1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
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
29 #ifndef _VMM_H_
30 #define _VMM_H_
31
32 #include <sys/cpuset.h>
33 #include <sys/sdt.h>
34 #include <x86/segments.h>
35
36 struct vcpu;
37 struct vm_snapshot_meta;
38
39 #ifdef _KERNEL
40 SDT_PROVIDER_DECLARE(vmm);
41 #endif
42
43 enum vm_suspend_how {
44 VM_SUSPEND_NONE,
45 VM_SUSPEND_RESET,
46 VM_SUSPEND_POWEROFF,
47 VM_SUSPEND_HALT,
48 VM_SUSPEND_TRIPLEFAULT,
49 VM_SUSPEND_DESTROY,
50 VM_SUSPEND_LAST
51 };
52
53 /*
54 * Identifiers for architecturally defined registers.
55 */
56 enum vm_reg_name {
57 VM_REG_GUEST_RAX,
58 VM_REG_GUEST_RBX,
59 VM_REG_GUEST_RCX,
60 VM_REG_GUEST_RDX,
61 VM_REG_GUEST_RSI,
62 VM_REG_GUEST_RDI,
63 VM_REG_GUEST_RBP,
64 VM_REG_GUEST_R8,
65 VM_REG_GUEST_R9,
66 VM_REG_GUEST_R10,
67 VM_REG_GUEST_R11,
68 VM_REG_GUEST_R12,
69 VM_REG_GUEST_R13,
70 VM_REG_GUEST_R14,
71 VM_REG_GUEST_R15,
72 VM_REG_GUEST_CR0,
73 VM_REG_GUEST_CR3,
74 VM_REG_GUEST_CR4,
75 VM_REG_GUEST_DR7,
76 VM_REG_GUEST_RSP,
77 VM_REG_GUEST_RIP,
78 VM_REG_GUEST_RFLAGS,
79 VM_REG_GUEST_ES,
80 VM_REG_GUEST_CS,
81 VM_REG_GUEST_SS,
82 VM_REG_GUEST_DS,
83 VM_REG_GUEST_FS,
84 VM_REG_GUEST_GS,
85 VM_REG_GUEST_LDTR,
86 VM_REG_GUEST_TR,
87 VM_REG_GUEST_IDTR,
88 VM_REG_GUEST_GDTR,
89 VM_REG_GUEST_EFER,
90 VM_REG_GUEST_CR2,
91 VM_REG_GUEST_PDPTE0,
92 VM_REG_GUEST_PDPTE1,
93 VM_REG_GUEST_PDPTE2,
94 VM_REG_GUEST_PDPTE3,
95 VM_REG_GUEST_INTR_SHADOW,
96 VM_REG_GUEST_DR0,
97 VM_REG_GUEST_DR1,
98 VM_REG_GUEST_DR2,
99 VM_REG_GUEST_DR3,
100 VM_REG_GUEST_DR6,
101 VM_REG_GUEST_ENTRY_INST_LENGTH,
102 VM_REG_GUEST_FS_BASE,
103 VM_REG_GUEST_GS_BASE,
104 VM_REG_GUEST_KGS_BASE,
105 VM_REG_GUEST_TPR,
106 VM_REG_LAST
107 };
108
109 enum x2apic_state {
110 X2APIC_DISABLED,
111 X2APIC_ENABLED,
112 X2APIC_STATE_LAST
113 };
114
115 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
116 #define VM_INTINFO_DEL_ERRCODE 0x800
117 #define VM_INTINFO_RSVD 0x7ffff000
118 #define VM_INTINFO_VALID 0x80000000
119 #define VM_INTINFO_TYPE 0x700
120 #define VM_INTINFO_HWINTR (0 << 8)
121 #define VM_INTINFO_NMI (2 << 8)
122 #define VM_INTINFO_HWEXCEPTION (3 << 8)
123 #define VM_INTINFO_SWINTR (4 << 8)
124
125 #ifdef _KERNEL
126 struct vm;
127 struct vm_exception;
128 struct vm_mem;
129 struct seg_desc;
130 struct vm_exit;
131 struct vm_run;
132 struct vhpet;
133 struct vioapic;
134 struct vlapic;
135 struct vmspace;
136 struct vm_object;
137 struct vm_guest_paging;
138 struct pmap;
139 enum snapshot_req;
140
141 struct vm_eventinfo {
142 cpuset_t *rptr; /* rendezvous cookie */
143 int *sptr; /* suspend cookie */
144 int *iptr; /* reqidle cookie */
145 };
146
147 #define DECLARE_VMMOPS_FUNC(ret_type, opname, args) \
148 typedef ret_type (*vmmops_##opname##_t) args; \
149 ret_type vmmops_##opname args
150
