xref: /freebsd/sys/amd64/include/vmm.h (revision bd1da0a002e9a43cfb5220835c7a42804d90dc56)
1 /*-
2  * Copyright (c) 2011 NetApp, Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  */
28 
29 #ifndef _VMM_H_
30 #define	_VMM_H_
31 
32 #include <x86/segments.h>
33 
34 enum vm_suspend_how {
35 	VM_SUSPEND_NONE,
36 	VM_SUSPEND_RESET,
37 	VM_SUSPEND_POWEROFF,
38 	VM_SUSPEND_HALT,
39 	VM_SUSPEND_TRIPLEFAULT,
40 	VM_SUSPEND_LAST
41 };
42 
43 /*
44  * Identifiers for architecturally defined registers.
45  */
46 enum vm_reg_name {
47 	VM_REG_GUEST_RAX,
48 	VM_REG_GUEST_RBX,
49 	VM_REG_GUEST_RCX,
50 	VM_REG_GUEST_RDX,
51 	VM_REG_GUEST_RSI,
52 	VM_REG_GUEST_RDI,
53 	VM_REG_GUEST_RBP,
54 	VM_REG_GUEST_R8,
55 	VM_REG_GUEST_R9,
56 	VM_REG_GUEST_R10,
57 	VM_REG_GUEST_R11,
58 	VM_REG_GUEST_R12,
59 	VM_REG_GUEST_R13,
60 	VM_REG_GUEST_R14,
61 	VM_REG_GUEST_R15,
62 	VM_REG_GUEST_CR0,
63 	VM_REG_GUEST_CR3,
64 	VM_REG_GUEST_CR4,
65 	VM_REG_GUEST_DR7,
66 	VM_REG_GUEST_RSP,
67 	VM_REG_GUEST_RIP,
68 	VM_REG_GUEST_RFLAGS,
69 	VM_REG_GUEST_ES,
70 	VM_REG_GUEST_CS,
71 	VM_REG_GUEST_SS,
72 	VM_REG_GUEST_DS,
73 	VM_REG_GUEST_FS,
74 	VM_REG_GUEST_GS,
75 	VM_REG_GUEST_LDTR,
76 	VM_REG_GUEST_TR,
77 	VM_REG_GUEST_IDTR,
78 	VM_REG_GUEST_GDTR,
79 	VM_REG_GUEST_EFER,
80 	VM_REG_GUEST_CR2,
81 	VM_REG_GUEST_PDPTE0,
82 	VM_REG_GUEST_PDPTE1,
83 	VM_REG_GUEST_PDPTE2,
84 	VM_REG_GUEST_PDPTE3,
85 	VM_REG_GUEST_INTR_SHADOW,
86 	VM_REG_LAST
87 };
88 
89 enum x2apic_state {
90 	X2APIC_DISABLED,
91 	X2APIC_ENABLED,
92 	X2APIC_STATE_LAST
93 };
94 
95 #define	VM_INTINFO_VECTOR(info)	((info) & 0xff)
96 #define	VM_INTINFO_DEL_ERRCODE	0x800
97 #define	VM_INTINFO_RSVD		0x7ffff000
98 #define	VM_INTINFO_VALID	0x80000000
99 #define	VM_INTINFO_TYPE		0x700
100 #define	VM_INTINFO_HWINTR	(0 << 8)
101 #define	VM_INTINFO_NMI		(2 << 8)
102 #define	VM_INTINFO_HWEXCEPTION	(3 << 8)
103 #define	VM_INTINFO_SWINTR	(4 << 8)
104 
105 #ifdef _KERNEL
106 
107 #define	VM_MAX_NAMELEN	32
108 
109 struct vm;
110 struct vm_exception;
111 struct vm_memory_segment;
112 struct seg_desc;
113 struct vm_exit;
114 struct vm_run;
115 struct vhpet;
116 struct vioapic;
117 struct vlapic;
118 struct vmspace;
119 struct vm_object;
120 struct vm_guest_paging;
121 struct pmap;
122 
123 typedef int	(*vmm_init_func_t)(int ipinum);
124 typedef int	(*vmm_cleanup_func_t)(void);
125 typedef void	(*vmm_resume_func_t)(void);
126 typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
127 typedef int	(*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
128 				  struct pmap *pmap, void *rendezvous_cookie,
129 				  void *suspend_cookie);
130 typedef void	(*vmi_cleanup_func_t)(void *vmi);
131 typedef int	(*vmi_get_register_t)(void *vmi, int vcpu, int num,
132 				      uint64_t *retval);
133 typedef int	(*vmi_set_register_t)(void *vmi, int vcpu, int num,
134 				      uint64_t val);
135 typedef int	(*vmi_get_desc_t)(void *vmi, int vcpu, int num,
136 				  struct seg_desc *desc);
137 typedef int	(*vmi_set_desc_t)(void *vmi, int vcpu, int num,
138 				  struct seg_desc *desc);
139 typedef int	(*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
140 typedef int	(*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
141 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
142 typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace);
143 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
