xref: /freebsd/sys/amd64/include/vmm.h (revision 0629b152762b06325dd75a41bcb0a2789514141b)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2011 NetApp, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30 
31 #ifndef _VMM_H_
32 #define	_VMM_H_
33 
34 #include <x86/segments.h>
35 
36 enum vm_suspend_how {
37 	VM_SUSPEND_NONE,
38 	VM_SUSPEND_RESET,
39 	VM_SUSPEND_POWEROFF,
40 	VM_SUSPEND_HALT,
41 	VM_SUSPEND_TRIPLEFAULT,
42 	VM_SUSPEND_LAST
43 };
44 
45 /*
46  * Identifiers for architecturally defined registers.
47  */
48 enum vm_reg_name {
49 	VM_REG_GUEST_RAX,
50 	VM_REG_GUEST_RBX,
51 	VM_REG_GUEST_RCX,
52 	VM_REG_GUEST_RDX,
53 	VM_REG_GUEST_RSI,
54 	VM_REG_GUEST_RDI,
55 	VM_REG_GUEST_RBP,
56 	VM_REG_GUEST_R8,
57 	VM_REG_GUEST_R9,
58 	VM_REG_GUEST_R10,
59 	VM_REG_GUEST_R11,
60 	VM_REG_GUEST_R12,
61 	VM_REG_GUEST_R13,
62 	VM_REG_GUEST_R14,
63 	VM_REG_GUEST_R15,
64 	VM_REG_GUEST_CR0,
65 	VM_REG_GUEST_CR3,
66 	VM_REG_GUEST_CR4,
67 	VM_REG_GUEST_DR7,
68 	VM_REG_GUEST_RSP,
69 	VM_REG_GUEST_RIP,
70 	VM_REG_GUEST_RFLAGS,
71 	VM_REG_GUEST_ES,
72 	VM_REG_GUEST_CS,
73 	VM_REG_GUEST_SS,
74 	VM_REG_GUEST_DS,
75 	VM_REG_GUEST_FS,
76 	VM_REG_GUEST_GS,
77 	VM_REG_GUEST_LDTR,
78 	VM_REG_GUEST_TR,
79 	VM_REG_GUEST_IDTR,
80 	VM_REG_GUEST_GDTR,
81 	VM_REG_GUEST_EFER,
82 	VM_REG_GUEST_CR2,
83 	VM_REG_GUEST_PDPTE0,
84 	VM_REG_GUEST_PDPTE1,
85 	VM_REG_GUEST_PDPTE2,
86 	VM_REG_GUEST_PDPTE3,
87 	VM_REG_GUEST_INTR_SHADOW,
88 	VM_REG_GUEST_DR0,
89 	VM_REG_GUEST_DR1,
90 	VM_REG_GUEST_DR2,
91 	VM_REG_GUEST_DR3,
92 	VM_REG_GUEST_DR6,
93 	VM_REG_LAST
94 };
95 
96 enum x2apic_state {
97 	X2APIC_DISABLED,
98 	X2APIC_ENABLED,
99 	X2APIC_STATE_LAST
100 };
101 
102 #define	VM_INTINFO_VECTOR(info)	((info) & 0xff)
103 #define	VM_INTINFO_DEL_ERRCODE	0x800
104 #define	VM_INTINFO_RSVD		0x7ffff000
105 #define	VM_INTINFO_VALID	0x80000000
106 #define	VM_INTINFO_TYPE		0x700
107 #define	VM_INTINFO_HWINTR	(0 << 8)
108 #define	VM_INTINFO_NMI		(2 << 8)
109 #define	VM_INTINFO_HWEXCEPTION	(3 << 8)
110 #define	VM_INTINFO_SWINTR	(4 << 8)
111 
112 #ifdef _KERNEL
113 
114 #define	VM_MAX_NAMELEN	32
115 
116 struct vm;
117 struct vm_exception;
118 struct seg_desc;
119 struct vm_exit;
120 struct vm_run;
121 struct vhpet;
122 struct vioapic;
123 struct vlapic;
124 struct vmspace;
125 struct vm_object;
126 struct vm_guest_paging;
127 struct pmap;
128 
129 struct vm_eventinfo {
130 	void	*rptr;		/* rendezvous cookie */
131 	int	*sptr;		/* suspend cookie */
132 	int	*iptr;		/* reqidle cookie */
133 };
134 
135 typedef int	(*vmm_init_func_t)(int ipinum);
136 typedef int	(*vmm_cleanup_func_t)(void);
137 typedef void	(*vmm_resume_func_t)(void);
138 typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
139 typedef int	(*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
140 		    struct pmap *pmap, struct vm_eventinfo *info);
141 typedef void	(*vmi_cleanup_func_t)(void *vmi);
142 typedef int	(*vmi_get_register_t)(void *vmi, int vcpu, int num,
143 				      uint64_t *retval);
144 typedef int	(*vmi_set_register_t)(void *vmi, int vcpu, int num,
145 				      uint64_t val);
146 typedef int	(*vmi_get_desc_t)(void *vmi, int vcpu, int num,
147 				  struct seg_desc *desc);
148 typedef int	(*vmi_set_desc_t)(void *vmi, int vcpu, int num,
149 				  struct seg_desc *desc);
150 typedef int	(*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
151 typedef int	(*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
152 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
153 typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace);
154 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
155 typedef void	(*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
156 
157 struct vmm_ops {
158 	vmm_init_func_t		init;		/* module wide initialization */
159 	vmm_cleanup_func_t	cleanup;
160 	vmm_resume_func_t	resume;
161 
162 	vmi_init_func_t		vminit;		/* vm-specific initialization */