151 DECLARE_VMMOPS_FUNC(int, modinit, (int ipinum));
152 DECLARE_VMMOPS_FUNC(int, modcleanup, (void));
153 DECLARE_VMMOPS_FUNC(void, modresume, (void));
154 DECLARE_VMMOPS_FUNC(void, modsuspend, (void));
155 DECLARE_VMMOPS_FUNC(void *, init, (struct vm *vm, struct pmap *pmap));
156 DECLARE_VMMOPS_FUNC(int, run, (void *vcpui, register_t pc,
157 struct pmap *pmap, struct vm_eventinfo *info));
158 DECLARE_VMMOPS_FUNC(void, cleanup, (void *vmi));
159 DECLARE_VMMOPS_FUNC(void *, vcpu_init, (void *vmi, struct vcpu *vcpu,
160 int vcpu_id));
161 DECLARE_VMMOPS_FUNC(void, vcpu_cleanup, (void *vcpui));
162 DECLARE_VMMOPS_FUNC(int, getreg, (void *vcpui, int num, uint64_t *retval));
163 DECLARE_VMMOPS_FUNC(int, setreg, (void *vcpui, int num, uint64_t val));
164 DECLARE_VMMOPS_FUNC(int, getdesc, (void *vcpui, int num,
165 struct seg_desc *desc));
166 DECLARE_VMMOPS_FUNC(int, setdesc, (void *vcpui, int num,
167 struct seg_desc *desc));
168 DECLARE_VMMOPS_FUNC(int, getcap, (void *vcpui, int num, int *retval));
169 DECLARE_VMMOPS_FUNC(int, setcap, (void *vcpui, int num, int val));
170 DECLARE_VMMOPS_FUNC(struct vmspace *, vmspace_alloc,
171 (vm_offset_t min, vm_offset_t max));
172 DECLARE_VMMOPS_FUNC(void, vmspace_free, (struct vmspace *vmspace));
173 DECLARE_VMMOPS_FUNC(struct vlapic *, vlapic_init, (void *vcpui));
174 DECLARE_VMMOPS_FUNC(void, vlapic_cleanup, (struct vlapic *vlapic));
175 DECLARE_VMMOPS_FUNC(int, vcpu_snapshot, (void *vcpui,
176 struct vm_snapshot_meta *meta));
177 DECLARE_VMMOPS_FUNC(int, restore_tsc, (void *vcpui, uint64_t now));
178
179 struct vmm_ops {
180 vmmops_modinit_t modinit; /* module wide initialization */
181 vmmops_modcleanup_t modcleanup;
182 vmmops_modresume_t modsuspend;
183 vmmops_modresume_t modresume;
184
185 vmmops_init_t init; /* vm-specific initialization */
186 vmmops_run_t run;
187 vmmops_cleanup_t cleanup;
188 vmmops_vcpu_init_t vcpu_init;
189 vmmops_vcpu_cleanup_t vcpu_cleanup;
190 vmmops_getreg_t getreg;
191 vmmops_setreg_t setreg;
192 vmmops_getdesc_t getdesc;
193 vmmops_setdesc_t setdesc;
194 vmmops_getcap_t getcap;
195 vmmops_setcap_t setcap;
196 vmmops_vmspace_alloc_t vmspace_alloc;
197 vmmops_vmspace_free_t vmspace_free;
198 vmmops_vlapic_init_t vlapic_init;
199 vmmops_vlapic_cleanup_t vlapic_cleanup;
200
201 /* checkpoint operations */
202 vmmops_vcpu_snapshot_t vcpu_snapshot;
203 vmmops_restore_tsc_t restore_tsc;
204 };
205
206 extern const struct vmm_ops vmm_ops_intel;
207 extern const struct vmm_ops vmm_ops_amd;
208
209 int vm_create(const char *name, struct vm **retvm);
210 struct vcpu *vm_alloc_vcpu(struct vm *vm, int vcpuid);
211 void vm_disable_vcpu_creation(struct vm *vm);
212 void vm_lock_vcpus(struct vm *vm);
213 void vm_unlock_vcpus(struct vm *vm);
214 void vm_destroy(struct vm *vm);
215 int vm_reinit(struct vm *vm);
