attr = {
398 		.attributes = attributes,
399 		.address = gpa,
400 		.size = size,
401 		.flags = 0,
402 	};
403 
404 	/*
405 	 * KVM_SET_MEMORY_ATTRIBUTES overwrites _all_ attributes.  These flows
406 	 * need significant enhancements to support multiple attributes.
407 	 */
408 	TEST_ASSERT(!attributes || attributes == KVM_MEMORY_ATTRIBUTE_PRIVATE,
409 		    "Update me to support multiple attributes!");
410 
411 	vm_ioctl(vm, KVM_SET_MEMORY_ATTRIBUTES, &attr);
412 }
413 
414 
415 static inline void vm_mem_set_private(struct kvm_vm *vm, uint64_t gpa,
416 				      uint64_t size)
417 {
418 	vm_set_memory_attributes(vm, gpa, size, KVM_MEMORY_ATTRIBUTE_PRIVATE);
419 }
420 
421 static inline void vm_mem_set_shared(struct kvm_vm *vm, uint64_t gpa,
422 				     uint64_t size)
423 {
424 	vm_set_memory_attributes(vm, gpa, size, 0);
425 }
426 
427 void vm_guest_mem_fallocate(struct kvm_vm *vm, uint64_t gpa, uint64_t size,
428 			    bool punch_hole);
429 
430 static inline void vm_guest_mem_punch_hole(struct kvm_vm *vm, uint64_t gpa,
431 					   uint64_t size)
432 {
433 	vm_guest_mem_fallocate(vm, gpa, size, true);
434 }
435 
436 static inline void vm_guest_mem_allocate(struct kvm_vm *vm, uint64_t gpa,
437 					 uint64_t size)
438 {
439 	vm_guest_mem_fallocate(vm, gpa, size, false);
440 }
441 
442 void vm_enable_dirty_ring(struct kvm_vm *vm, uint32_t ring_size);
443 const char *vm_guest_mode_string(uint32_t i);
444 
445 void kvm_vm_free(struct kvm_vm *vmp);
446 void kvm_vm_restart(struct kvm_vm *vmp);
447 void kvm_vm_release(struct kvm_vm *vmp);
448 int kvm_memcmp_hva_gva(void *hva, struct kvm_vm *vm, const vm_vaddr_t gva,
449 		       size_t len);
450 void kvm_vm_elf_load(struct kvm_vm *vm, const char *filename);
451 int kvm_memfd_alloc(size_t size, bool hugepages);
452 
453 void vm_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent);
454 
455 static inline void kvm_vm_get_dirty_log(struct kvm_vm *vm, int slot, void *log)
456 {
457 	struct kvm_dirty_log args = { .dirty_bitmap = log, .slot = slot };
458 
459 	vm_ioctl(vm, KVM_GET_DIRTY_LOG, &args);
460 }
461 
462 static inline void kvm_vm_clear_dirty_log(struct kvm_vm *vm, int slot, void *log,
463 					  uint64_t first_page, uint32_t num_pages)
464 {
465 	struct kvm_clear_dirty_log args = {
466 		.dirty_bitmap = log,
467 		.slot = slot,
468 		.first_page = first_page,
469 		.num_pages = num_pages
470 	};
471 
472 	vm_ioctl(vm, KVM_CLEAR_DIRTY_LOG, &args);
473 }
474 
475 static inline uint32_t kvm_vm_reset_dirty_ring(struct kvm_vm *vm)
476 {
477 	return __vm_ioctl(vm, KVM_RESET_DIRTY_RINGS, NULL);
478 }
479 
480 static inline int vm_get_stats_fd(struct kvm_vm *vm)
481 {
482 	int fd = __vm_ioctl(vm, KVM_GET_STATS_FD, NULL);
483 
484 	TEST_ASSERT_VM_VCPU_IOCTL(fd >= 0, KVM_GET_STATS_FD, fd, vm);
485 	return fd;
486 }
487 
488 static inline void read_stats_header(int stats_fd, struct kvm_stats_header *header)
489 {
490 	ssize_t ret;
491 
492 	ret = pread(stats_fd, header, sizeof(*header), 0);
493 	TEST_ASSERT(ret == sizeof(*header),
494 		    "Failed to read '%lu' header bytes, ret = '%ld'",
495 		    sizeof(*header), ret);
496 }
497 
498 struct kvm_stats_desc *read_stats_descriptors(int stats_fd,
499 					      struct kvm_stats_header *header);
500 
501 static inline ssize_t get_stats_descriptor_size(struct kvm_stats_header *header)
502 {
503 	 /*
504 	  * The base size of the descriptor is defined by KVM's ABI, but the
505 	  * size of the name field is variable, as far as KVM's ABI is
506 	  * concerned. For a given instance of KVM, the name field is the same
507 	  * size for all stats and is provided in the overall stats header.
