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: stable/12/sys/amd64/include/vmm.h 365560 2020-09-10 10:49:59Z grehan $
29  */
30 
31 #ifndef _VMM_H_
32 #define	_VMM_H_
33 
34 #include <sys/sdt.h>
35 #include <x86/segments.h>
36 
37 #ifdef _KERNEL
38 SDT_PROVIDER_DECLARE(vmm);
39 #endif
40 
41 enum vm_suspend_how {
42 	VM_SUSPEND_NONE,
43 	VM_SUSPEND_RESET,
44 	VM_SUSPEND_POWEROFF,
45 	VM_SUSPEND_HALT,
46 	VM_SUSPEND_TRIPLEFAULT,
47 	VM_SUSPEND_LAST
48 };
49 
50 /*
51  * Identifiers for architecturally defined registers.
52  */
53 enum vm_reg_name {
54 	VM_REG_GUEST_RAX,
55 	VM_REG_GUEST_RBX,
56 	VM_REG_GUEST_RCX,
57 	VM_REG_GUEST_RDX,
58 	VM_REG_GUEST_RSI,
59 	VM_REG_GUEST_RDI,
60 	VM_REG_GUEST_RBP,
61 	VM_REG_GUEST_R8,
62 	VM_REG_GUEST_R9,
63 	VM_REG_GUEST_R10,
64 	VM_REG_GUEST_R11,
65 	VM_REG_GUEST_R12,
66 	VM_REG_GUEST_R13,
67 	VM_REG_GUEST_R14,
68 	VM_REG_GUEST_R15,
69 	VM_REG_GUEST_CR0,
70 	VM_REG_GUEST_CR3,
71 	VM_REG_GUEST_CR4,
72 	VM_REG_GUEST_DR7,
73 	VM_REG_GUEST_RSP,
74 	VM_REG_GUEST_RIP,
75 	VM_REG_GUEST_RFLAGS,
76 	VM_REG_GUEST_ES,
77 	VM_REG_GUEST_CS,
78 	VM_REG_GUEST_SS,
79 	VM_REG_GUEST_DS,
80 	VM_REG_GUEST_FS,
81 	VM_REG_GUEST_GS,
82 	VM_REG_GUEST_LDTR,
83 	VM_REG_GUEST_TR,
84 	VM_REG_GUEST_IDTR,
85 	VM_REG_GUEST_GDTR,
86 	VM_REG_GUEST_EFER,
87 	VM_REG_GUEST_CR2,
88 	VM_REG_GUEST_PDPTE0,
89 	VM_REG_GUEST_PDPTE1,
90 	VM_REG_GUEST_PDPTE2,
91 	VM_REG_GUEST_PDPTE3,
92 	VM_REG_GUEST_INTR_SHADOW,
93 	VM_REG_GUEST_DR0,
94 	VM_REG_GUEST_DR1,
95 	VM_REG_GUEST_DR2,
96 	VM_REG_GUEST_DR3,
97 	VM_REG_GUEST_DR6,
98 	VM_REG_GUEST_ENTRY_INST_LENGTH,
99 	VM_REG_LAST
100 };
101 
102 enum x2apic_state {
103 	X2APIC_DISABLED,
104 	X2APIC_ENABLED,
105 	X2APIC_STATE_LAST
106 };
107 
108 #define	VM_INTINFO_VECTOR(info)	((info) & 0xff)
109 #define	VM_INTINFO_DEL_ERRCODE	0x800
110 #define	VM_INTINFO_RSVD		0x7ffff000
111 #define	VM_INTINFO_VALID	0x80000000
112 #define	VM_INTINFO_TYPE		0x700
113 #define	VM_INTINFO_HWINTR	(0 << 8)
114 #define	VM_INTINFO_NMI		(2 << 8)
115 #define	VM_INTINFO_HWEXCEPTION	(3 << 8)
116 #define	VM_INTINFO_SWINTR	(4 << 8)
117 
118 #ifdef _KERNEL
119 
120 #define	VM_MAX_NAMELEN	32
121 
122 struct vm;
123 struct vm_exception;
124 struct seg_desc;
125 struct vm_exit;
126 struct vm_run;
127 struct vhpet;
128 struct vioapic;
129 struct vlapic;
130 struct vmspace;
131 struct vm_object;
132 struct vm_guest_paging;
133 struct pmap;
134 
135 struct vm_eventinfo {
136 	void	*rptr;		/* rendezvous cookie */
137 	int	*sptr;		/* suspend cookie */
138 	int	*iptr;		/* reqidle cookie */
139 };
140 
141 typedef int	(*vmm_init_func_t)(int ipinum);
142 typedef int	(*vmm_cleanup_func_t)(void);
143 typedef void	(*vmm_resume_func_t)(void);
144 typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
145 typedef int	(*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
