xref: /freebsd/sys/amd64/include/vmm.h (revision 0bf48626aaa33768078f5872b922b1487b3a9296)
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 <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_LAST
99 };
100 
101 enum x2apic_state {
102 	X2APIC_DISABLED,
103 	X2APIC_ENABLED,
104 	X2APIC_STATE_LAST
105 };
106 
107 #define	VM_INTINFO_VECTOR(info)	((info) & 0xff)
108 #define	VM_INTINFO_DEL_ERRCODE	0x800
109 #define	VM_INTINFO_RSVD		0x7ffff000
110 #define	VM_INTINFO_VALID	0x80000000
111 #define	VM_INTINFO_TYPE		0x700
112 #define	VM_INTINFO_HWINTR	(0 << 8)
113 #define	VM_INTINFO_NMI		(2 << 8)
114 #define	VM_INTINFO_HWEXCEPTION	(3 << 8)
115 #define	VM_INTINFO_SWINTR	(4 << 8)
116 
117 /*
118  * The VM name has to fit into the pathname length constraints of devfs,
119  * governed primarily by SPECNAMELEN.  The length is the total number of
120  * characters in the full path, relative to the mount point and not
121  * including any leading '/' characters.
122  * A prefix and a suffix are added to the name specified by the user.
123  * The prefix is usually "vmm/" or "vmm.io/", but can be a few characters
124  * longer for future use.
125  * The suffix is a string that identifies a bootrom image or some similar
126  * image that is attached to the VM. A separator character gets added to
127  * the suffix automatically when generating the full path, so it must be
128  * accounted for, reducing the effective length by 1.
129  * The effective length of a VM name is 229 bytes for FreeBSD 13 and 37
130  * bytes for FreeBSD 12.  A minimum length is set for safety and supports
131  * a SPECNAMELEN as small as 32 on old systems.
132  */
133 #define VM_MAX_PREFIXLEN 10
134 #define VM_MAX_SUFFIXLEN 15
135 #define VM_MIN_NAMELEN   6
136 #define VM_MAX_NAMELEN \
137     (SPECNAMELEN - VM_MAX_PREFIXLEN - VM_MAX_SUFFIXLEN - 1)
138 
139 #ifdef _KERNEL
140 CTASSERT(VM_MAX_NAMELEN >= VM_MIN_NAMELEN);
141 
142 struct vm;
143 struct vm_exception;
144 struct seg_desc;
145 struct vm_exit;
146 struct vm_run;
147 struct vhpet;
148 struct vioapic;
149 struct vlapic;
150 struct vmspace;
151 struct vm_object;
152 struct vm_guest_paging;
153 struct pmap;
154 
155 struct vm_eventinfo {
156 	void	*rptr;		/* rendezvous cookie */
157 	int	*sptr;		/* suspend cookie */
158 	int	*iptr;		/* reqidle cookie */
159 };
160 
161 typedef int	(*vmm_init_func_t)(int ipinum);
162 typedef int	(*vmm_cleanup_func_t)(void);
163 typedef void	(*vmm_resume_func_t)(void);
164 typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
165 typedef int	(*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
166 		    struct pmap *pmap, struct vm_eventinfo *info);
167 typedef void	(*vmi_cleanup_func_t)(void *vmi);
168 typedef int	(*vmi_get_register_t)(void *vmi, int vcpu, int num,
169 				      uint64_t *retval);
170 typedef int	(*vmi_set_register_t)(void *vmi, int vcpu, int num,
171 				      uint64_t val);
172 typedef int	(*vmi_get_desc_t)(void *vmi, int vcpu, int num,
173 				  struct seg_desc *desc);
174 typedef int	(*vmi_set_desc_t)(void *vmi, int vcpu, int num,
175 				  struct seg_desc *desc);
176 typedef int	(*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
177 typedef int	(*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
178 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
179 typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace);
180 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
181 typedef void	(*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
182 
183 struct vmm_ops {
184 	vmm_init_func_t		init;		/* module wide initialization */
185 	vmm_cleanup_func_t	cleanup;
186 	vmm_resume_func_t	resume;
187 
188 	vmi_init_func_t		vminit;		/* vm-specific initialization */
189 	vmi_run_func_t		vmrun;
190 	vmi_cleanup_func_t	vmcleanup;
191 	vmi_get_register_t	vmgetreg;
192 	vmi_set_register_t	vmsetreg;
193 	vmi_get_desc_t		vmgetdesc;
194 	vmi_set_desc_t		vmsetdesc;
195 	vmi_get_cap_t		vmgetcap;
196 	vmi_set_cap_t		vmsetcap;
197 	vmi_vmspace_alloc	vmspace_alloc;
198 	vmi_vmspace_free	vmspace_free;
199 	vmi_vlapic_init		vlapic_init;
200 	vmi_vlapic_cleanup	vlapic_cleanup;
201 };
202 
203 extern struct vmm_ops vmm_ops_intel;
204 extern struct vmm_ops vmm_ops_amd;
205 
206 int vm_create(const char *name, struct vm **retvm);
207 void vm_destroy(struct vm *vm);
208 int vm_reinit(struct vm *vm);
209 const char *vm_name(struct vm *vm);
210 uint16_t vm_get_maxcpus(struct vm *vm);
211 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
212     uint16_t *threads, uint16_t *maxcpus);
213 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
214     uint16_t threads, uint16_t maxcpus);
215 
216 /*
217  * APIs that modify the guest memory map require all vcpus to be frozen.
