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_LAST
50 };
51
52 /*
53 * Identifiers for architecturally defined registers.
54 */
55 enum vm_reg_name {
56 VM_REG_GUEST_RAX,
57 VM_REG_GUEST_RBX,
58 VM_REG_GUEST_RCX,
59 VM_REG_GUEST_RDX,
60 VM_REG_GUEST_RSI,
61 VM_REG_GUEST_RDI,
62 VM_REG_GUEST_RBP,
63 VM_REG_GUEST_R8,
64 VM_REG_GUEST_R9,
65 VM_REG_GUEST_R10,
66 VM_REG_GUEST_R11,
67 VM_REG_GUEST_R12,
68 VM_REG_GUEST_R13,
69 VM_REG_GUEST_R14,
70 VM_REG_GUEST_R15,
71 VM_REG_GUEST_CR0,
72 VM_REG_GUEST_CR3,
73 VM_REG_GUEST_CR4,
74 VM_REG_GUEST_DR7,
75 VM_REG_GUEST_RSP,
76 VM_REG_GUEST_RIP,
77 VM_REG_GUEST_RFLAGS,
78 VM_REG_GUEST_ES,
79 VM_REG_GUEST_CS,
80 VM_REG_GUEST_SS,
81 VM_REG_GUEST_DS,
82 VM_REG_GUEST_FS,
83 VM_REG_GUEST_GS,
84 VM_REG_GUEST_LDTR,
85 VM_REG_GUEST_TR,
86 VM_REG_GUEST_IDTR,
87 VM_REG_GUEST_GDTR,
88 VM_REG_GUEST_EFER,
89 VM_REG_GUEST_CR2,
90 VM_REG_GUEST_PDPTE0,
91 VM_REG_GUEST_PDPTE1,
92 VM_REG_GUEST_PDPTE2,
93 VM_REG_GUEST_PDPTE3,
94 VM_REG_GUEST_INTR_SHADOW,
95 VM_REG_GUEST_DR0,
96 VM_REG_GUEST_DR1,
97 VM_REG_GUEST_DR2,
98 VM_REG_GUEST_DR3,
99 VM_REG_GUEST_DR6,
100 VM_REG_GUEST_ENTRY_INST_LENGTH,
101 VM_REG_GUEST_FS_BASE,
102 VM_REG_GUEST_GS_BASE,
103 VM_REG_GUEST_KGS_BASE,
104 VM_REG_GUEST_TPR,
105 VM_REG_LAST
106 };
107
108 enum x2apic_state {
109 X2APIC_DISABLED,
110 X2APIC_ENABLED,
111 X2APIC_STATE_LAST
112 };
113
114 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
115 #define VM_INTINFO_DEL_ERRCODE 0x800
116 #define VM_INTINFO_RSVD 0x7ffff000
117 #define VM_INTINFO_VALID 0x80000000
118 #define VM_INTINFO_TYPE 0x700
119 #define VM_INTINFO_HWINTR (0 << 8)
120 #define VM_INTINFO_NMI (2 << 8)
121 #define VM_INTINFO_HWEXCEPTION (3 << 8)
122 #define VM_INTINFO_SWINTR (4 << 8)
123
124 /*
125 * The VM name has to fit into the pathname length constraints of devfs,
126 * governed primarily by SPECNAMELEN. The length is the total number of
127 * characters in the full path, relative to the mount point and not
128 * including any leading '/' characters.
129 * A prefix and a suffix are added to the name specified by the user.
130 * The prefix is usually "vmm/" or "vmm.io/", but can be a few characters
131 * longer for future use.
132 * The suffix is a string that identifies a bootrom image or some similar
133 * image that is attached to the VM. A separator character gets added to
134 * the suffix automatically when generating the full path, so it must be
135 * accounted for, reducing the effective length by 1.
136 * The effective length of a VM name is 229 bytes for FreeBSD 13 and 37
137 * bytes for FreeBSD 12. A minimum length is set for safety and supports
138 * a SPECNAMELEN as small as 32 on old systems.