216 const char *vm_name(struct vm *vm);
217 uint16_t vm_get_maxcpus(struct vm *vm);
218 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
219 uint16_t *threads, uint16_t *maxcpus);
220 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
221 uint16_t threads, uint16_t maxcpus);
222
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 int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval);
229 int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val);
230 int vm_get_seg_desc(struct vcpu *vcpu, int reg,
231 struct seg_desc *ret_desc);
232 int vm_set_seg_desc(struct vcpu *vcpu, int reg,
233 struct seg_desc *desc);
234 int vm_run(struct vcpu *vcpu);
235 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
236 int vm_inject_nmi(struct vcpu *vcpu);
237 int vm_nmi_pending(struct vcpu *vcpu);
238 void vm_nmi_clear(struct vcpu *vcpu);
239 int vm_inject_extint(struct vcpu *vcpu);
240 int vm_extint_pending(struct vcpu *vcpu);
241 void vm_extint_clear(struct vcpu *vcpu);
242 int vcpu_vcpuid(struct vcpu *vcpu);
243 struct vm *vcpu_vm(struct vcpu *vcpu);
244 struct vcpu *vm_vcpu(struct vm *vm, int cpu);
245 struct vlapic *vm_lapic(struct vcpu *vcpu);
246 struct vioapic *vm_ioapic(struct vm *vm);
247 struct vhpet *vm_hpet(struct vm *vm);
248 int vm_get_capability(struct vcpu *vcpu, int type, int *val);
249 int vm_set_capability(struct vcpu *vcpu, int type, int val);
250 int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state);
251 int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state);
252 int vm_apicid2vcpuid(struct vm *vm, int apicid);
253 int vm_activate_cpu(struct vcpu *vcpu);
254 int vm_suspend_cpu(struct vm *vm, struct vcpu *vcpu);
255 int vm_resume_cpu(struct vm *vm, struct vcpu *vcpu);
256 int vm_restart_instruction(struct vcpu *vcpu);
257 struct vm_exit *vm_exitinfo(struct vcpu *vcpu);
258 cpuset_t *vm_exitinfo_cpuset(struct vcpu *vcpu);
259 void vm_exit_suspended(struct vcpu *vcpu, uint64_t rip);
260 void vm_exit_debug(struct vcpu *vcpu, uint64_t rip);
261 void vm_exit_rendezvous(struct vcpu *vcpu, uint64_t rip);
262 void vm_exit_astpending(struct vcpu *vcpu, uint64_t rip);
263 void vm_exit_reqidle(struct vcpu *vcpu, uint64_t rip);
264 int vm_snapshot_req(struct vm *vm, struct vm_snapshot_meta *meta);
265 int vm_restore_time(struct vm *vm);
266
267 #ifdef _SYS__CPUSET_H_
268 /*
269 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
270 * The rendezvous 'func(arg)' is not allowed to do anything that will
271 * cause the thread to be put to sleep.
272 *
273 * The caller cannot hold any locks when initiating the rendezvous.
274 *
275 * The implementation of this API may cause vcpus other than those specified
276 * by 'dest' to be stalled. The caller should not rely on any vcpus making
277 * forward progress when the rendezvous is in progress.