163 	vmi_run_func_t		vmrun;
164 	vmi_cleanup_func_t	vmcleanup;
165 	vmi_get_register_t	vmgetreg;
166 	vmi_set_register_t	vmsetreg;
167 	vmi_get_desc_t		vmgetdesc;
168 	vmi_set_desc_t		vmsetdesc;
169 	vmi_get_cap_t		vmgetcap;
170 	vmi_set_cap_t		vmsetcap;
171 	vmi_vmspace_alloc	vmspace_alloc;
172 	vmi_vmspace_free	vmspace_free;
173 	vmi_vlapic_init		vlapic_init;
174 	vmi_vlapic_cleanup	vlapic_cleanup;
175 };
176 
177 extern struct vmm_ops vmm_ops_intel;
178 extern struct vmm_ops vmm_ops_amd;
179 
180 int vm_create(const char *name, struct vm **retvm);
181 void vm_destroy(struct vm *vm);
182 int vm_reinit(struct vm *vm);
183 const char *vm_name(struct vm *vm);
184 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
185     uint16_t *threads, uint16_t *maxcpus);
186 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
187     uint16_t threads, uint16_t maxcpus);
188 
189 /*
190  * APIs that modify the guest memory map require all vcpus to be frozen.
191  */
192 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
193     size_t len, int prot, int flags);
194 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
195 void vm_free_memseg(struct vm *vm, int ident);
196 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
197 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
198 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
199 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
200 
201 /*
202  * APIs that inspect the guest memory map require only a *single* vcpu to
203  * be frozen. This acts like a read lock on the guest memory map since any
204  * modification requires *all* vcpus to be frozen.
205  */
206 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
207     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
208 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
209     struct vm_object **objptr);
210 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
211     int prot, void **cookie);
212 void vm_gpa_release(void *cookie);
213 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
214 
215 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
216 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
217 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
218 		    struct seg_desc *ret_desc);
219 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
220 		    struct seg_desc *desc);
221 int vm_run(struct vm *vm, struct vm_run *vmrun);
222 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
223 int vm_inject_nmi(struct vm *vm, int vcpu);
224 int vm_nmi_pending(struct vm *vm, int vcpuid);
225 void vm_nmi_clear(struct vm *vm, int vcpuid);
226 int vm_inject_extint(struct vm *vm, int vcpu);
227 int vm_extint_pending(struct vm *vm, int vcpuid);
228 void vm_extint_clear(struct vm *vm, int vcpuid);
229 struct vlapic *vm_lapic(struct vm *vm, int cpu);
230 struct vioapic *vm_ioapic(struct vm *vm);
231 struct vhpet *vm_hpet(struct vm *vm);
232 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
233 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
234 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
235 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
236 int vm_apicid2vcpuid(struct vm *vm, int apicid);
237 int vm_activate_cpu(struct vm *vm, int vcpu);
238 int vm_suspend_cpu(struct vm *vm, int vcpu);
239 int vm_resume_cpu(struct vm *vm, int vcpu);
240 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
241 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
242 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip);
243 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
244 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
245 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
246 
247 #ifdef _SYS__CPUSET_H_
248 /*
249  * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
250  * The rendezvous 'func(arg)' is not allowed to do anything that will
251  * cause the thread to be put to sleep.