218  */
219 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
220     size_t len, int prot, int flags);
221 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
222 void vm_free_memseg(struct vm *vm, int ident);
223 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
224 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
225 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
226 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
227 
228 /*
229  * APIs that inspect the guest memory map require only a *single* vcpu to
230  * be frozen. This acts like a read lock on the guest memory map since any
231  * modification requires *all* vcpus to be frozen.
232  */
233 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
234     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
235 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
236     struct vm_object **objptr);
237 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
238 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
239     int prot, void **cookie);
240 void vm_gpa_release(void *cookie);
241 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
242 
243 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
244 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
245 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
246 		    struct seg_desc *ret_desc);
247 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
248 		    struct seg_desc *desc);
249 int vm_run(struct vm *vm, struct vm_run *vmrun);
250 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
251 int vm_inject_nmi(struct vm *vm, int vcpu);
252 int vm_nmi_pending(struct vm *vm, int vcpuid);
253 void vm_nmi_clear(struct vm *vm, int vcpuid);
254 int vm_inject_extint(struct vm *vm, int vcpu);
255 int vm_extint_pending(struct vm *vm, int vcpuid);
256 void vm_extint_clear(struct vm *vm, int vcpuid);
257 struct vlapic *vm_lapic(struct vm *vm, int cpu);
258 struct vioapic *vm_ioapic(struct vm *vm);
259 struct vhpet *vm_hpet(struct vm *vm);
260 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
261 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
262 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
263 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
264 int vm_apicid2vcpuid(struct vm *vm, int apicid);
265 int vm_activate_cpu(struct vm *vm, int vcpu);
266 int vm_suspend_cpu(struct vm *vm, int vcpu);
267 int vm_resume_cpu(struct vm *vm, int vcpu);
268 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
269 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
270 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip);
271 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
272 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
273 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
274 
275 #ifdef _SYS__CPUSET_H_
276 /*
277  * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
278  * The rendezvous 'func(arg)' is not allowed to do anything that will
279  * cause the thread to be put to sleep.
280  *
281  * If the rendezvous is being initiated from a vcpu context then the
282  * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
283  *
284  * The caller cannot hold any locks when initiating the rendezvous.
285  *
286  * The implementation of this API may cause vcpus other than those specified
287  * by 'dest' to be stalled. The caller should not rely on any vcpus making
288  * forward progress when the rendezvous is in progress.
289  */
290 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
291 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
292     vm_rendezvous_func_t func, void *arg);
293 cpuset_t vm_active_cpus(struct vm *vm);
294 cpuset_t vm_debug_cpus(struct vm *vm);
295 cpuset_t vm_suspended_cpus(struct vm *vm);
296 #endif	/* _SYS__CPUSET_H_ */
297 
298 static __inline int
299 vcpu_rendezvous_pending(struct vm_eventinfo *info)
300 {
301 
302 	return (*((uintptr_t *)(info->rptr)) != 0);
303 }
304 
305 static __inline int
306 vcpu_suspended(struct vm_eventinfo *info)
307 {
308 
309 	return (*info->sptr);
310 }
311 
312 static __inline int
313 vcpu_reqidle(struct vm_eventinfo *info)
314 {
315 
316 	return (*info->iptr);
317 }
318 
319 int vcpu_debugged(struct vm *vm, int vcpuid);
320 
321 /*
322  * Return true if device indicated by bus/slot/func is supposed to be a
323  * pci passthrough device.