146 		    struct pmap *pmap, struct vm_eventinfo *info);
147 typedef void	(*vmi_cleanup_func_t)(void *vmi);
148 typedef int	(*vmi_get_register_t)(void *vmi, int vcpu, int num,
149 				      uint64_t *retval);
150 typedef int	(*vmi_set_register_t)(void *vmi, int vcpu, int num,
151 				      uint64_t val);
152 typedef int	(*vmi_get_desc_t)(void *vmi, int vcpu, int num,
153 				  struct seg_desc *desc);
154 typedef int	(*vmi_set_desc_t)(void *vmi, int vcpu, int num,
155 				  struct seg_desc *desc);
156 typedef int	(*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
157 typedef int	(*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
158 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
159 typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace);
160 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
161 typedef void	(*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
162 
163 struct vmm_ops {
164 	vmm_init_func_t		init;		/* module wide initialization */
165 	vmm_cleanup_func_t	cleanup;
166 	vmm_resume_func_t	resume;
167 
168 	vmi_init_func_t		vminit;		/* vm-specific initialization */
169 	vmi_run_func_t		vmrun;
170 	vmi_cleanup_func_t	vmcleanup;
171 	vmi_get_register_t	vmgetreg;
172 	vmi_set_register_t	vmsetreg;
173 	vmi_get_desc_t		vmgetdesc;
174 	vmi_set_desc_t		vmsetdesc;
175 	vmi_get_cap_t		vmgetcap;
176 	vmi_set_cap_t		vmsetcap;
177 	vmi_vmspace_alloc	vmspace_alloc;
178 	vmi_vmspace_free	vmspace_free;
179 	vmi_vlapic_init		vlapic_init;
180 	vmi_vlapic_cleanup	vlapic_cleanup;
181 };
182 
183 extern struct vmm_ops vmm_ops_intel;
184 extern struct vmm_ops vmm_ops_amd;
185 
186 int vm_create(const char *name, struct vm **retvm);
187 void vm_destroy(struct vm *vm);
188 int vm_reinit(struct vm *vm);
189 const char *vm_name(struct vm *vm);
190 uint16_t vm_get_maxcpus(struct vm *vm);
191 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
192     uint16_t *threads, uint16_t *maxcpus);
193 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
194     uint16_t threads, uint16_t maxcpus);
195 
196 /*
197  * APIs that modify the guest memory map require all vcpus to be frozen.
198  */
199 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
200     size_t len, int prot, int flags);
201 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
202 void vm_free_memseg(struct vm *vm, int ident);
203 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
204 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
205 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
206 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
207 
208 /*
209  * APIs that inspect the guest memory map require only a *single* vcpu to
210  * be frozen. This acts like a read lock on the guest memory map since any
211  * modification requires *all* vcpus to be frozen.