252  *
253  * If the rendezvous is being initiated from a vcpu context then the
254  * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
255  *
256  * The caller cannot hold any locks when initiating the rendezvous.
257  *
258  * The implementation of this API may cause vcpus other than those specified
259  * by 'dest' to be stalled. The caller should not rely on any vcpus making
260  * forward progress when the rendezvous is in progress.
261  */
262 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
263 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
264     vm_rendezvous_func_t func, void *arg);
265 cpuset_t vm_active_cpus(struct vm *vm);
266 cpuset_t vm_debug_cpus(struct vm *vm);
267 cpuset_t vm_suspended_cpus(struct vm *vm);
268 #endif	/* _SYS__CPUSET_H_ */
269 
270 static __inline int
271 vcpu_rendezvous_pending(struct vm_eventinfo *info)
272 {
273 
274 	return (*((uintptr_t *)(info->rptr)) != 0);
275 }
276 
277 static __inline int
278 vcpu_suspended(struct vm_eventinfo *info)
279 {
280 
281 	return (*info->sptr);
282 }
283 
284 static __inline int
285 vcpu_reqidle(struct vm_eventinfo *info)
286 {
287 
288 	return (*info->iptr);
289 }
290 
291 int vcpu_debugged(struct vm *vm, int vcpuid);
292 
293 /*
294  * Return 1 if device indicated by bus/slot/func is supposed to be a
295  * pci passthrough device.
296  *
297  * Return 0 otherwise.
298  */
299 int vmm_is_pptdev(int bus, int slot, int func);
300 
301 void *vm_iommu_domain(struct vm *vm);
302 
303 enum vcpu_state {
304 	VCPU_IDLE,
305 	VCPU_FROZEN,
306 	VCPU_RUNNING,
307 	VCPU_SLEEPING,
308 };
309 
310 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
311     bool from_idle);
312 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
313 
314 static int __inline
315 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
316 {
317 	return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
318 }
319 
320 #ifdef _SYS_PROC_H_
321 static int __inline
322 vcpu_should_yield(struct vm *vm, int vcpu)
323 {
324 
325 	if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
326 		return (1);
327 	else if (curthread->td_owepreempt)
328 		return (1);
329 	else
330 		return (0);
331 }
332 #endif
333 
334 void *vcpu_stats(struct vm *vm, int vcpu);
335 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
336 struct vmspace *vm_get_vmspace(struct vm *vm);
337 struct vatpic *vm_atpic(struct vm *vm);
338 struct vatpit *vm_atpit(struct vm *vm);
339 struct vpmtmr *vm_pmtmr(struct vm *vm);
340 struct vrtc *vm_rtc(struct vm *vm);
341 
342 /*
343  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
344  * success and non-zero on failure.
345  *
346  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
347  * this function directly because they enforce the trap-like or fault-like
348  * behavior of an exception.
349  *
350  * This function should only be called in the context of the thread that is
351  * executing this vcpu.
352  */
353 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
354     uint32_t errcode, int restart_instruction);
355 
356 /*
357  * This function is called after a VM-exit that occurred during exception or
358  * interrupt delivery through the IDT. The format of 'intinfo' is described
359  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
360  *
361  * If a VM-exit handler completes the event delivery successfully then it
362  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
363  * if the task switch emulation is triggered via a task gate then it should
364  * call this function with 'intinfo=0' to indicate that the external event
365  * is not pending anymore.
366  *
367  * Return value is 0 on success and non-zero on failure.
368  */
369 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
370 
371 /*
372  * This function is called before every VM-entry to retrieve a pending
373  * event that should be injected into the guest. This function combines
374  * nested events into a double or triple fault.
375  *
376  * Returns 0 if there are no events that need to be injected into the guest
377  * and non-zero otherwise.
378  */
379 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
380 
381 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
382 
383 enum vm_reg_name vm_segment_name(int seg_encoding);
384 
385 struct vm_copyinfo {
386 	uint64_t	gpa;
387 	size_t		len;
388 	void		*hva;
389 	void		*cookie;
390 };
391 
392 /*
393  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
394  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
395  * a copyin or PROT_WRITE for a copyout.
396  *
397  * retval	is_fault	Interpretation
398  *   0		   0		Success
399  *   0		   1		An exception was injected into the guest
400  * EFAULT	  N/A		Unrecoverable error
401  *
402  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
403  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
404  * 'vm_copy_teardown()' after the copy is done.