324  *
325  * Return false otherwise.
326  */
327 bool vmm_is_pptdev(int bus, int slot, int func);
328 
329 void *vm_iommu_domain(struct vm *vm);
330 
331 enum vcpu_state {
332 	VCPU_IDLE,
333 	VCPU_FROZEN,
334 	VCPU_RUNNING,
335 	VCPU_SLEEPING,
336 };
337 
338 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
339     bool from_idle);
340 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
341 
342 static int __inline
343 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
344 {
345 	return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
346 }
347 
348 #ifdef _SYS_PROC_H_
349 static int __inline
350 vcpu_should_yield(struct vm *vm, int vcpu)
351 {
352 
353 	if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
354 		return (1);
355 	else if (curthread->td_owepreempt)
356 		return (1);
357 	else
358 		return (0);
359 }
360 #endif
361 
362 void *vcpu_stats(struct vm *vm, int vcpu);
363 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
364 struct vmspace *vm_get_vmspace(struct vm *vm);
365 struct vatpic *vm_atpic(struct vm *vm);
366 struct vatpit *vm_atpit(struct vm *vm);
367 struct vpmtmr *vm_pmtmr(struct vm *vm);
368 struct vrtc *vm_rtc(struct vm *vm);
369 
370 /*
371  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
372  * success and non-zero on failure.
373  *
374  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
375  * this function directly because they enforce the trap-like or fault-like
376  * behavior of an exception.
377  *
378  * This function should only be called in the context of the thread that is
379  * executing this vcpu.
380  */
381 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
382     uint32_t errcode, int restart_instruction);
383 
384 /*
385  * This function is called after a VM-exit that occurred during exception or
386  * interrupt delivery through the IDT. The format of 'intinfo' is described
387  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
388  *
389  * If a VM-exit handler completes the event delivery successfully then it
390  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
391  * if the task switch emulation is triggered via a task gate then it should
392  * call this function with 'intinfo=0' to indicate that the external event
393  * is not pending anymore.
394  *
395  * Return value is 0 on success and non-zero on failure.
396  */
397 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
398 
399 /*
400  * This function is called before every VM-entry to retrieve a pending
401  * event that should be injected into the guest. This function combines
402  * nested events into a double or triple fault.
403  *
404  * Returns 0 if there are no events that need to be injected into the guest
405  * and non-zero otherwise.
406  */
407 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
408 
409 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
410 
411 enum vm_reg_name vm_segment_name(int seg_encoding);
412 
413 struct vm_copyinfo {
414 	uint64_t	gpa;
415 	size_t		len;
416 	void		*hva;
417 	void		*cookie;
418 };
419 
420 /*
421  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
422  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
423  * a copyin or PROT_WRITE for a copyout.
424  *
425  * retval	is_fault	Interpretation
426  *   0		   0		Success
427  *   0		   1		An exception was injected into the guest
428  * EFAULT	  N/A		Unrecoverable error
429  *
430  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
431  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
432  * 'vm_copy_teardown()' after the copy is done.
433  */
434 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
435     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
436     int num_copyinfo, int *is_fault);
437 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
438     int num_copyinfo);
439 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
440     void *kaddr, size_t len);
441 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
442     struct vm_copyinfo *copyinfo, size_t len);
443 
444 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
445 #endif	/* KERNEL */
446 
447 #define	VM_MAXCPU	16			/* maximum virtual cpus */
448 
449 /*
450  * Identifiers for optional vmm capabilities
451  */
452 enum vm_cap_type {
453 	VM_CAP_HALT_EXIT,
454 	VM_CAP_MTRAP_EXIT,
455 	VM_CAP_PAUSE_EXIT,
456 	VM_CAP_UNRESTRICTED_GUEST,
457 	VM_CAP_ENABLE_INVPCID,
458 	VM_CAP_MAX
459 };
460 
461 enum vm_intr_trigger {
462 	EDGE_TRIGGER,
463 	LEVEL_TRIGGER
464 };
465 
466 /*
467  * The 'access' field has the format specified in Table 21-2 of the Intel
468  * Architecture Manual vol 3b.
