xref: /freebsd/sys/amd64/include/vmm.h (revision 3e5645b78f476816ca3b5acc28b29bbafbb9c444)
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 
280 	if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
281 		return (1);
282 	else if (curthread->td_owepreempt)
283 		return (1);
284 	else
285 		return (0);
286 }
287 #endif
288 
289 void *vcpu_stats(struct vm *vm, int vcpu);
290 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
291 struct vmspace *vm_get_vmspace(struct vm *vm);
292 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
293 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
294 struct vatpic *vm_atpic(struct vm *vm);
295 struct vatpit *vm_atpit(struct vm *vm);
296 struct vpmtmr *vm_pmtmr(struct vm *vm);
297 struct vrtc *vm_rtc(struct vm *vm);
298 
299 /*
300  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
301  * success and non-zero on failure.
302  *
303  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
304  * this function directly because they enforce the trap-like or fault-like
305  * behavior of an exception.
306  *
307  * This function should only be called in the context of the thread that is
308  * executing this vcpu.
309  */
310 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
311     uint32_t errcode, int restart_instruction);
312 
313 /*
314  * This function is called after a VM-exit that occurred during exception or
315  * interrupt delivery through the IDT. The format of 'intinfo' is described
316  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
317  *
318  * If a VM-exit handler completes the event delivery successfully then it
319  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
320  * if the task switch emulation is triggered via a task gate then it should
321  * call this function with 'intinfo=0' to indicate that the external event
322  * is not pending anymore.
323  *
324  * Return value is 0 on success and non-zero on failure.
325  */
326 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
327 
328 /*
329  * This function is called before every VM-entry to retrieve a pending
330  * event that should be injected into the guest. This function combines
331  * nested events into a double or triple fault.
332  *
333  * Returns 0 if there are no events that need to be injected into the guest
334  * and non-zero otherwise.
335  */
336 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
337 
338 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
339 
340 enum vm_reg_name vm_segment_name(int seg_encoding);
341 
342 struct vm_copyinfo {
343 	uint64_t	gpa;
344 	size_t		len;
345 	void		*hva;
346 	void		*cookie;
347 };
348 
349 /*
350  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
351  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
352  * a copyin or PROT_WRITE for a copyout.
353  *
354  * retval	is_fault	Intepretation
355  *   0		   0		Success
356  *   0		   1		An exception was injected into the guest
357  * EFAULT	  N/A		Unrecoverable error
358  *
359  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
360  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
361  * 'vm_copy_teardown()' after the copy is done.
362  */
363 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
364     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
365     int num_copyinfo, int *is_fault);
366 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
367     int num_copyinfo);
368 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
369     void *kaddr, size_t len);
370 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
371     struct vm_copyinfo *copyinfo, size_t len);
372 
373 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
374 #endif	/* KERNEL */
375 
376 #define	VM_MAXCPU	16			/* maximum virtual cpus */
377 
378 /*
379  * Identifiers for optional vmm capabilities
380  */
381 enum vm_cap_type {
382 	VM_CAP_HALT_EXIT,
383 	VM_CAP_MTRAP_EXIT,
384 	VM_CAP_PAUSE_EXIT,
385 	VM_CAP_UNRESTRICTED_GUEST,
386 	VM_CAP_ENABLE_INVPCID,
387 	VM_CAP_MAX
388 };
389 
390 enum vm_intr_trigger {
391 	EDGE_TRIGGER,
392 	LEVEL_TRIGGER
393 };
394 
395 /*
396  * The 'access' field has the format specified in Table 21-2 of the Intel
397  * Architecture Manual vol 3b.
398  *
399  * XXX The contents of the 'access' field are architecturally defined except
400  * bit 16 - Segment Unusable.
401  */
402 struct seg_desc {
403 	uint64_t	base;
404 	uint32_t	limit;
405 	uint32_t	access;
406 };
407 #define	SEG_DESC_TYPE(access)		((access) & 0x001f)
408 #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3)
409 #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0)
410 #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0)
411 #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0)
412 #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0)
413 
414 enum vm_cpu_mode {
415 	CPU_MODE_REAL,
416 	CPU_MODE_PROTECTED,
417 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
418 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
419 };
420 
421 enum vm_paging_mode {
422 	PAGING_MODE_FLAT,
423 	PAGING_MODE_32,
424 	PAGING_MODE_PAE,
425 	PAGING_MODE_64,
426 };
427 
428 struct vm_guest_paging {
429 	uint64_t	cr3;
430 	int		cpl;
431 	enum vm_cpu_mode cpu_mode;
432 	enum vm_paging_mode paging_mode;
433 };
434 
435 /*
436  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
437  * consumers of instruction decoding. The only reason why their contents
438  * need to be exposed is because they are part of the 'vm_exit' structure.