144 typedef void	(*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
145 
146 struct vmm_ops {
147 	vmm_init_func_t		init;		/* module wide initialization */
148 	vmm_cleanup_func_t	cleanup;
149 	vmm_resume_func_t	resume;
150 
151 	vmi_init_func_t		vminit;		/* vm-specific initialization */
152 	vmi_run_func_t		vmrun;
153 	vmi_cleanup_func_t	vmcleanup;
154 	vmi_get_register_t	vmgetreg;
155 	vmi_set_register_t	vmsetreg;
156 	vmi_get_desc_t		vmgetdesc;
157 	vmi_set_desc_t		vmsetdesc;
158 	vmi_get_cap_t		vmgetcap;
159 	vmi_set_cap_t		vmsetcap;
160 	vmi_vmspace_alloc	vmspace_alloc;
161 	vmi_vmspace_free	vmspace_free;
162 	vmi_vlapic_init		vlapic_init;
163 	vmi_vlapic_cleanup	vlapic_cleanup;
164 };
165 
166 extern struct vmm_ops vmm_ops_intel;
167 extern struct vmm_ops vmm_ops_amd;
168 
169 int vm_create(const char *name, struct vm **retvm);
170 void vm_destroy(struct vm *vm);
171 int vm_reinit(struct vm *vm);
172 const char *vm_name(struct vm *vm);
173 int vm_malloc(struct vm *vm, vm_paddr_t gpa, size_t len);
174 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
175 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
176 void *vm_gpa_hold(struct vm *, vm_paddr_t gpa, size_t len, int prot,
177 		  void **cookie);
178 void vm_gpa_release(void *cookie);
179 int vm_gpabase2memseg(struct vm *vm, vm_paddr_t gpabase,
180 	      struct vm_memory_segment *seg);
181 int vm_get_memobj(struct vm *vm, vm_paddr_t gpa, size_t len,
182 		  vm_offset_t *offset, struct vm_object **object);
183 boolean_t vm_mem_allocated(struct vm *vm, vm_paddr_t gpa);
184 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
185 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
186 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
187 		    struct seg_desc *ret_desc);
188 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
189 		    struct seg_desc *desc);
190 int vm_run(struct vm *vm, struct vm_run *vmrun);
191 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
192 int vm_inject_nmi(struct vm *vm, int vcpu);
193 int vm_nmi_pending(struct vm *vm, int vcpuid);
194 void vm_nmi_clear(struct vm *vm, int vcpuid);
195 int vm_inject_extint(struct vm *vm, int vcpu);
196 int vm_extint_pending(struct vm *vm, int vcpuid);
197 void vm_extint_clear(struct vm *vm, int vcpuid);
198 struct vlapic *vm_lapic(struct vm *vm, int cpu);
199 struct vioapic *vm_ioapic(struct vm *vm);
200 struct vhpet *vm_hpet(struct vm *vm);
201 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
202 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
203 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
204 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
205 int vm_apicid2vcpuid(struct vm *vm, int apicid);
206 int vm_activate_cpu(struct vm *vm, int vcpu);
207 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
208 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
209 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
210 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
211 
212 #ifdef _SYS__CPUSET_H_
213 /*
214  * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
215  * The rendezvous 'func(arg)' is not allowed to do anything that will
216  * cause the thread to be put to sleep.
217  *
218  * If the rendezvous is being initiated from a vcpu context then the
219  * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
220  *
221  * The caller cannot hold any locks when initiating the rendezvous.