469  *
470  * XXX The contents of the 'access' field are architecturally defined except
471  * bit 16 - Segment Unusable.
472  */
473 struct seg_desc {
474 	uint64_t	base;
475 	uint32_t	limit;
476 	uint32_t	access;
477 };
478 #define	SEG_DESC_TYPE(access)		((access) & 0x001f)
479 #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3)
480 #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0)
481 #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0)
482 #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0)
483 #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0)
484 
485 enum vm_cpu_mode {
486 	CPU_MODE_REAL,
487 	CPU_MODE_PROTECTED,
488 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */
489 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */
490 };
491 
492 enum vm_paging_mode {
493 	PAGING_MODE_FLAT,
494 	PAGING_MODE_32,
495 	PAGING_MODE_PAE,
496 	PAGING_MODE_64,
497 };
498 
499 struct vm_guest_paging {
500 	uint64_t	cr3;
501 	int		cpl;
502 	enum vm_cpu_mode cpu_mode;
503 	enum vm_paging_mode paging_mode;
504 };
505 
506 /*
507  * The data structures 'vie' and 'vie_op' are meant to be opaque to the
508  * consumers of instruction decoding. The only reason why their contents
509  * need to be exposed is because they are part of the 'vm_exit' structure.
510  */
511 struct vie_op {
512 	uint8_t		op_byte;	/* actual opcode byte */
513 	uint8_t		op_type;	/* type of operation (e.g. MOV) */
514 	uint16_t	op_flags;
515 };
516 
517 #define	VIE_INST_SIZE	15
518 struct vie {
519 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */
520 	uint8_t		num_valid;		/* size of the instruction */
521 	uint8_t		num_processed;
522 
523 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */
524 	uint8_t		rex_w:1,		/* REX prefix */
525 			rex_r:1,
526 			rex_x:1,
527 			rex_b:1,
528 			rex_present:1,
529 			repz_present:1,		/* REP/REPE/REPZ prefix */
530 			repnz_present:1,	/* REPNE/REPNZ prefix */
531 			opsize_override:1,	/* Operand size override */
532 			addrsize_override:1,	/* Address size override */
533 			segment_override:1;	/* Segment override */
534 
535 	uint8_t		mod:2,			/* ModRM byte */
536 			reg:4,
537 			rm:4;
538 
539 	uint8_t		ss:2,			/* SIB byte */
540 			index:4,
541 			base:4;
542 
543 	uint8_t		disp_bytes;
544 	uint8_t		imm_bytes;
545 
546 	uint8_t		scale;
547 	int		base_register;		/* VM_REG_GUEST_xyz */
548 	int		index_register;		/* VM_REG_GUEST_xyz */
549 	int		segment_register;	/* VM_REG_GUEST_xyz */
550 
551 	int64_t		displacement;		/* optional addr displacement */
552 	int64_t		immediate;		/* optional immediate operand */
553 
554 	uint8_t		decoded;	/* set to 1 if successfully decoded */
555 
556 	struct vie_op	op;			/* opcode description */
557 };
558 
559 enum vm_exitcode {
560 	VM_EXITCODE_INOUT,
561 	VM_EXITCODE_VMX,
562 	VM_EXITCODE_BOGUS,
563 	VM_EXITCODE_RDMSR,
564 	VM_EXITCODE_WRMSR,
565 	VM_EXITCODE_HLT,
566 	VM_EXITCODE_MTRAP,
567 	VM_EXITCODE_PAUSE,
568 	VM_EXITCODE_PAGING,
569 	VM_EXITCODE_INST_EMUL,
570 	VM_EXITCODE_SPINUP_AP,
571 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */
572 	VM_EXITCODE_RENDEZVOUS,
573 	VM_EXITCODE_IOAPIC_EOI,
574 	VM_EXITCODE_SUSPENDED,
575 	VM_EXITCODE_INOUT_STR,
576 	VM_EXITCODE_TASK_SWITCH,
577 	VM_EXITCODE_MONITOR,
578 	VM_EXITCODE_MWAIT,
579 	VM_EXITCODE_SVM,
580 	VM_EXITCODE_REQIDLE,
581 	VM_EXITCODE_DEBUG,
582 	VM_EXITCODE_VMINSN,
583 	VM_EXITCODE_MAX
584 };
585 