405  */
406 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
407     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
408     int num_copyinfo, int *is_fault);
409 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
410     int num_copyinfo);
411 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
412     void *kaddr, size_t len);
413 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
414     struct vm_copyinfo *copyinfo, size_t len);
415 
416 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
417 #endif	/* KERNEL */
418 
419 #define	VM_MAXCPU	16			/* maximum virtual cpus */
420 
421 /*
422  * Identifiers for optional vmm capabilities
423  */
424 enum vm_cap_type {
425 	VM_CAP_HALT_EXIT,
426 	VM_CAP_MTRAP_EXIT,
427 	VM_CAP_PAUSE_EXIT,
428 	VM_CAP_UNRESTRICTED_GUEST,
429 	VM_CAP_ENABLE_INVPCID,
430 	VM_CAP_MAX
431 };
432 
433 enum vm_intr_trigger {
434 	EDGE_TRIGGER,
435 	LEVEL_TRIGGER
436 };
437 
438 /*
439  * The 'access' field has the format specified in Table 21-2 of the Intel
440  * Architecture Manual vol 3b.
441  *
442  * XXX The contents of the 'access' field are architecturally defined except
443  * bit 16 - Segment Unusable.
444  */
445 struct seg_desc {
446 	uint64_t	base;
447 	uint32_t	limit;
448 	uint32_t	access;
449 };
450 #define	SEG_DESC_TYPE(access)		((access) & 0x001f)
451 #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3)
452 #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0)
453 #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0)
454 #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0)
455 #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0)
456 
457 enum vm_cpu_mode {
458 	CPU_MODE_REAL,
459 	CPU_MODE_PROTECTED,
460 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
461 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
462 };
463 
464 enum vm_paging_mode {
465 	PAGING_MODE_FLAT,
466 	PAGING_MODE_32,
467 	PAGING_MODE_PAE,
468 	PAGING_MODE_64,
469 };
470 
471 struct vm_guest_paging {
472 	uint64_t	cr3;
473 	int		cpl;
474 	enum vm_cpu_mode cpu_mode;
475 	enum vm_paging_mode paging_mode;
476 };
477 
478 /*
479  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
480  * consumers of instruction decoding. The only reason why their contents
481  * need to be exposed is because they are part of the 'vm_exit' structure.
482  */
483 struct vie_op {
484 	uint8_t		op_byte;	/* actual opcode byte */
485 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
486 	uint16_t	op_flags;
487 };
488 
489 #define	VIE_INST_SIZE	15
490 struct vie {
491 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
492 	uint8_t		num_valid;		/* size of the instruction */
493 	uint8_t		num_processed;
494 
495 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */
496 	uint8_t		rex_w:1,		/* REX prefix */
497 			rex_r:1,
498 			rex_x:1,
499 			rex_b:1,
500 			rex_present:1,
501 			repz_present:1,		/* REP/REPE/REPZ prefix */
502 			repnz_present:1,	/* REPNE/REPNZ prefix */
503 			opsize_override:1,	/* Operand size override */
504 			addrsize_override:1,	/* Address size override */
505 			segment_override:1;	/* Segment override */
506 
507 	uint8_t		mod:2,			/* ModRM byte */
508 			reg:4,
509 			rm:4;
510 
511 	uint8_t		ss:2,			/* SIB byte */
512 			index:4,
513 			base:4;
514 
515 	uint8_t		disp_bytes;
516 	uint8_t		imm_bytes;
517 
518 	uint8_t		scale;
519 	int		base_register;		/* VM_REG_GUEST_xyz */
520 	int		index_register;		/* VM_REG_GUEST_xyz */
521 	int		segment_register;	/* VM_REG_GUEST_xyz */
522 
523 	int64_t		displacement;		/* optional addr displacement */
524 	int64_t		immediate;		/* optional immediate operand */
525 
526 	uint8_t		decoded;	/* set to 1 if successfully decoded */
527 
528 	struct vie_op	op;			/* opcode description */
529 };
530 
531 enum vm_exitcode {
532 	VM_EXITCODE_INOUT,
533 	VM_EXITCODE_VMX,
534 	VM_EXITCODE_BOGUS,
535 	VM_EXITCODE_RDMSR,
536 	VM_EXITCODE_WRMSR,
537 	VM_EXITCODE_HLT,