508 	  */
509 	return sizeof(struct kvm_stats_desc) + header->name_size;
510 }
511 
512 static inline struct kvm_stats_desc *get_stats_descriptor(struct kvm_stats_desc *stats,
513 							  int index,
514 							  struct kvm_stats_header *header)
515 {
516 	/*
517 	 * Note, size_desc includes the size of the name field, which is
518 	 * variable. i.e. this is NOT equivalent to &stats_desc[i].
519 	 */
520 	return (void *)stats + index * get_stats_descriptor_size(header);
521 }
522 
523 void read_stat_data(int stats_fd, struct kvm_stats_header *header,
524 		    struct kvm_stats_desc *desc, uint64_t *data,
525 		    size_t max_elements);
526 
527 void __vm_get_stat(struct kvm_vm *vm, const char *stat_name, uint64_t *data,
528 		   size_t max_elements);
529 
530 static inline uint64_t vm_get_stat(struct kvm_vm *vm, const char *stat_name)
531 {
532 	uint64_t data;
533 
534 	__vm_get_stat(vm, stat_name, &data, 1);
535 	return data;
536 }
537 
538 void vm_create_irqchip(struct kvm_vm *vm);
539 
540 static inline int __vm_create_guest_memfd(struct kvm_vm *vm, uint64_t size,
541 					uint64_t flags)
542 {
543 	struct kvm_create_guest_memfd guest_memfd = {
544 		.size = size,
545 		.flags = flags,
546 	};
547 
548 	return __vm_ioctl(vm, KVM_CREATE_GUEST_MEMFD, &guest_memfd);
549 }
550 
551 static inline int vm_create_guest_memfd(struct kvm_vm *vm, uint64_t size,
552 					uint64_t flags)
553 {
554 	int fd = __vm_create_guest_memfd(vm, size, flags);
555 
556 	TEST_ASSERT(fd >= 0, KVM_IOCTL_ERROR(KVM_CREATE_GUEST_MEMFD, fd));
557 	return fd;
558 }
559 
560 void vm_set_user_memory_region(struct kvm_vm *vm, uint32_t slot, uint32_t flags,
561 			       uint64_t gpa, uint64_t size, void *hva);
562 int __vm_set_user_memory_region(struct kvm_vm *vm, uint32_t slot, uint32_t flags,
563 				uint64_t gpa, uint64_t size, void *hva);
564 void vm_set_user_memory_region2(struct kvm_vm *vm, uint32_t slot, uint32_t flags,
565 				uint64_t gpa, uint64_t size, void *hva,
566 				uint32_t guest_memfd, uint64_t guest_memfd_offset);
567 int __vm_set_user_memory_region2(struct kvm_vm *vm, uint32_t slot, uint32_t flags,
568 				 uint64_t gpa, uint64_t size, void *hva,
569 				 uint32_t guest_memfd, uint64_t guest_memfd_offset);
570 
571 void vm_userspace_mem_region_add(struct kvm_vm *vm,
572 	enum vm_mem_backing_src_type src_type,
573 	uint64_t guest_paddr, uint32_t slot, uint64_t npages,
574 	uint32_t flags);
575 void vm_mem_add(struct kvm_vm *vm, enum vm_mem_backing_src_type src_type,
576 		uint64_t guest_paddr, uint32_t slot, uint64_t npages,
577 		uint32_t flags, int guest_memfd_fd, uint64_t guest_memfd_offset);
578 
579 #ifndef vm_arch_has_protected_memory
580 static inline bool vm_arch_has_protected_memory(struct kvm_vm *vm)
581 {
582 	return false;
583 }
584 #endif
585 
586 void vm_mem_region_set_flags(struct kvm_vm *vm, uint32_t slot, uint32_t flags);
587 void vm_mem_region_move(struct kvm_vm *vm, uint32_t slot, uint64_t new_gpa);