212  */
213 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
214     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
215 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
216     struct vm_object **objptr);
217 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
218 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
219     int prot, void **cookie);
220 void vm_gpa_release(void *cookie);
221 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
222 
223 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
224 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
225 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
226 		    struct seg_desc *ret_desc);
227 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
228 		    struct seg_desc *desc);
229 int vm_run(struct vm *vm, struct vm_run *vmrun);
230 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
231 int vm_inject_nmi(struct vm *vm, int vcpu);
232 int vm_nmi_pending(struct vm *vm, int vcpuid);
233 void vm_nmi_clear(struct vm *vm, int vcpuid);
234 int vm_inject_extint(struct vm *vm, int vcpu);
235 int vm_extint_pending(struct vm *vm, int vcpuid);
236 void vm_extint_clear(struct vm *vm, int vcpuid);
237 struct vlapic *vm_lapic(struct vm *vm, int cpu);
238 struct vioapic *vm_ioapic(struct vm *vm);
239 struct vhpet *vm_hpet(struct vm *vm);
240 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
241 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
242 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
243 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
244 int vm_apicid2vcpuid(struct vm *vm, int apicid);
245 int vm_activate_cpu(struct vm *vm, int vcpu);
246 int vm_suspend_cpu(struct vm *vm, int vcpu);
247 int vm_resume_cpu(struct vm *vm, int vcpu);
248 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
249 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
250 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip);
251 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
252 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
253 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
254 
255 #ifdef _SYS__CPUSET_H_
256 /*
257  * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
258  * The rendezvous 'func(arg)' is not allowed to do anything that will
259  * cause the thread to be put to sleep.
260  *
261  * If the rendezvous is being initiated from a vcpu context then the
262  * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
263  *
264  * The caller cannot hold any locks when initiating the rendezvous.
265  *
266  * The implementation of this API may cause vcpus other than those specified
267  * by 'dest' to be stalled. The caller should not rely on any vcpus making
268  * forward progress when the rendezvous is in progress.
269  */
270 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
271 int vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
272     vm_rendezvous_func_t func, void *arg);
273 cpuset_t vm_active_cpus(struct vm *vm);
274 cpuset_t vm_debug_cpus(struct vm *vm);
275 cpuset_t vm_suspended_cpus(struct vm *vm);
276 #endif	/* _SYS__CPUSET_H_ */
277 
278 static __inline int
vcpu_rendezvous_pending(struct vm_eventinfo * info)279 vcpu_rendezvous_pending(struct vm_eventinfo *info)
280 {
281 
282 	return (*((uintptr_t *)(info->rptr)) != 0);
283 }
284 
285 static __inline int
vcpu_suspended(struct vm_eventinfo * info)286 vcpu_suspended(struct vm_eventinfo *info)
287 {
288 
289 	return (*info->sptr);
290 }
291 
292 static __inline int
vcpu_reqidle(struct vm_eventinfo * info)293 vcpu_reqidle(struct vm_eventinfo *info)
294 {
295 
296 	return (*info->iptr);
297 }
298 
299 int vcpu_debugged(struct vm *vm, int vcpuid);
300 
301 /*
302  * Return true if device indicated by bus/slot/func is supposed to be a
303  * pci passthrough device.
304  *
305  * Return false otherwise.
306  */
307 bool vmm_is_pptdev(int bus, int slot, int func);
308 
309 void *vm_iommu_domain(struct vm *vm);
310 
311 enum vcpu_state {
312 	VCPU_IDLE,
313 	VCPU_FROZEN,
314 	VCPU_RUNNING,
315 	VCPU_SLEEPING,
316 };
317 
318 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
319     bool from_idle);
320 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
321 
322 static int __inline
vcpu_is_running(struct vm * vm,int vcpu,int * hostcpu)323 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
324 {
325 	return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
326 }
327 
328 #ifdef _SYS_PROC_H_
329 static int __inline
vcpu_should_yield(struct vm * vm,int vcpu)330 vcpu_should_yield(struct vm *vm, int vcpu)
331 {
332 
333 	if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
334 		return (1);
335 	else if (curthread->td_owepreempt)
336 		return (1);
337 	else
338 		return (0);
339 }
340 #endif
341 
342 void *vcpu_stats(struct vm *vm, int vcpu);
343 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
344 struct vmspace *vm_get_vmspace(struct vm *vm);
345 struct vatpic *vm_atpic(struct vm *vm);
346 struct vatpit *vm_atpit(struct vm *vm);
347 struct vpmtmr *vm_pmtmr(struct vm *vm);
348 struct vrtc *vm_rtc(struct vm *vm);
349 
350 /*
351  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
352  * success and non-zero on failure.