538 	VM_EXITCODE_MTRAP,
539 	VM_EXITCODE_PAUSE,
540 	VM_EXITCODE_PAGING,
541 	VM_EXITCODE_INST_EMUL,
542 	VM_EXITCODE_SPINUP_AP,
543 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
544 	VM_EXITCODE_RENDEZVOUS,
545 	VM_EXITCODE_IOAPIC_EOI,
546 	VM_EXITCODE_SUSPENDED,
547 	VM_EXITCODE_INOUT_STR,
548 	VM_EXITCODE_TASK_SWITCH,
549 	VM_EXITCODE_MONITOR,
550 	VM_EXITCODE_MWAIT,
551 	VM_EXITCODE_SVM,
552 	VM_EXITCODE_REQIDLE,
553 	VM_EXITCODE_DEBUG,
554 	VM_EXITCODE_MAX
555 };
556 
557 struct vm_inout {
558 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
559 	uint16_t	in:1;
560 	uint16_t	string:1;
561 	uint16_t	rep:1;
562 	uint16_t	port;
563 	uint32_t	eax;		/* valid for out */
564 };
565 
566 struct vm_inout_str {
567 	struct vm_inout	inout;		/* must be the first element */
568 	struct vm_guest_paging paging;
569 	uint64_t	rflags;
570 	uint64_t	cr0;
571 	uint64_t	index;
572 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
573 	int		addrsize;
574 	enum vm_reg_name seg_name;
575 	struct seg_desc seg_desc;
576 };
577 
578 enum task_switch_reason {
579 	TSR_CALL,
580 	TSR_IRET,
581 	TSR_JMP,
582 	TSR_IDT_GATE,	/* task gate in IDT */
583 };
584 
585 struct vm_task_switch {
586 	uint16_t	tsssel;		/* new TSS selector */
587 	int		ext;		/* task switch due to external event */
588 	uint32_t	errcode;
589 	int		errcode_valid;	/* push 'errcode' on the new stack */
590 	enum task_switch_reason reason;
591 	struct vm_guest_paging paging;
592 };
593 
594 struct vm_exit {
595 	enum vm_exitcode	exitcode;
596 	int			inst_length;	/* 0 means unknown */
597 	uint64_t		rip;
598 	union {
599 		struct vm_inout	inout;
600 		struct vm_inout_str inout_str;
601 		struct {
602 			uint64_t	gpa;
603 			int		fault_type;
604 		} paging;
605 		struct {
606 			uint64_t	gpa;
607 			uint64_t	gla;
608 			uint64_t	cs_base;
609 			int		cs_d;		/* CS.D */
610 			struct vm_guest_paging paging;
611 			struct vie	vie;
612 		} inst_emul;
613 		/*
614 		 * VMX specific payload. Used when there is no "better"
615 		 * exitcode to represent the VM-exit.
616 		 */
617 		struct {
618 			int		status;		/* vmx inst status */
619 			/*
620 			 * 'exit_reason' and 'exit_qualification' are valid
621 			 * only if 'status' is zero.
622 			 */
623 			uint32_t	exit_reason;
624 			uint64_t	exit_qualification;
625 			/*
626 			 * 'inst_error' and 'inst_type' are valid
627 			 * only if 'status' is non-zero.
628 			 */
629 			int		inst_type;
630 			int		inst_error;
631 		} vmx;
632 		/*
633 		 * SVM specific payload.
634 		 */
635 		struct {
636 			uint64_t	exitcode;
637 			uint64_t	exitinfo1;
638 			uint64_t	exitinfo2;
639 		} svm;
640 		struct {
641 			uint32_t	code;		/* ecx value */
642 			uint64_t	wval;
643 		} msr;
644 		struct {
645 			int		vcpu;
646 			uint64_t	rip;
647 		} spinup_ap;
648 		struct {
649 			uint64_t	rflags;
650 			uint64_t	intr_status;
651 		} hlt;
652 		struct {
653 			int		vector;
654 		} ioapic_eoi;
655 		struct {
656 			enum vm_suspend_how how;
657 		} suspended;
658 		struct vm_task_switch task_switch;
659 	} u;
660 };
661 
662 /* APIs to inject faults into the guest */
663 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
664     int errcode);
665 
666 static __inline void
667 vm_inject_ud(void *vm, int vcpuid)
668 {
669 	vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
670 }
671 
672 static __inline void
673 vm_inject_gp(void *vm, int vcpuid)
674 {
675 	vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
676 }
677 
678 static __inline void
679 vm_inject_ac(void *vm, int vcpuid, int errcode)
680 {
681 	vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
682 }
683 
684 static __inline void
685 vm_inject_ss(void *vm, int vcpuid, int errcode)
686 {
687 	vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
688 }
689 
690 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
691 
692 int vm_restart_instruction(void *vm, int vcpuid);
693 
694 #endif	/* _VMM_H_ */
695