139 */
140 #define VM_MAX_PREFIXLEN 10
141 #define VM_MAX_SUFFIXLEN 15
142 #define VM_MIN_NAMELEN 6
143 #define VM_MAX_NAMELEN \
144 (SPECNAMELEN - VM_MAX_PREFIXLEN - VM_MAX_SUFFIXLEN - 1)
145
146 #ifdef _KERNEL
147 #include <sys/kassert.h>
148
149 CTASSERT(VM_MAX_NAMELEN >= VM_MIN_NAMELEN);
150
151 struct vm;
152 struct vm_exception;
153 struct vm_mem;
154 struct seg_desc;
155 struct vm_exit;
156 struct vm_run;
157 struct vhpet;
158 struct vioapic;
159 struct vlapic;
160 struct vmspace;
161 struct vm_object;
162 struct vm_guest_paging;
163 struct pmap;
164 enum snapshot_req;
165
166 struct vm_eventinfo {
167 cpuset_t *rptr; /* rendezvous cookie */
168 int *sptr; /* suspend cookie */
169 int *iptr; /* reqidle cookie */
170 };
171
172 typedef int (*vmm_init_func_t)(int ipinum);
173 typedef int (*vmm_cleanup_func_t)(void);
174 typedef void (*vmm_suspend_func_t)(void);
175 typedef void (*vmm_resume_func_t)(void);
176 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
177 typedef int (*vmi_run_func_t)(void *vcpui, register_t rip,
178 struct pmap *pmap, struct vm_eventinfo *info);
179 typedef void (*vmi_cleanup_func_t)(void *vmi);
180 typedef void * (*vmi_vcpu_init_func_t)(void *vmi, struct vcpu *vcpu,
181 int vcpu_id);
182 typedef void (*vmi_vcpu_cleanup_func_t)(void *vcpui);
183 typedef int (*vmi_get_register_t)(void *vcpui, int num, uint64_t *retval);
184 typedef int (*vmi_set_register_t)(void *vcpui, int num, uint64_t val);
185 typedef int (*vmi_get_desc_t)(void *vcpui, int num, struct seg_desc *desc);
186 typedef int (*vmi_set_desc_t)(void *vcpui, int num, struct seg_desc *desc);
187 typedef int (*vmi_get_cap_t)(void *vcpui, int num, int *retval);
188 typedef int (*vmi_set_cap_t)(void *vcpui, int num, int val);
189 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
190 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
191 typedef struct vlapic * (*vmi_vlapic_init)(void *vcpui);
192 typedef void (*vmi_vlapic_cleanup)(struct vlapic *vlapic);
193 typedef int (*vmi_snapshot_vcpu_t)(void *vcpui, struct vm_snapshot_meta *meta);
194 typedef int (*vmi_restore_tsc_t)(void *vcpui, uint64_t now);
195
196 struct vmm_ops {
197 vmm_init_func_t modinit; /* module wide initialization */
198 vmm_cleanup_func_t modcleanup;
199 vmm_resume_func_t modsuspend;
200 vmm_resume_func_t modresume;
201
202 vmi_init_func_t init; /* vm-specific initialization */
203 vmi_run_func_t run;
204 vmi_cleanup_func_t cleanup;
205 vmi_vcpu_init_func_t vcpu_init;
206 vmi_vcpu_cleanup_func_t vcpu_cleanup;
207 vmi_get_register_t getreg;
208 vmi_set_register_t setreg;
209 vmi_get_desc_t getdesc;
210 vmi_set_desc_t setdesc;
211 vmi_get_cap_t getcap;
212 vmi_set_cap_t setcap;
213 vmi_vmspace_alloc vmspace_alloc;
214 vmi_vmspace_free vmspace_free;
215 vmi_vlapic_init vlapic_init;
216 vmi_vlapic_cleanup vlapic_cleanup;
217
218 /* checkpoint operations */
219 vmi_snapshot_vcpu_t vcpu_snapshot;
220 vmi_restore_tsc_t restore_tsc;
221 };
222
223 extern const struct vmm_ops vmm_ops_intel;
224 extern const struct vmm_ops vmm_ops_amd;
225
226 extern u_int vm_maxcpu; /* maximum virtual cpus */
227
228 int vm_create(const char *name, struct vm **retvm);
229 struct vcpu *vm_alloc_vcpu(struct vm *vm, int vcpuid);
230 void vm_disable_vcpu_creation(struct vm *vm);
231 void vm_slock_vcpus(struct vm *vm);
232 void vm_unlock_vcpus(struct vm *vm);
233 void vm_destroy(struct vm *vm);
234 int vm_reinit(struct vm *vm);
235 const char *vm_name(struct vm *vm);
236 uint16_t vm_get_maxcpus(struct vm *vm);