353  *
354  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
355  * this function directly because they enforce the trap-like or fault-like
356  * behavior of an exception.
357  *
358  * This function should only be called in the context of the thread that is
359  * executing this vcpu.
360  */
361 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
362     uint32_t errcode, int restart_instruction);
363 
364 /*
365  * This function is called after a VM-exit that occurred during exception or
366  * interrupt delivery through the IDT. The format of 'intinfo' is described
367  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
368  *
369  * If a VM-exit handler completes the event delivery successfully then it
370  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
371  * if the task switch emulation is triggered via a task gate then it should
372  * call this function with 'intinfo=0' to indicate that the external event
373  * is not pending anymore.
374  *
375  * Return value is 0 on success and non-zero on failure.
376  */
377 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
378 
379 /*
380  * This function is called before every VM-entry to retrieve a pending
381  * event that should be injected into the guest. This function combines
382  * nested events into a double or triple fault.
383  *
384  * Returns 0 if there are no events that need to be injected into the guest
385  * and non-zero otherwise.
386  */
387 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
388 
389 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
390 
391 enum vm_reg_name vm_segment_name(int seg_encoding);
392 
393 struct vm_copyinfo {
394 	uint64_t	gpa;
395 	size_t		len;
396 	void		*hva;
397 	void		*cookie;
398 };
399 
400 /*
401  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
402  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
403  * a copyin or PROT_WRITE for a copyout.
404  *
405  * retval	is_fault	Interpretation
406  *   0		   0		Success
407  *   0		   1		An exception was injected into the guest
408  * EFAULT	  N/A		Unrecoverable error
409  *
410  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
411  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
412  * 'vm_copy_teardown()' after the copy is done.
413  */
414 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
415     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
416     int num_copyinfo, int *is_fault);
417 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
418     int num_copyinfo);
419 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
420     void *kaddr, size_t len);
421 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
422     struct vm_copyinfo *copyinfo, size_t len);
423 
424 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
425 #endif	/* KERNEL */
426 
427 #define	VM_MAXCPU	16			/* maximum virtual cpus */
428 
429 /*
430  * Identifiers for optional vmm capabilities
431  */
432 enum vm_cap_type {
433 	VM_CAP_HALT_EXIT,
434 	VM_CAP_MTRAP_EXIT,
435 	VM_CAP_PAUSE_EXIT,
436 	VM_CAP_UNRESTRICTED_GUEST,
437 	VM_CAP_ENABLE_INVPCID,
438 	VM_CAP_BPT_EXIT,
439 	VM_CAP_RDPID,
440 	VM_CAP_RDTSCP,
441 	VM_CAP_MAX
442 };
443 
444 enum vm_intr_trigger {
445 	EDGE_TRIGGER,
446 	LEVEL_TRIGGER
447 };
448 
449 /*
450  * The 'access' field has the format specified in Table 21-2 of the Intel
451  * Architecture Manual vol 3b.
452  *
453  * XXX The contents of the 'access' field are architecturally defined except
454  * bit 16 - Segment Unusable.