222  *
223  * The implementation of this API may cause vcpus other than those specified
224  * by 'dest' to be stalled. The caller should not rely on any vcpus making
225  * forward progress when the rendezvous is in progress.
226  */
227 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
228 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
229     vm_rendezvous_func_t func, void *arg);
230 cpuset_t vm_active_cpus(struct vm *vm);
231 cpuset_t vm_suspended_cpus(struct vm *vm);
232 #endif	/* _SYS__CPUSET_H_ */
233 
234 static __inline int
235 vcpu_rendezvous_pending(void *rendezvous_cookie)
236 {
237 
238 	return (*(uintptr_t *)rendezvous_cookie != 0);
239 }
240 
241 static __inline int
242 vcpu_suspended(void *suspend_cookie)
243 {
244 
245 	return (*(int *)suspend_cookie);
246 }
247 
248 /*
249  * Return 1 if device indicated by bus/slot/func is supposed to be a
250  * pci passthrough device.
251  *
252  * Return 0 otherwise.
253  */
254 int vmm_is_pptdev(int bus, int slot, int func);
255 
256 void *vm_iommu_domain(struct vm *vm);
257 
258 enum vcpu_state {
259 	VCPU_IDLE,
260 	VCPU_FROZEN,
261 	VCPU_RUNNING,
262 	VCPU_SLEEPING,
263 };
264 
265 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
266     bool from_idle);
267 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
268 
269 static int __inline
270 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
271 {
272 	return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
273 }
274 
275 #ifdef _SYS_PROC_H_
276 static int __inline
277 vcpu_should_yield(struct vm *vm, int vcpu)
278 {
279 	return (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED));
280 }
281 #endif
282 
283 void *vcpu_stats(struct vm *vm, int vcpu);
284 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
285 struct vmspace *vm_get_vmspace(struct vm *vm);
286 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
287 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
288 struct vatpic *vm_atpic(struct vm *vm);
289 struct vatpit *vm_atpit(struct vm *vm);
290 struct vpmtmr *vm_pmtmr(struct vm *vm);
291 struct vrtc *vm_rtc(struct vm *vm);
292 
293 /*
294  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
295  * success and non-zero on failure.
296  *
297  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
298  * this function directly because they enforce the trap-like or fault-like
299  * behavior of an exception.
300  *
301  * This function should only be called in the context of the thread that is
302  * executing this vcpu.
303  */
304 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
305     uint32_t errcode, int restart_instruction);
306 
307 /*
308  * This function is called after a VM-exit that occurred during exception or
309  * interrupt delivery through the IDT. The format of 'intinfo' is described
310  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
311  *
312  * If a VM-exit handler completes the event delivery successfully then it
313  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
314  * if the task switch emulation is triggered via a task gate then it should
315  * call this function with 'intinfo=0' to indicate that the external event
316  * is not pending anymore.
317  *
318  * Return value is 0 on success and non-zero on failure.
319  */
320 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
321 
322 /*
323  * This function is called before every VM-entry to retrieve a pending
324  * event that should be injected into the guest. This function combines
325  * nested events into a double or triple fault.
326  *
327  * Returns 0 if there are no events that need to be injected into the guest
328  * and non-zero otherwise.
329  */
330 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
331 
332 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
333 
334 enum vm_reg_name vm_segment_name(int seg_encoding);
335 
336 struct vm_copyinfo {
337 	uint64_t	gpa;
338 	size_t		len;
339 	void		*hva;
340 	void		*cookie;
341 };
342 
343 /*
344  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
345  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
346  * a copyin or PROT_WRITE for a copyout.
347  *
348  * Returns 0 on success.
349  * Returns 1 if an exception was injected into the guest.
350  * Returns -1 otherwise.
351  *
352  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
353  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
354  * 'vm_copy_teardown()' after the copy is done.