278 */
279 typedef void (*vm_rendezvous_func_t)(struct vcpu *vcpu, void *arg);
280 int vm_smp_rendezvous(struct vcpu *vcpu, cpuset_t dest,
281 vm_rendezvous_func_t func, void *arg);
282
283 cpuset_t vm_active_cpus(struct vm *vm);
284 cpuset_t vm_debug_cpus(struct vm *vm);
285 cpuset_t vm_suspended_cpus(struct vm *vm);
286 cpuset_t vm_start_cpus(struct vm *vm, const cpuset_t *tostart);
287 void vm_await_start(struct vm *vm, const cpuset_t *waiting);
288 #endif /* _SYS__CPUSET_H_ */
289
290 static __inline int
vcpu_rendezvous_pending(struct vcpu * vcpu,struct vm_eventinfo * info)291 vcpu_rendezvous_pending(struct vcpu *vcpu, struct vm_eventinfo *info)
292 {
293 /*
294 * This check isn't done with atomic operations or under a lock because
295 * there's no need to. If the vcpuid bit is set, the vcpu is part of a
296 * rendezvous and the bit won't be cleared until the vcpu enters the
297 * rendezvous. On rendezvous exit, the cpuset is cleared and the vcpu
298 * will see an empty cpuset. So, the races are harmless.
299 */
300 return (CPU_ISSET(vcpu_vcpuid(vcpu), info->rptr));
301 }
302
303 static __inline int
vcpu_suspended(struct vm_eventinfo * info)304 vcpu_suspended(struct vm_eventinfo *info)
305 {
306
307 return (*info->sptr);
308 }
309
310 static __inline int
vcpu_reqidle(struct vm_eventinfo * info)311 vcpu_reqidle(struct vm_eventinfo *info)
312 {
313
314 return (*info->iptr);
315 }
316
317 int vcpu_debugged(struct vcpu *vcpu);
318
319 /*
320 * Return true if device indicated by bus/slot/func is supposed to be a
321 * pci passthrough device.
322 *
323 * Return false otherwise.
324 */
325 bool vmm_is_pptdev(int bus, int slot, int func);
326
327 void *vm_iommu_domain(struct vm *vm);
328
329 enum vcpu_state {
330 VCPU_IDLE,
331 VCPU_FROZEN,
332 VCPU_RUNNING,
333 VCPU_SLEEPING,
334 };
335
336 int vcpu_set_state(struct vcpu *vcpu, enum vcpu_state state, bool from_idle);
337 int vcpu_set_state_all(struct vm *vm, enum vcpu_state state);
338 enum vcpu_state vcpu_get_state(struct vcpu *vcpu, int *hostcpu);
339
340 static int __inline
vcpu_is_running(struct vcpu * vcpu,int * hostcpu)341 vcpu_is_running(struct vcpu *vcpu, int *hostcpu)
342 {
343 return (vcpu_get_state(vcpu, hostcpu) == VCPU_RUNNING);
344 }
345
346 #ifdef _SYS_PROC_H_
347 static int __inline
vcpu_should_yield(struct vcpu * vcpu)348 vcpu_should_yield(struct vcpu *vcpu)
349 {
350 struct thread *td;
351
352 td = curthread;
353 return (td->td_ast != 0 || td->td_owepreempt != 0);
354 }
355 #endif
356
357 void *vcpu_stats(struct vcpu *vcpu);
358 void vcpu_notify_event(struct vcpu *vcpu);
359 void vcpu_notify_lapic(struct vcpu *vcpu);
360 struct vm_mem *vm_mem(struct vm *vm);
361 struct vatpic *vm_atpic(struct vm *vm);
362 struct vatpit *vm_atpit(struct vm *vm);
363 struct vpmtmr *vm_pmtmr(struct vm *vm);
364 struct vrtc *vm_rtc(struct vm *vm);
365
366 /*
367 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
368 * success and non-zero on failure.
369 *
370 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
371 * this function directly because they enforce the trap-like or fault-like
372 * behavior of an exception.
373 *
374 * This function should only be called in the context of the thread that is
375 * executing this vcpu.
376 */
377 int vm_inject_exception(struct vcpu *vcpu, int vector, int err_valid,
378 uint32_t errcode, int restart_instruction);
379
380 /*
381 * This function is called after a VM-exit that occurred during exception or
382 * interrupt delivery through the IDT. The format of 'intinfo' is described
383 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
384 *
385 * If a VM-exit handler completes the event delivery successfully then it
386 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
387 * if the task switch emulation is triggered via a task gate then it should
388 * call this function with 'intinfo=0' to indicate that the external event
389 * is not pending anymore.