588 void vm_mem_region_delete(struct kvm_vm *vm, uint32_t slot);
589 struct kvm_vcpu *__vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpu_id);
590 void vm_populate_vaddr_bitmap(struct kvm_vm *vm);
591 vm_vaddr_t vm_vaddr_unused_gap(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min);
592 vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min);
593 vm_vaddr_t __vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
594 			    enum kvm_mem_region_type type);
595 vm_vaddr_t vm_vaddr_alloc_shared(struct kvm_vm *vm, size_t sz,
596 				 vm_vaddr_t vaddr_min,
597 				 enum kvm_mem_region_type type);
598 vm_vaddr_t vm_vaddr_alloc_pages(struct kvm_vm *vm, int nr_pages);
599 vm_vaddr_t __vm_vaddr_alloc_page(struct kvm_vm *vm,
600 				 enum kvm_mem_region_type type);
601 vm_vaddr_t vm_vaddr_alloc_page(struct kvm_vm *vm);
602 
603 void virt_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr,
604 	      unsigned int npages);
605 void *addr_gpa2hva(struct kvm_vm *vm, vm_paddr_t gpa);
606 void *addr_gva2hva(struct kvm_vm *vm, vm_vaddr_t gva);
607 vm_paddr_t addr_hva2gpa(struct kvm_vm *vm, void *hva);
608 void *addr_gpa2alias(struct kvm_vm *vm, vm_paddr_t gpa);
609 
610 #ifndef vcpu_arch_put_guest
611 #define vcpu_arch_put_guest(mem, val) do { (mem) = (val); } while (0)
612 #endif
613 
614 static inline vm_paddr_t vm_untag_gpa(struct kvm_vm *vm, vm_paddr_t gpa)
615 {
616 	return gpa & ~vm->gpa_tag_mask;
617 }
618 
619 void vcpu_run(struct kvm_vcpu *vcpu);
620 int _vcpu_run(struct kvm_vcpu *vcpu);
621 
622 static inline int __vcpu_run(struct kvm_vcpu *vcpu)
623 {
624 	return __vcpu_ioctl(vcpu, KVM_RUN, NULL);
625 }
626 
627 void vcpu_run_complete_io(struct kvm_vcpu *vcpu);
628 struct kvm_reg_list *vcpu_get_reg_list(struct kvm_vcpu *vcpu);
629 
630 static inline void vcpu_enable_cap(struct kvm_vcpu *vcpu, uint32_t cap,
631 				   uint64_t arg0)
632 {
633 	struct kvm_enable_cap enable_cap = { .cap = cap, .args = { arg0 } };
634 
635 	vcpu_ioctl(vcpu, KVM_ENABLE_CAP, &enable_cap);
636 }
637 
638 static inline void vcpu_guest_debug_set(struct kvm_vcpu *vcpu,
639 					struct kvm_guest_debug *debug)
640 {
641 	vcpu_ioctl(vcpu, KVM_SET_GUEST_DEBUG, debug);
642 }
643 
644 static inline void vcpu_mp_state_get(struct kvm_vcpu *vcpu,
645 				     struct kvm_mp_state *mp_state)
646 {
647 	vcpu_ioctl(vcpu, KVM_GET_MP_STATE, mp_state);
648 }
649 static inline void vcpu_mp_state_set(struct kvm_vcpu *vcpu,
650 				     struct kvm_mp_state *mp_state)
651 {
652 	vcpu_ioctl(vcpu, KVM_SET_MP_STATE, mp_state);
653 }
654 
655 static inline void vcpu_regs_get(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
656 {
657 	vcpu_ioctl(vcpu, KVM_GET_REGS, regs);
658 }
659 
660 static inline void vcpu_regs_set(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
661 {
662 	vcpu_ioctl(vcpu, KVM_SET_REGS, regs);
663 }
664 static inline void vcpu_sregs_get(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
