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