355  */
356 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
357     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
358     int num_copyinfo);
359 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
360     int num_copyinfo);
361 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
362     void *kaddr, size_t len);
363 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
364     struct vm_copyinfo *copyinfo, size_t len);
365 
366 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
367 #endif	/* KERNEL */
368 
369 #define	VM_MAXCPU	16			/* maximum virtual cpus */
370 
371 /*
372  * Identifiers for optional vmm capabilities
373  */
374 enum vm_cap_type {
375 	VM_CAP_HALT_EXIT,
376 	VM_CAP_MTRAP_EXIT,
377 	VM_CAP_PAUSE_EXIT,
378 	VM_CAP_UNRESTRICTED_GUEST,
379 	VM_CAP_ENABLE_INVPCID,
380 	VM_CAP_MAX
381 };
382 
383 enum vm_intr_trigger {
384 	EDGE_TRIGGER,
385 	LEVEL_TRIGGER
386 };
387 
388 /*
389  * The 'access' field has the format specified in Table 21-2 of the Intel
390  * Architecture Manual vol 3b.
391  *
392  * XXX The contents of the 'access' field are architecturally defined except
393  * bit 16 - Segment Unusable.
394  */
395 struct seg_desc {
396 	uint64_t	base;
397 	uint32_t	limit;
398 	uint32_t	access;
399 };
400 #define	SEG_DESC_TYPE(access)		((access) & 0x001f)
401 #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3)
402 #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0)
403 #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0)
404 #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0)
405 #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0)
406 
407 enum vm_cpu_mode {
408 	CPU_MODE_REAL,
409 	CPU_MODE_PROTECTED,
410 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
411 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
412 };
413 
414 enum vm_paging_mode {
415 	PAGING_MODE_FLAT,
416 	PAGING_MODE_32,
417 	PAGING_MODE_PAE,
418 	PAGING_MODE_64,
419 };
420 
421 struct vm_guest_paging {
422 	uint64_t	cr3;
423 	int		cpl;
424 	enum vm_cpu_mode cpu_mode;
425 	enum vm_paging_mode paging_mode;
426 };
427 
428 /*
429  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
430  * consumers of instruction decoding. The only reason why their contents
431  * need to be exposed is because they are part of the 'vm_exit' structure.
432  */
433 struct vie_op {
434 	uint8_t		op_byte;	/* actual opcode byte */
435 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
436 	uint16_t	op_flags;
437 };
438 
439 #define	VIE_INST_SIZE	15
440 struct vie {
441 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
442 	uint8_t		num_valid;		/* size of the instruction */
443 	uint8_t		num_processed;
444 
445 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */
446 	uint8_t		rex_w:1,		/* REX prefix */
447 			rex_r:1,
448 			rex_x:1,
449 			rex_b:1,
450 			rex_present:1,
451 			repz_present:1,		/* REP/REPE/REPZ prefix */
452 			repnz_present:1,	/* REPNE/REPNZ prefix */
453 			opsize_override:1,	/* Operand size override */
454 			addrsize_override:1,	/* Address size override */
455 			segment_override:1;	/* Segment override */
456 
457 	uint8_t		mod:2,			/* ModRM byte */
458 			reg:4,
459 			rm:4;
460 
461 	uint8_t		ss:2,			/* SIB byte */
462 			index:4,
463 			base:4;
464 
465 	uint8_t		disp_bytes;
466 	uint8_t		imm_bytes;
467 
468 	uint8_t		scale;
469 	int		base_register;		/* VM_REG_GUEST_xyz */
470 	int		index_register;		/* VM_REG_GUEST_xyz */
471 	int		segment_register;	/* VM_REG_GUEST_xyz */
472 
473 	int64_t		displacement;		/* optional addr displacement */
474 	int64_t		immediate;		/* optional immediate operand */
475 
476 	uint8_t		decoded;	/* set to 1 if successfully decoded */
477 
478 	struct vie_op	op;			/* opcode description */
479 };
480 
481 enum vm_exitcode {
482 	VM_EXITCODE_INOUT,
483 	VM_EXITCODE_VMX,
484 	VM_EXITCODE_BOGUS,
485 	VM_EXITCODE_RDMSR,
486 	VM_EXITCODE_WRMSR,