665 {
666 	vcpu_ioctl(vcpu, KVM_GET_SREGS, sregs);
667 
668 }
669 static inline void vcpu_sregs_set(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
670 {
671 	vcpu_ioctl(vcpu, KVM_SET_SREGS, sregs);
672 }
673 static inline int _vcpu_sregs_set(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
674 {
675 	return __vcpu_ioctl(vcpu, KVM_SET_SREGS, sregs);
676 }
677 static inline void vcpu_fpu_get(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
678 {
679 	vcpu_ioctl(vcpu, KVM_GET_FPU, fpu);
680 }
681 static inline void vcpu_fpu_set(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
682 {
683 	vcpu_ioctl(vcpu, KVM_SET_FPU, fpu);
684 }
685 
686 static inline int __vcpu_get_reg(struct kvm_vcpu *vcpu, uint64_t id, void *addr)
687 {
688 	struct kvm_one_reg reg = { .id = id, .addr = (uint64_t)addr };
689 
690 	return __vcpu_ioctl(vcpu, KVM_GET_ONE_REG, &reg);
691 }
692 static inline int __vcpu_set_reg(struct kvm_vcpu *vcpu, uint64_t id, uint64_t val)
693 {
694 	struct kvm_one_reg reg = { .id = id, .addr = (uint64_t)&val };
695 
696 	return __vcpu_ioctl(vcpu, KVM_SET_ONE_REG, &reg);
697 }
698 static inline void vcpu_get_reg(struct kvm_vcpu *vcpu, uint64_t id, void *addr)
699 {
700 	struct kvm_one_reg reg = { .id = id, .addr = (uint64_t)addr };
701 
702 	vcpu_ioctl(vcpu, KVM_GET_ONE_REG, &reg);
703 }
704 static inline void vcpu_set_reg(struct kvm_vcpu *vcpu, uint64_t id, uint64_t val)
705 {
706 	struct kvm_one_reg reg = { .id = id, .addr = (uint64_t)&val };
707 
708 	vcpu_ioctl(vcpu, KVM_SET_ONE_REG, &reg);
709 }
710 
711 #ifdef __KVM_HAVE_VCPU_EVENTS
712 static inline void vcpu_events_get(struct kvm_vcpu *vcpu,
713 				   struct kvm_vcpu_events *events)
714 {
715 	vcpu_ioctl(vcpu, KVM_GET_VCPU_EVENTS, events);
716 }
717 static inline void vcpu_events_set(struct kvm_vcpu *vcpu,
718 				   struct kvm_vcpu_events *events)
719 {
720 	vcpu_ioctl(vcpu, KVM_SET_VCPU_EVENTS, events);
721 }
722 #endif
723 #ifdef __x86_64__
724 static inline void vcpu_nested_state_get(struct kvm_vcpu *vcpu,
725 					 struct kvm_nested_state *state)
726 {
727 	vcpu_ioctl(vcpu, KVM_GET_NESTED_STATE, state);
728 }
729 static inline int __vcpu_nested_state_set(struct kvm_vcpu *vcpu,
730 					  struct kvm_nested_state *state)
731 {
732 	return __vcpu_ioctl(vcpu, KVM_SET_NESTED_STATE, state);
733 }
734 
735 static inline void vcpu_nested_state_set(struct kvm_vcpu *vcpu,
736 					 struct kvm_nested_state *state)
737 {
738 	vcpu_ioctl(vcpu, KVM_SET_NESTED_STATE, state);
739 }
740 #endif
741 static inline int vcpu_get_stats_fd(struct kvm_vcpu *vcpu)
742 {
743 	int fd = __vcpu_ioctl(vcpu, KVM_GET_STATS_FD, NULL);
744 
745 	TEST_ASSERT_VM_VCPU_IOCTL(fd >= 0, KVM_CHECK_EXTENSION, fd, vcpu->vm);
746 	return fd;
747 }
748 
749 int __kvm_has_device_attr(int dev_fd, uint32_t group, uint64_t attr);
750 
751 static inline void kvm_has_device_attr(int dev_fd, uint32_t group, uint64_t attr)
752 {