455  */
456 struct seg_desc {
457 	uint64_t	base;
458 	uint32_t	limit;
459 	uint32_t	access;
460 };
461 #define	SEG_DESC_TYPE(access)		((access) & 0x001f)
462 #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3)
463 #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0)
464 #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0)
465 #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0)
466 #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0)
467 
468 enum vm_cpu_mode {
469 	CPU_MODE_REAL,
470 	CPU_MODE_PROTECTED,
471 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
472 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
473 };
474 
475 enum vm_paging_mode {
476 	PAGING_MODE_FLAT,
477 	PAGING_MODE_32,
478 	PAGING_MODE_PAE,
479 	PAGING_MODE_64,
480 };
481 
482 struct vm_guest_paging {
483 	uint64_t	cr3;
484 	int		cpl;
485 	enum vm_cpu_mode cpu_mode;
486 	enum vm_paging_mode paging_mode;
487 };
488 
489 /*
490  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
491  * consumers of instruction decoding. The only reason why their contents
492  * need to be exposed is because they are part of the 'vm_exit' structure.
493  */
494 struct vie_op {
495 	uint8_t		op_byte;	/* actual opcode byte */
496 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
497 	uint16_t	op_flags;
498 };
499 
500 #define	VIE_INST_SIZE	15
501 struct vie {
502 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
503 	uint8_t		num_valid;		/* size of the instruction */
504 	uint8_t		num_processed;
505 
506 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */
507 	uint8_t		rex_w:1,		/* REX prefix */
508 			rex_r:1,
509 			rex_x:1,
510 			rex_b:1,
511 			rex_present:1,
512 			repz_present:1,		/* REP/REPE/REPZ prefix */
513 			repnz_present:1,	/* REPNE/REPNZ prefix */
514 			opsize_override:1,	/* Operand size override */
515 			addrsize_override:1,	/* Address size override */
516 			segment_override:1;	/* Segment override */
517 
518 	uint8_t		mod:2,			/* ModRM byte */
519 			reg:4,
520 			rm:4;
521 
522 	uint8_t		ss:2,			/* SIB byte */
523 			index:4,
524 			base:4;
525 
526 	uint8_t		disp_bytes;
527 	uint8_t		imm_bytes;
528 
529 	uint8_t		scale;
530 	int		base_register;		/* VM_REG_GUEST_xyz */
531 	int		index_register;		/* VM_REG_GUEST_xyz */
532 	int		segment_register;	/* VM_REG_GUEST_xyz */
533 
534 	int64_t		displacement;		/* optional addr displacement */
535 	int64_t		immediate;		/* optional immediate operand */
536 
537 	uint8_t		decoded;	/* set to 1 if successfully decoded */
538 
539 	struct vie_op	op;			/* opcode description */
540 };
541 
542 enum vm_exitcode {
543 	VM_EXITCODE_INOUT,
544 	VM_EXITCODE_VMX,
545 	VM_EXITCODE_BOGUS,
546 	VM_EXITCODE_RDMSR,
547 	VM_EXITCODE_WRMSR,
548 	VM_EXITCODE_HLT,
549 	VM_EXITCODE_MTRAP,
550 	VM_EXITCODE_PAUSE,
551 	VM_EXITCODE_PAGING,
552 	VM_EXITCODE_INST_EMUL,
553 	VM_EXITCODE_SPINUP_AP,
554 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
555 	VM_EXITCODE_RENDEZVOUS,
556 	VM_EXITCODE_IOAPIC_EOI,
557 	VM_EXITCODE_SUSPENDED,
558 	VM_EXITCODE_INOUT_STR,
559 	VM_EXITCODE_TASK_SWITCH,
560 	VM_EXITCODE_MONITOR,
561 	VM_EXITCODE_MWAIT,
562 	VM_EXITCODE_SVM,
563 	VM_EXITCODE_REQIDLE,
564 	VM_EXITCODE_DEBUG,
565 	VM_EXITCODE_VMINSN,
566 	VM_EXITCODE_BPT,
567 	VM_EXITCODE_MAX
568 };
569 
570 struct vm_inout {
571 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
572 	uint16_t	in:1;
573 	uint16_t	string:1;
574 	uint16_t	rep:1;
575 	uint16_t	port;
576 	uint32_t	eax;		/* valid for out */