487 	VM_EXITCODE_HLT,
488 	VM_EXITCODE_MTRAP,
489 	VM_EXITCODE_PAUSE,
490 	VM_EXITCODE_PAGING,
491 	VM_EXITCODE_INST_EMUL,
492 	VM_EXITCODE_SPINUP_AP,
493 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
494 	VM_EXITCODE_RENDEZVOUS,
495 	VM_EXITCODE_IOAPIC_EOI,
496 	VM_EXITCODE_SUSPENDED,
497 	VM_EXITCODE_INOUT_STR,
498 	VM_EXITCODE_TASK_SWITCH,
499 	VM_EXITCODE_MONITOR,
500 	VM_EXITCODE_MWAIT,
501 	VM_EXITCODE_SVM,
502 	VM_EXITCODE_MAX
503 };
504 
505 struct vm_inout {
506 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
507 	uint16_t	in:1;
508 	uint16_t	string:1;
509 	uint16_t	rep:1;
510 	uint16_t	port;
511 	uint32_t	eax;		/* valid for out */
512 };
513 
514 struct vm_inout_str {
515 	struct vm_inout	inout;		/* must be the first element */
516 	struct vm_guest_paging paging;
517 	uint64_t	rflags;
518 	uint64_t	cr0;
519 	uint64_t	index;
520 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
521 	int		addrsize;
522 	enum vm_reg_name seg_name;
523 	struct seg_desc seg_desc;
524 };
525 
526 enum task_switch_reason {
527 	TSR_CALL,
528 	TSR_IRET,
529 	TSR_JMP,
530 	TSR_IDT_GATE,	/* task gate in IDT */
531 };
532 
533 struct vm_task_switch {
534 	uint16_t	tsssel;		/* new TSS selector */
535 	int		ext;		/* task switch due to external event */
536 	uint32_t	errcode;
537 	int		errcode_valid;	/* push 'errcode' on the new stack */
538 	enum task_switch_reason reason;
539 	struct vm_guest_paging paging;
540 };
541 
542 struct vm_exit {
543 	enum vm_exitcode	exitcode;
544 	int			inst_length;	/* 0 means unknown */
545 	uint64_t		rip;
546 	union {
547 		struct vm_inout	inout;
548 		struct vm_inout_str inout_str;
549 		struct {
550 			uint64_t	gpa;
551 			int		fault_type;
552 		} paging;
553 		struct {
554 			uint64_t	gpa;
555 			uint64_t	gla;
556 			uint64_t	cs_base;
557 			int		cs_d;		/* CS.D */
558 			struct vm_guest_paging paging;
559 			struct vie	vie;
560 		} inst_emul;
561 		/*
562 		 * VMX specific payload. Used when there is no "better"
563 		 * exitcode to represent the VM-exit.
564 		 */
565 		struct {
566 			int		status;		/* vmx inst status */
567 			/*
568 			 * 'exit_reason' and 'exit_qualification' are valid
569 			 * only if 'status' is zero.
570 			 */
571 			uint32_t	exit_reason;
572 			uint64_t	exit_qualification;
573 			/*
574 			 * 'inst_error' and 'inst_type' are valid
575 			 * only if 'status' is non-zero.
576 			 */
577 			int		inst_type;
578 			int		inst_error;
579 		} vmx;
580 		/*
581 		 * SVM specific payload.
582 		 */
583 		struct {
584 			uint64_t	exitcode;
585 			uint64_t	exitinfo1;
586 			uint64_t	exitinfo2;
587 		} svm;
588 		struct {
589 			uint32_t	code;		/* ecx value */
590 			uint64_t	wval;
591 		} msr;
592 		struct {
593 			int		vcpu;
594 			uint64_t	rip;
595 		} spinup_ap;
596 		struct {
597 			uint64_t	rflags;
598 		} hlt;
599 		struct {
600 			int		vector;
601 		} ioapic_eoi;
602 		struct {
603 			enum vm_suspend_how how;
604 		} suspended;
605 		struct vm_task_switch task_switch;
606 	} u;
607 };
608 
609 /* APIs to inject faults into the guest */
610 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
611     int errcode);
612 
613 static __inline void
614 vm_inject_ud(void *vm, int vcpuid)
615 {
616 	vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
617 }
618 
619 static __inline void
620 vm_inject_gp(void *vm, int vcpuid)
621 {
622 	vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
623 }
624 
625 static __inline void
626 vm_inject_ac(void *vm, int vcpuid, int errcode)
627 {
628 	vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
629 }
630 
631 static __inline void
632 vm_inject_ss(void *vm, int vcpuid, int errcode)
633 {
634 	vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
635 }
636 
637 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
638 
639 int vm_restart_instruction(void *vm, int vcpuid);
640 
641 #endif	/* _VMM_H_ */
642