753 	int ret = __kvm_has_device_attr(dev_fd, group, attr);
754 
755 	TEST_ASSERT(!ret, "KVM_HAS_DEVICE_ATTR failed, rc: %i errno: %i", ret, errno);
756 }
757 
758 int __kvm_device_attr_get(int dev_fd, uint32_t group, uint64_t attr, void *val);
759 
760 static inline void kvm_device_attr_get(int dev_fd, uint32_t group,
761 				       uint64_t attr, void *val)
762 {
763 	int ret = __kvm_device_attr_get(dev_fd, group, attr, val);
764 
765 	TEST_ASSERT(!ret, KVM_IOCTL_ERROR(KVM_GET_DEVICE_ATTR, ret));
766 }
767 
768 int __kvm_device_attr_set(int dev_fd, uint32_t group, uint64_t attr, void *val);
769 
770 static inline void kvm_device_attr_set(int dev_fd, uint32_t group,
771 				       uint64_t attr, void *val)
772 {
773 	int ret = __kvm_device_attr_set(dev_fd, group, attr, val);
774 
775 	TEST_ASSERT(!ret, KVM_IOCTL_ERROR(KVM_SET_DEVICE_ATTR, ret));
776 }
777 
778 static inline int __vcpu_has_device_attr(struct kvm_vcpu *vcpu, uint32_t group,
779 					 uint64_t attr)
780 {
781 	return __kvm_has_device_attr(vcpu->fd, group, attr);
782 }
783 
784 static inline void vcpu_has_device_attr(struct kvm_vcpu *vcpu, uint32_t group,
785 					uint64_t attr)
786 {
787 	kvm_has_device_attr(vcpu->fd, group, attr);
788 }
789 
790 static inline int __vcpu_device_attr_get(struct kvm_vcpu *vcpu, uint32_t group,
791 					 uint64_t attr, void *val)
792 {
793 	return __kvm_device_attr_get(vcpu->fd, group, attr, val);
794 }
795 
796 static inline void vcpu_device_attr_get(struct kvm_vcpu *vcpu, uint32_t group,
797 					uint64_t attr, void *val)
798 {
799 	kvm_device_attr_get(vcpu->fd, group, attr, val);
800 }
801 
802 static inline int __vcpu_device_attr_set(struct kvm_vcpu *vcpu, uint32_t group,
803 					 uint64_t attr, void *val)
804 {
805 	return __kvm_device_attr_set(vcpu->fd, group, attr, val);
806 }
807 
808 static inline void vcpu_device_attr_set(struct kvm_vcpu *vcpu, uint32_t group,
809 					uint64_t attr, void *val)
810 {
811 	kvm_device_attr_set(vcpu->fd, group, attr, val);
812 }
813 
814 int __kvm_test_create_device(struct kvm_vm *vm, uint64_t type);
815 int __kvm_create_device(struct kvm_vm *vm, uint64_t type);
816 
817 static inline int kvm_create_device(struct kvm_vm *vm, uint64_t type)
818 {
819 	int fd = __kvm_create_device(vm, type);
820 
821 	TEST_ASSERT(fd >= 0, KVM_IOCTL_ERROR(KVM_CREATE_DEVICE, fd));
822 	return fd;
823 }
824 
825 void *vcpu_map_dirty_ring(struct kvm_vcpu *vcpu);
826 
827 /*
828  * VM VCPU Args Set
829  *
830  * Input Args:
831  *   vm - Virtual Machine
832  *   num - number of arguments
833  *   ... - arguments, each of type uint64_t
834  *
835  * Output Args: None
836  *
837  * Return: None
838  *
839  * Sets the first @num input parameters for the function at @vcpu's entry point,
840  * per the C calling convention of the architecture, to the values given as
841  * variable args. Each of the variable args is expected to be of type uint64_t.
842  * The maximum @num can be is specific to the architecture.