390 *
391 * Return value is 0 on success and non-zero on failure.
392 */
393 int vm_exit_intinfo(struct vcpu *vcpu, uint64_t intinfo);
394
395 /*
396 * This function is called before every VM-entry to retrieve a pending
397 * event that should be injected into the guest. This function combines
398 * nested events into a double or triple fault.
399 *
400 * Returns 0 if there are no events that need to be injected into the guest
401 * and non-zero otherwise.
402 */
403 int vm_entry_intinfo(struct vcpu *vcpu, uint64_t *info);
404
405 int vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2);
406
407 /*
408 * Function used to keep track of the guest's TSC offset. The
409 * offset is used by the virtualization extensions to provide a consistent
410 * value for the Time Stamp Counter to the guest.
411 */
412 void vm_set_tsc_offset(struct vcpu *vcpu, uint64_t offset);
413
414 enum vm_reg_name vm_segment_name(int seg_encoding);
415
416 struct vm_copyinfo {
417 uint64_t gpa;
418 size_t len;
419 void *hva;
420 void *cookie;
421 };
422
423 /*
424 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
425 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
426 * a copyin or PROT_WRITE for a copyout.
427 *
428 * retval is_fault Interpretation
429 * 0 0 Success
430 * 0 1 An exception was injected into the guest
431 * EFAULT N/A Unrecoverable error
432 *
433 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
434 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
435 * 'vm_copy_teardown()' after the copy is done.
436 */
437 int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging,
438 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
439 int num_copyinfo, int *is_fault);
440 void vm_copy_teardown(struct vm_copyinfo *copyinfo, int num_copyinfo);
441 void vm_copyin(struct vm_copyinfo *copyinfo, void *kaddr, size_t len);
442 void vm_copyout(const void *kaddr, struct vm_copyinfo *copyinfo, size_t len);
443
444 int vcpu_trace_exceptions(struct vcpu *vcpu);
445 int vcpu_trap_wbinvd(struct vcpu *vcpu);
446 #endif /* KERNEL */
447
448 /*
449 * Identifiers for optional vmm capabilities
450 */
451 enum vm_cap_type {
452 VM_CAP_HALT_EXIT,
453 VM_CAP_MTRAP_EXIT,
454 VM_CAP_PAUSE_EXIT,
455 VM_CAP_UNRESTRICTED_GUEST,
456 VM_CAP_ENABLE_INVPCID,
457 VM_CAP_BPT_EXIT,
458 VM_CAP_RDPID,
459 VM_CAP_RDTSCP,
460 VM_CAP_IPI_EXIT,
461 VM_CAP_MASK_HWINTR,
462 VM_CAP_RFLAGS_TF,
463 VM_CAP_MAX
464 };
465
466 enum vm_intr_trigger {
467 EDGE_TRIGGER,
468 LEVEL_TRIGGER
469 };
470
471 /*
472 * The 'access' field has the format specified in Table 21-2 of the Intel
473 * Architecture Manual vol 3b.
474 *
475 * XXX The contents of the 'access' field are architecturally defined except
476 * bit 16 - Segment Unusable.
477 */
478 struct seg_desc {
479 uint64_t base;
480 uint32_t limit;
481 uint32_t access;
482 };
483 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
484 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
485 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
486 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
487 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
488 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
489
490 enum vm_cpu_mode {
491 CPU_MODE_REAL,
492 CPU_MODE_PROTECTED,
493 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
494 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
495 };
496
497 enum vm_paging_mode {
498 PAGING_MODE_FLAT,
499 PAGING_MODE_32,
500 PAGING_MODE_PAE,
501 PAGING_MODE_64,
502 PAGING_MODE_64_LA57,
503 };
504
505 struct vm_guest_paging {
506 uint64_t cr3;
507 int cpl;
508 enum vm_cpu_mode cpu_mode;
509 enum vm_paging_mode paging_mode;
510 };
511
512 /*
513 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
514 * consumers of instruction decoding. The only reason why their contents
515 * need to be exposed is because they are part of the 'vm_exit' structure.