577 };
578 
579 struct vm_inout_str {
580 	struct vm_inout	inout;		/* must be the first element */
581 	struct vm_guest_paging paging;
582 	uint64_t	rflags;
583 	uint64_t	cr0;
584 	uint64_t	index;
585 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
586 	int		addrsize;
587 	enum vm_reg_name seg_name;
588 	struct seg_desc seg_desc;
589 };
590 
591 enum task_switch_reason {
592 	TSR_CALL,
593 	TSR_IRET,
594 	TSR_JMP,
595 	TSR_IDT_GATE,	/* task gate in IDT */
596 };
597 
598 struct vm_task_switch {
599 	uint16_t	tsssel;		/* new TSS selector */
600 	int		ext;		/* task switch due to external event */
601 	uint32_t	errcode;
602 	int		errcode_valid;	/* push 'errcode' on the new stack */
603 	enum task_switch_reason reason;
604 	struct vm_guest_paging paging;
605 };
606 
607 struct vm_exit {
608 	enum vm_exitcode	exitcode;
609 	int			inst_length;	/* 0 means unknown */
610 	uint64_t		rip;
611 	union {
612 		struct vm_inout	inout;
613 		struct vm_inout_str inout_str;
614 		struct {
615 			uint64_t	gpa;
616 			int		fault_type;
617 		} paging;
618 		struct {
619 			uint64_t	gpa;
620 			uint64_t	gla;
621 			uint64_t	cs_base;
622 			int		cs_d;		/* CS.D */
623 			struct vm_guest_paging paging;
624 			struct vie	vie;
625 		} inst_emul;
626 		/*
627 		 * VMX specific payload. Used when there is no "better"
628 		 * exitcode to represent the VM-exit.
629 		 */
630 		struct {
631 			int		status;		/* vmx inst status */
632 			/*
633 			 * 'exit_reason' and 'exit_qualification' are valid
634 			 * only if 'status' is zero.
635 			 */
636 			uint32_t	exit_reason;
637 			uint64_t	exit_qualification;
638 			/*
639 			 * 'inst_error' and 'inst_type' are valid
640 			 * only if 'status' is non-zero.
641 			 */
642 			int		inst_type;
643 			int		inst_error;
644 		} vmx;
645 		/*
646 		 * SVM specific payload.
647 		 */
648 		struct {
649 			uint64_t	exitcode;
650 			uint64_t	exitinfo1;
651 			uint64_t	exitinfo2;
652 		} svm;
653 		struct {
654 			int		inst_length;
655 		} bpt;
656 		struct {
657 			uint32_t	code;		/* ecx value */
658 			uint64_t	wval;
659 		} msr;
660 		struct {
661 			int		vcpu;
662 			uint64_t	rip;
663 		} spinup_ap;
664 		struct {
665 			uint64_t	rflags;
666 			uint64_t	intr_status;
667 		} hlt;
668 		struct {
669 			int		vector;
670 		} ioapic_eoi;
671 		struct {
672 			enum vm_suspend_how how;
673 		} suspended;
674 		struct vm_task_switch task_switch;
675 	} u;
676 };
677 
678 /* APIs to inject faults into the guest */
679 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
680     int errcode);
681 
682 static __inline void
vm_inject_ud(void * vm,int vcpuid)683 vm_inject_ud(void *vm, int vcpuid)
684 {
685 	vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
686 }
687 
688 static __inline void
vm_inject_gp(void * vm,int vcpuid)689 vm_inject_gp(void *vm, int vcpuid)
690 {
691 	vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
692 }
693 
694 static __inline void
vm_inject_ac(void * vm,int vcpuid,int errcode)695 vm_inject_ac(void *vm, int vcpuid, int errcode)
696 {
697 	vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
698 }
699 
700 static __inline void
vm_inject_ss(void * vm,int vcpuid,int errcode)701 vm_inject_ss(void *vm, int vcpuid, int errcode)
702 {
703 	vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
704 }
705 
706 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
707 
708 int vm_restart_instruction(void *vm, int vcpuid);
709 
710 #endif	/* _VMM_H_ */
711