843  */
844 void vcpu_args_set(struct kvm_vcpu *vcpu, unsigned int num, ...);
845 
846 void kvm_irq_line(struct kvm_vm *vm, uint32_t irq, int level);
847 int _kvm_irq_line(struct kvm_vm *vm, uint32_t irq, int level);
848 
849 #define KVM_MAX_IRQ_ROUTES		4096
850 
851 struct kvm_irq_routing *kvm_gsi_routing_create(void);
852 void kvm_gsi_routing_irqchip_add(struct kvm_irq_routing *routing,
853 		uint32_t gsi, uint32_t pin);
854 int _kvm_gsi_routing_write(struct kvm_vm *vm, struct kvm_irq_routing *routing);
855 void kvm_gsi_routing_write(struct kvm_vm *vm, struct kvm_irq_routing *routing);
856 
857 const char *exit_reason_str(unsigned int exit_reason);
858 
859 vm_paddr_t vm_phy_page_alloc(struct kvm_vm *vm, vm_paddr_t paddr_min,
860 			     uint32_t memslot);
861 vm_paddr_t __vm_phy_pages_alloc(struct kvm_vm *vm, size_t num,
862 				vm_paddr_t paddr_min, uint32_t memslot,
863 				bool protected);
864 vm_paddr_t vm_alloc_page_table(struct kvm_vm *vm);
865 
866 static inline vm_paddr_t vm_phy_pages_alloc(struct kvm_vm *vm, size_t num,
867 					    vm_paddr_t paddr_min, uint32_t memslot)
868 {
869 	/*
870 	 * By default, allocate memory as protected for VMs that support
871 	 * protected memory, as the majority of memory for such VMs is
872 	 * protected, i.e. using shared memory is effectively opt-in.
873 	 */
874 	return __vm_phy_pages_alloc(vm, num, paddr_min, memslot,
875 				    vm_arch_has_protected_memory(vm));
876 }
877 
878 /*
879  * ____vm_create() does KVM_CREATE_VM and little else.  __vm_create() also
880  * loads the test binary into guest memory and creates an IRQ chip (x86 only).
881  * __vm_create() does NOT create vCPUs, @nr_runnable_vcpus is used purely to
882  * calculate the amount of memory needed for per-vCPU data, e.g. stacks.
883  */
884 struct kvm_vm *____vm_create(struct vm_shape shape);
885 struct kvm_vm *__vm_create(struct vm_shape shape, uint32_t nr_runnable_vcpus,
886 			   uint64_t nr_extra_pages);
887 
888 static inline struct kvm_vm *vm_create_barebones(void)
889 {
890 	return ____vm_create(VM_SHAPE_DEFAULT);
891 }
892 
893 static inline struct kvm_vm *vm_create_barebones_type(unsigned long type)
894 {
895 	const struct vm_shape shape = {
896 		.mode = VM_MODE_DEFAULT,
897 		.type = type,
898 	};
899 
900 	return ____vm_create(shape);
901 }
902 
903 static inline struct kvm_vm *vm_create(uint32_t nr_runnable_vcpus)
904 {
905 	return __vm_create(VM_SHAPE_DEFAULT, nr_runnable_vcpus, 0);
906 }
907 
908 struct kvm_vm *__vm_create_with_vcpus(struct vm_shape shape, uint32_t nr_vcpus,
909 				      uint64_t extra_mem_pages,
910 				      void *guest_code, struct kvm_vcpu *vcpus[]);
911 
912 static inline struct kvm_vm *vm_create_with_vcpus(uint32_t nr_vcpus,
913 						  void *guest_code,
914 						  struct kvm_vcpu *vcpus[])
915 {
916 	return __vm_create_with_vcpus(VM_SHAPE_DEFAULT, nr_vcpus, 0,
917 				      guest_code, vcpus);
918 }
919 
920 
921 struct kvm_vm *__vm_create_shape_with_one_vcpu(struct vm_shape shape,
922 					       struct kvm_vcpu **vcpu,
923 					       uint64_t extra_mem_pages,
924 					       void *guest_code);
925 
926 /*
927  * Create a VM with a single vCPU with reasonable defaults and @extra_mem_pages
928  * additional pages of guest memory.  Returns the VM and vCPU (via out param).