586 struct vm_inout {
587 	uint16_t	bytes:3;	/* 1 or 2 or 4 */
588 	uint16_t	in:1;
589 	uint16_t	string:1;
590 	uint16_t	rep:1;
591 	uint16_t	port;
592 	uint32_t	eax;		/* valid for out */
593 };
594 
595 struct vm_inout_str {
596 	struct vm_inout	inout;		/* must be the first element */
597 	struct vm_guest_paging paging;
598 	uint64_t	rflags;
599 	uint64_t	cr0;
600 	uint64_t	index;
601 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */
602 	int		addrsize;
603 	enum vm_reg_name seg_name;
604 	struct seg_desc seg_desc;
605 };
606 
607 enum task_switch_reason {
608 	TSR_CALL,
609 	TSR_IRET,
610 	TSR_JMP,
611 	TSR_IDT_GATE,	/* task gate in IDT */
612 };
613 
614 struct vm_task_switch {
615 	uint16_t	tsssel;		/* new TSS selector */
616 	int		ext;		/* task switch due to external event */
617 	uint32_t	errcode;
618 	int		errcode_valid;	/* push 'errcode' on the new stack */
619 	enum task_switch_reason reason;
620 	struct vm_guest_paging paging;
621 };
622 
623 struct vm_exit {
624 	enum vm_exitcode	exitcode;
625 	int			inst_length;	/* 0 means unknown */
626 	uint64_t		rip;
627 	union {
628 		struct vm_inout	inout;
629 		struct vm_inout_str inout_str;
630 		struct {
631 			uint64_t	gpa;
632 			int		fault_type;
633 		} paging;
634 		struct {
635 			uint64_t	gpa;
636 			uint64_t	gla;
637 			uint64_t	cs_base;
638 			int		cs_d;		/* CS.D */
639 			struct vm_guest_paging paging;
640 			struct vie	vie;
641 		} inst_emul;
642 		/*
643 		 * VMX specific payload. Used when there is no "better"
644 		 * exitcode to represent the VM-exit.
645 		 */
646 		struct {
647 			int		status;		/* vmx inst status */
648 			/*
649 			 * 'exit_reason' and 'exit_qualification' are valid
650 			 * only if 'status' is zero.
651 			 */
652 			uint32_t	exit_reason;
653 			uint64_t	exit_qualification;
654 			/*
655 			 * 'inst_error' and 'inst_type' are valid
656 			 * only if 'status' is non-zero.
657 			 */
658 			int		inst_type;
659 			int		inst_error;
660 		} vmx;
661 		/*
662 		 * SVM specific payload.
663 		 */
664 		struct {
665 			uint64_t	exitcode;
666 			uint64_t	exitinfo1;
667 			uint64_t	exitinfo2;
668 		} svm;
669 		struct {
670 			uint32_t	code;		/* ecx value */
671 			uint64_t	wval;
672 		} msr;
673 		struct {
674 			int		vcpu;
675 			uint64_t	rip;
676 		} spinup_ap;
677 		struct {
678 			uint64_t	rflags;
679 			uint64_t	intr_status;
680 		} hlt;
681 		struct {
682 			int		vector;
683 		} ioapic_eoi;
684 		struct {
685 			enum vm_suspend_how how;
686 		} suspended;
687 		struct vm_task_switch task_switch;
688 	} u;
689 };
690 
691 /* APIs to inject faults into the guest */
692 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
693     int errcode);
694 
695 static __inline void
696 vm_inject_ud(void *vm, int vcpuid)
697 {
698 	vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
699 }
700 
701 static __inline void
702 vm_inject_gp(void *vm, int vcpuid)
703 {
704 	vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
705 }
706 
707 static __inline void
708 vm_inject_ac(void *vm, int vcpuid, int errcode)
709 {
710 	vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
711 }
712 
713 static __inline void
714 vm_inject_ss(void *vm, int vcpuid, int errcode)
715 {
716 	vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
717 }
718 
719 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
720 
721 int vm_restart_instruction(void *vm, int vcpuid);
722 
723 #endif	/* _VMM_H_ */
724