516 */
517 struct vie_op {
518 uint8_t op_byte; /* actual opcode byte */
519 uint8_t op_type; /* type of operation (e.g. MOV) */
520 uint16_t op_flags;
521 };
522 _Static_assert(sizeof(struct vie_op) == 4, "ABI");
523 _Static_assert(_Alignof(struct vie_op) == 2, "ABI");
524
525 #define VIE_INST_SIZE 15
526 struct vie {
527 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
528 uint8_t num_valid; /* size of the instruction */
529
530 /* The following fields are all zeroed upon restart. */
531 #define vie_startzero num_processed
532 uint8_t num_processed;
533
534 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
535 uint8_t rex_w:1, /* REX prefix */
536 rex_r:1,
537 rex_x:1,
538 rex_b:1,
539 rex_present:1,
540 repz_present:1, /* REP/REPE/REPZ prefix */
541 repnz_present:1, /* REPNE/REPNZ prefix */
542 opsize_override:1, /* Operand size override */
543 addrsize_override:1, /* Address size override */
544 segment_override:1; /* Segment override */
545
546 uint8_t mod:2, /* ModRM byte */
547 reg:4,
548 rm:4;
549
550 uint8_t ss:2, /* SIB byte */
551 vex_present:1, /* VEX prefixed */
552 vex_l:1, /* L bit */
553 index:4, /* SIB byte */
554 base:4; /* SIB byte */
555
556 uint8_t disp_bytes;
557 uint8_t imm_bytes;
558
559 uint8_t scale;
560
561 uint8_t vex_reg:4, /* vvvv: first source register specifier */
562 vex_pp:2, /* pp */
563 _sparebits:2;
564
565 uint8_t _sparebytes[2];
566
567 int base_register; /* VM_REG_GUEST_xyz */
568 int index_register; /* VM_REG_GUEST_xyz */
569 int segment_register; /* VM_REG_GUEST_xyz */
570
571 int64_t displacement; /* optional addr displacement */
572 int64_t immediate; /* optional immediate operand */
573
574 uint8_t decoded; /* set to 1 if successfully decoded */
575
576 uint8_t _sparebyte;
577
578 struct vie_op op; /* opcode description */
579 };
580 _Static_assert(sizeof(struct vie) == 64, "ABI");
581 _Static_assert(__offsetof(struct vie, disp_bytes) == 22, "ABI");
582 _Static_assert(__offsetof(struct vie, scale) == 24, "ABI");
583 _Static_assert(__offsetof(struct vie, base_register) == 28, "ABI");
584
585 enum vm_exitcode {
586 VM_EXITCODE_INOUT,
587 VM_EXITCODE_VMX,
588 VM_EXITCODE_BOGUS,
589 VM_EXITCODE_RDMSR,
590 VM_EXITCODE_WRMSR,
591 VM_EXITCODE_HLT,
592 VM_EXITCODE_MTRAP,
593 VM_EXITCODE_PAUSE,
594 VM_EXITCODE_PAGING,
595 VM_EXITCODE_INST_EMUL,
596 VM_EXITCODE_SPINUP_AP,
597 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
598 VM_EXITCODE_RENDEZVOUS,
599 VM_EXITCODE_IOAPIC_EOI,
600 VM_EXITCODE_SUSPENDED,
601 VM_EXITCODE_INOUT_STR,
602 VM_EXITCODE_TASK_SWITCH,
603 VM_EXITCODE_MONITOR,
604 VM_EXITCODE_MWAIT,
605 VM_EXITCODE_SVM,
606 VM_EXITCODE_REQIDLE,
607 VM_EXITCODE_DEBUG,
608 VM_EXITCODE_VMINSN,
609 VM_EXITCODE_BPT,
610 VM_EXITCODE_IPI,
611 VM_EXITCODE_DB,
612 VM_EXITCODE_MAX
613 };
614
615 struct vm_inout {
616 uint16_t bytes:3; /* 1 or 2 or 4 */
617 uint16_t in:1;
618 uint16_t string:1;
619 uint16_t rep:1;
620 uint16_t port;
621 uint32_t eax; /* valid for out */
622 };
623
624 struct vm_inout_str {
625 struct vm_inout inout; /* must be the first element */
626 struct vm_guest_paging paging;
627 uint64_t rflags;
628 uint64_t cr0;
629 uint64_t index;
630 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