929  */
930 static inline struct kvm_vm *__vm_create_with_one_vcpu(struct kvm_vcpu **vcpu,
931 						       uint64_t extra_mem_pages,
932 						       void *guest_code)
933 {
934 	return __vm_create_shape_with_one_vcpu(VM_SHAPE_DEFAULT, vcpu,
935 					       extra_mem_pages, guest_code);
936 }
937 
938 static inline struct kvm_vm *vm_create_with_one_vcpu(struct kvm_vcpu **vcpu,
939 						     void *guest_code)
940 {
941 	return __vm_create_with_one_vcpu(vcpu, 0, guest_code);
942 }
943 
944 static inline struct kvm_vm *vm_create_shape_with_one_vcpu(struct vm_shape shape,
945 							   struct kvm_vcpu **vcpu,
946 							   void *guest_code)
947 {
948 	return __vm_create_shape_with_one_vcpu(shape, vcpu, 0, guest_code);
949 }
950 
951 struct kvm_vcpu *vm_recreate_with_one_vcpu(struct kvm_vm *vm);
952 
953 void kvm_pin_this_task_to_pcpu(uint32_t pcpu);
954 void kvm_print_vcpu_pinning_help(void);
955 void kvm_parse_vcpu_pinning(const char *pcpus_string, uint32_t vcpu_to_pcpu[],
956 			    int nr_vcpus);
957 
958 unsigned long vm_compute_max_gfn(struct kvm_vm *vm);
959 unsigned int vm_calc_num_guest_pages(enum vm_guest_mode mode, size_t size);
960 unsigned int vm_num_host_pages(enum vm_guest_mode mode, unsigned int num_guest_pages);
961 unsigned int vm_num_guest_pages(enum vm_guest_mode mode, unsigned int num_host_pages);
962 static inline unsigned int
963 vm_adjust_num_guest_pages(enum vm_guest_mode mode, unsigned int num_guest_pages)
964 {
965 	unsigned int n;
966 	n = vm_num_guest_pages(mode, vm_num_host_pages(mode, num_guest_pages));
967 #ifdef __s390x__
968 	/* s390 requires 1M aligned guest sizes */
969 	n = (n + 255) & ~255;
970 #endif
971 	return n;
972 }
973 
974 #define sync_global_to_guest(vm, g) ({				\
975 	typeof(g) *_p = addr_gva2hva(vm, (vm_vaddr_t)&(g));	\
976 	memcpy(_p, &(g), sizeof(g));				\
977 })
978 
979 #define sync_global_from_guest(vm, g) ({			\
980 	typeof(g) *_p = addr_gva2hva(vm, (vm_vaddr_t)&(g));	\
981 	memcpy(&(g), _p, sizeof(g));				\
982 })
983 
984 /*
985  * Write a global value, but only in the VM's (guest's) domain.  Primarily used
986  * for "globals" that hold per-VM values (VMs always duplicate code and global
987  * data into their own region of physical memory), but can be used anytime it's
988  * undesirable to change the host's copy of the global.
989  */
990 #define write_guest_global(vm, g, val) ({			\
991 	typeof(g) *_p = addr_gva2hva(vm, (vm_vaddr_t)&(g));	\
992 	typeof(g) _val = val;					\
993 								\
994 	memcpy(_p, &(_val), sizeof(g));				\
995 })
996 
997 void assert_on_unhandled_exception(struct kvm_vcpu *vcpu);
998 
999 void vcpu_arch_dump(FILE *stream, struct kvm_vcpu *vcpu,
1000 		    uint8_t indent);
1001 
1002 static inline void vcpu_dump(FILE *stream, struct kvm_vcpu *vcpu,
1003 			     uint8_t indent)
1004 {
1005 	vcpu_arch_dump(stream, vcpu, indent);
1006 }
1007 
1008 /*
1009  * Adds a vCPU with reasonable defaults (e.g. a stack)
1010  *
1011  * Input Args:
1012  *   vm - Virtual Machine
1013  *   vcpu_id - The id of the VCPU to add to the VM.
