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