631 int addrsize;
632 enum vm_reg_name seg_name;
633 struct seg_desc seg_desc;
634 int cs_d;
635 uint64_t cs_base;
636 };
637
638 enum task_switch_reason {
639 TSR_CALL,
640 TSR_IRET,
641 TSR_JMP,
642 TSR_IDT_GATE, /* task gate in IDT */
643 };
644
645 struct vm_task_switch {
646 uint16_t tsssel; /* new TSS selector */
647 int ext; /* task switch due to external event */
648 uint32_t errcode;
649 int errcode_valid; /* push 'errcode' on the new stack */
650 enum task_switch_reason reason;
651 struct vm_guest_paging paging;
652 };
653
654 struct vm_exit {
655 enum vm_exitcode exitcode;
656 int inst_length; /* 0 means unknown */
657 uint64_t rip;
658 union {
659 struct vm_inout inout;
660 struct vm_inout_str inout_str;
661 struct {
662 uint64_t gpa;
663 int fault_type;
664 } paging;
665 struct {
666 uint64_t gpa;
667 uint64_t gla;
668 uint64_t cs_base;
669 int cs_d; /* CS.D */
670 struct vm_guest_paging paging;
671 struct vie vie;
672 } inst_emul;
673 /*
674 * VMX specific payload. Used when there is no "better"
675 * exitcode to represent the VM-exit.
676 */
677 struct {
678 int status; /* vmx inst status */
679 /*
680 * 'exit_reason' and 'exit_qualification' are valid
681 * only if 'status' is zero.
682 */
683 uint32_t exit_reason;
684 uint64_t exit_qualification;
685 /*
686 * 'inst_error' and 'inst_type' are valid
687 * only if 'status' is non-zero.
688 */
689 int inst_type;
690 int inst_error;
691 } vmx;
692 /*
693 * SVM specific payload.
694 */
695 struct {
696 uint64_t exitcode;
697 uint64_t exitinfo1;
698 uint64_t exitinfo2;
699 } svm;
700 struct {
701 int inst_length;
702 } bpt;
703 struct {
704 int trace_trap;
705 int pushf_intercept;
706 int tf_shadow_val;
707 struct vm_guest_paging paging;
708 } dbg;
709 struct {
710 uint32_t code; /* ecx value */
711 uint64_t wval;
712 } msr;
713 struct {
714 int vcpu;
715 uint64_t rip;
716 } spinup_ap;
717 struct {
718 uint64_t rflags;
719 uint64_t intr_status;
720 } hlt;
721 struct {
722 int vector;
723 } ioapic_eoi;
724 struct {
725 enum vm_suspend_how how;
726 } suspended;
727 struct {
728 /*
729 * The destination vCPU mask is saved in vcpu->cpuset
730 * and is copied out to userspace separately to avoid
731 * ABI concerns.
732 */
733 uint32_t mode;
734 uint8_t vector;
735 } ipi;
736 struct vm_task_switch task_switch;
737 } u;
738 };
739
740 /* APIs to inject faults into the guest */
741 void vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid,
742 int errcode);
743
744 static __inline void
vm_inject_ud(struct vcpu * vcpu)745 vm_inject_ud(struct vcpu *vcpu)
746 {
747 vm_inject_fault(vcpu, IDT_UD, 0, 0);
748 }
749
750 static __inline void
vm_inject_gp(struct vcpu * vcpu)751 vm_inject_gp(struct vcpu *vcpu)
752 {
753 vm_inject_fault(vcpu, IDT_GP, 1, 0);
754 }
755
756 static __inline void
vm_inject_ac(struct vcpu * vcpu,int errcode)757 vm_inject_ac(struct vcpu *vcpu, int errcode)
758 {
759 vm_inject_fault(vcpu, IDT_AC, 1, errcode);
760 }
761
762 static __inline void
vm_inject_ss(struct vcpu * vcpu,int errcode)763 vm_inject_ss(struct vcpu *vcpu, int errcode)
764 {
765 vm_inject_fault(vcpu, IDT_SS, 1, errcode);
766 }
767
768 void vm_inject_pf(struct vcpu *vcpu, int error_code, uint64_t cr2);
769
770 #endif /* _VMM_H_ */
771