1014  */
1015 struct kvm_vcpu *vm_arch_vcpu_add(struct kvm_vm *vm, uint32_t vcpu_id);
1016 void vcpu_arch_set_entry_point(struct kvm_vcpu *vcpu, void *guest_code);
1017 
1018 static inline struct kvm_vcpu *vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpu_id,
1019 					   void *guest_code)
1020 {
1021 	struct kvm_vcpu *vcpu = vm_arch_vcpu_add(vm, vcpu_id);
1022 
1023 	vcpu_arch_set_entry_point(vcpu, guest_code);
1024 
1025 	return vcpu;
1026 }
1027 
1028 /* Re-create a vCPU after restarting a VM, e.g. for state save/restore tests. */
1029 struct kvm_vcpu *vm_arch_vcpu_recreate(struct kvm_vm *vm, uint32_t vcpu_id);
1030 
1031 static inline struct kvm_vcpu *vm_vcpu_recreate(struct kvm_vm *vm,
1032 						uint32_t vcpu_id)
1033 {
1034 	return vm_arch_vcpu_recreate(vm, vcpu_id);
1035 }
1036 
1037 void vcpu_arch_free(struct kvm_vcpu *vcpu);
1038 
1039 void virt_arch_pgd_alloc(struct kvm_vm *vm);
1040 
1041 static inline void virt_pgd_alloc(struct kvm_vm *vm)
1042 {
1043 	virt_arch_pgd_alloc(vm);
1044 }
1045 
1046 /*
1047  * VM Virtual Page Map
1048  *
1049  * Input Args:
1050  *   vm - Virtual Machine
1051  *   vaddr - VM Virtual Address
1052  *   paddr - VM Physical Address
1053  *   memslot - Memory region slot for new virtual translation tables
1054  *
1055  * Output Args: None
1056  *
1057  * Return: None
1058  *
1059  * Within @vm, creates a virtual translation for the page starting
1060  * at @vaddr to the page starting at @paddr.
1061  */
1062 void virt_arch_pg_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr);
1063 
1064 static inline void virt_pg_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr)
1065 {
1066 	virt_arch_pg_map(vm, vaddr, paddr);
1067 }
1068 
1069 
1070 /*
1071  * Address Guest Virtual to Guest Physical
1072  *
1073  * Input Args:
1074  *   vm - Virtual Machine
1075  *   gva - VM virtual address
1076  *
1077  * Output Args: None
1078  *
1079  * Return:
1080  *   Equivalent VM physical address
1081  *
1082  * Returns the VM physical address of the translated VM virtual
1083  * address given by @gva.
1084  */
1085 vm_paddr_t addr_arch_gva2gpa(struct kvm_vm *vm, vm_vaddr_t gva);
1086 
1087 static inline vm_paddr_t addr_gva2gpa(struct kvm_vm *vm, vm_vaddr_t gva)
1088 {
1089 	return addr_arch_gva2gpa(vm, gva);
1090 }
1091 
1092 /*
1093  * Virtual Translation Tables Dump
1094  *
1095  * Input Args:
1096  *   stream - Output FILE stream
1097  *   vm     - Virtual Machine
1098  *   indent - Left margin indent amount
1099  *
1100  * Output Args: None
1101  *
1102  * Return: None
1103  *
1104  * Dumps to the FILE stream given by @stream, the contents of all the
1105  * virtual translation tables for the VM given by @vm.
1106  */
1107 void virt_arch_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent);
1108 
1109 static inline void virt_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
1110 {
1111 	virt_arch_dump(stream, vm, indent);
1112 }
1113 
1114 
1115 static inline int __vm_disable_nx_huge_pages(struct kvm_vm *vm)
1116 {
1117 	return __vm_enable_cap(vm, KVM_CAP_VM_DISABLE_NX_HUGE_PAGES, 0);
1118 }
1119 
1120 /*
1121  * Arch hook that is invoked via a constructor, i.e. before exeucting main(),
1122  * to allow for arch-specific setup that is common to all tests, e.g. computing
1123  * the default guest "mode".
1124  */
1125 void kvm_selftest_arch_init(void);
1126 
1127 void kvm_arch_vm_post_create(struct kvm_vm *vm);
1128 
1129 bool vm_is_gpa_protected(struct kvm_vm *vm, vm_paddr_t paddr);
1130 
1131 uint32_t guest_get_vcpuid(void);
1132 
1133 #endif /* SELFTEST_KVM_UTIL_H */
1134