439  */
440 struct vie_op {
441 	uint8_t		op_byte;	/* actual opcode byte */
442 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
443 	uint16_t	op_flags;
444 };
445 
446 #define	VIE_INST_SIZE	15
447 struct vie {
448 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
449 	uint8_t		num_valid;		/* size of the instruction */
450 	uint8_t		num_processed;
451 
452 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */
453 	uint8_t		rex_w:1,		/* REX prefix */
454 			rex_r:1,
455 			rex_x:1,
456 			rex_b:1,
457 			rex_present:1,
458 			repz_present:1,		/* REP/REPE/REPZ prefix */
459 			repnz_present:1,	/* REPNE/REPNZ prefix */
460 			opsize_override:1,	/* Operand size override */
461 			addrsize_override:1,	/* Address size override */
462 			segment_override:1;	/* Segment override */
463 
464 	uint8_t		mod:2,			/* ModRM byte */
465 			reg:4,
466 			rm:4;
467 
468 	uint8_t		ss:2,			/* SIB byte */
469 			index:4,
470 			base:4;
471 
472 	uint8_t		disp_bytes;
473 	uint8_t		imm_bytes;
474 
475 	uint8_t		scale;
476 	int		base_register;		/* VM_REG_GUEST_xyz */
477 	int		index_register;		/* VM_REG_GUEST_xyz */
478 	int		segment_register;	/* VM_REG_GUEST_xyz */
479 
480 	int64_t		displacement;		/* optional addr displacement */
481 	int64_t		immediate;		/* optional immediate operand */
482 
483 	uint8_t		decoded;	/* set to 1 if successfully decoded */
484 
485 	struct vie_op	op;			/* opcode description */
486 };
487 
488 enum vm_exitcode {
489 	VM_EXITCODE_INOUT,
490 	VM_EXITCODE_VMX,
491 	VM_EXITCODE_BOGUS,
492 	VM_EXITCODE_RDMSR,
493 	VM_EXITCODE_WRMSR,
494 	VM_EXITCODE_HLT,
495 	VM_EXITCODE_MTRAP,
496 	VM_EXITCODE_PAUSE,
497 	VM_EXITCODE_PAGING,
498 	VM_EXITCODE_INST_EMUL,
499 	VM_EXITCODE_SPINUP_AP,
500 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
501 	VM_EXITCODE_RENDEZVOUS,
502 	VM_EXITCODE_IOAPIC_EOI,
503 	VM_EXITCODE_SUSPENDED,
504 	VM_EXITCODE_INOUT_STR,
505 	VM_EXITCODE_TASK_SWITCH,
506 	VM_EXITCODE_MONITOR,
507 	VM_EXITCODE_MWAIT,
508 	VM_EXITCODE_SVM,
509 	VM_EXITCODE_MAX
510 };
511 
512 struct vm_inout {
513 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
514 	uint16_t	in:1;
515 	uint16_t	string:1;
516 	uint16_t	rep:1;
517 	uint16_t	port;
518 	uint32_t	eax;		/* valid for out */
519 };
520 
521 struct vm_inout_str {
522 	struct vm_inout	inout;		/* must be the first element */
523 	struct vm_guest_paging paging;
524 	uint64_t	rflags;
525 	uint64_t	cr0;
526 	uint64_t	index;
527 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
528 	int		addrsize;
529 	enum vm_reg_name seg_name;
530 	struct seg_desc seg_desc;
531 };
532 
533 enum task_switch_reason {
534 	TSR_CALL,
535 	TSR_IRET,
536 	TSR_JMP,
537 	TSR_IDT_GATE,	/* task gate in IDT */
538 };
539 
540 struct vm_task_switch {
541 	uint16_t	tsssel;		/* new TSS selector */
542 	int		ext;		/* task switch due to external event */
543 	uint32_t	errcode;
544 	int		errcode_valid;	/* push 'errcode' on the new stack */
545 	enum task_switch_reason reason;
546 	struct vm_guest_paging paging;
547 };
548 
549 struct vm_exit {
550 	enum vm_exitcode	exitcode;
551 	int			inst_length;	/* 0 means unknown */
552 	uint64_t		rip;
553 	union {
554 		struct vm_inout	inout;
555 		struct vm_inout_str inout_str;
556 		struct {
557 			uint64_t	gpa;
558 			int		fault_type;
559 		} paging;
560 		struct {
561 			uint64_t	gpa;
562 			uint64_t	gla;
563 			uint64_t	cs_base;
564 			int		cs_d;		/* CS.D */
565 			struct vm_guest_paging paging;
566 			struct vie	vie;
567 		} inst_emul;
568 		/*
569 		 * VMX specific payload. Used when there is no "better"
570 		 * exitcode to represent the VM-exit.
571 		 */
572 		struct {
573 			int		status;		/* vmx inst status */
574 			/*
575 			 * 'exit_reason' and 'exit_qualification' are valid
576 			 * only if 'status' is zero.
577 			 */
578 			uint32_t	exit_reason;
579 			uint64_t	exit_qualification;
580 			/*
581 			 * 'inst_error' and 'inst_type' are valid
582 			 * only if 'status' is non-zero.
583 			 */
584 			int		inst_type;
585 			int		inst_error;
586 		} vmx;
587 		/*
588 		 * SVM specific payload.
589 		 */
590 		struct {
591 			uint64_t	exitcode;
592 			uint64_t	exitinfo1;
593 			uint64_t	exitinfo2;
594 		} svm;
595 		struct {
596 			uint32_t	code;		/* ecx value */
597 			uint64_t	wval;
598 		} msr;
599 		struct {
600 			int		vcpu;
601 			uint64_t	rip;
602 		} spinup_ap;
603 		struct {
604 			uint64_t	rflags;
605 		} hlt;
606 		struct {
607 			int		vector;
608 		} ioapic_eoi;
609 		struct {
610 			enum vm_suspend_how how;
611 		} suspended;
612 		struct vm_task_switch task_switch;
613 	} u;
614 };
615 
616 /* APIs to inject faults into the guest */
617 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
618     int errcode);
619 
620 static __inline void
621 vm_inject_ud(void *vm, int vcpuid)
622 {
623 	vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
624 }
625 
626 static __inline void
627 vm_inject_gp(void *vm, int vcpuid)
628 {
629 	vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
630 }
631 
632 static __inline void
633 vm_inject_ac(void *vm, int vcpuid, int errcode)
634 {
635 	vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
636 }
637 
638 static __inline void
639 vm_inject_ss(void *vm, int vcpuid, int errcode)
640 {
641 	vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
642 }
643 
644 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
645 
646 int vm_restart_instruction(void *vm, int vcpuid);
647 
648 #endif	/* _VMM_H_ */
649