237 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
238 uint16_t *threads, uint16_t *maxcpus);
239 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
240 uint16_t threads, uint16_t maxcpus);
241
242 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
243 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
244 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
245 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
246
247 int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval);
248 int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val);
249 int vm_get_seg_desc(struct vcpu *vcpu, int reg,
250 struct seg_desc *ret_desc);
251 int vm_set_seg_desc(struct vcpu *vcpu, int reg,
252 struct seg_desc *desc);
253 int vm_run(struct vcpu *vcpu);
254 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
255 int vm_inject_nmi(struct vcpu *vcpu);
256 int vm_nmi_pending(struct vcpu *vcpu);
257 void vm_nmi_clear(struct vcpu *vcpu);
258 int vm_inject_extint(struct vcpu *vcpu);
259 int vm_extint_pending(struct vcpu *vcpu);
260 void vm_extint_clear(struct vcpu *vcpu);
261 int vcpu_vcpuid(struct vcpu *vcpu);
262 struct vm *vcpu_vm(struct vcpu *vcpu);
263 struct vcpu *vm_vcpu(struct vm *vm, int cpu);
264 struct vlapic *vm_lapic(struct vcpu *vcpu);
265 struct vioapic *vm_ioapic(struct vm *vm);
266 struct vhpet *vm_hpet(struct vm *vm);
267 int vm_get_capability(struct vcpu *vcpu, int type, int *val);
268 int vm_set_capability(struct vcpu *vcpu, int type, int val);
269 int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state);
270 int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state);
271 int vm_apicid2vcpuid(struct vm *vm, int apicid);
272 int vm_activate_cpu(struct vcpu *vcpu);
273 int vm_suspend_cpu(struct vm *vm, struct vcpu *vcpu);
274 int vm_resume_cpu(struct vm *vm, struct vcpu *vcpu);
275 int vm_restart_instruction(struct vcpu *vcpu);
276 struct vm_exit *vm_exitinfo(struct vcpu *vcpu);
277 cpuset_t *vm_exitinfo_cpuset(struct vcpu *vcpu);
278 void vm_exit_suspended(struct vcpu *vcpu, uint64_t rip);
279 void vm_exit_debug(struct vcpu *vcpu, uint64_t rip);
280 void vm_exit_rendezvous(struct vcpu *vcpu, uint64_t rip);
281 void vm_exit_astpending(struct vcpu *vcpu, uint64_t rip);
282 void vm_exit_reqidle(struct vcpu *vcpu, uint64_t rip);
283 int vm_snapshot_req(struct vm *vm, struct vm_snapshot_meta *meta);
284 int vm_restore_time(struct vm *vm);
285
286 #ifdef _SYS__CPUSET_H_
287 /*
288 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
289 * The rendezvous 'func(arg)' is not allowed to do anything that will
290 * cause the thread to be put to sleep.
291 *
292 * The caller cannot hold any locks when initiating the rendezvous.
293 *
294 * The implementation of this API may cause vcpus other than those specified
295 * by 'dest' to be stalled. The caller should not rely on any vcpus making
296 * forward progress when the rendezvous is in progress.
297 */
298 typedef void (*vm_rendezvous_func_t)(struct vcpu *vcpu, void *arg);
299 int vm_smp_rendezvous(struct vcpu *vcpu, cpuset_t dest,
300 vm_rendezvous_func_t func, void *arg);
301
302 cpuset_t vm_active_cpus(struct vm *vm);
303 cpuset_t vm_debug_cpus(struct vm *vm);
304 cpuset_t vm_suspended_cpus(struct vm *vm);
305 cpuset_t vm_start_cpus(struct vm *vm, const cpuset_t *tostart);
306 void vm_await_start(struct vm *vm, const cpuset_t *waiting);
307 #endif /* _SYS__CPUSET_H_ */
308
309 static __inline int
vcpu_rendezvous_pending(struct vcpu * vcpu,struct vm_eventinfo * info)310 vcpu_rendezvous_pending(struct vcpu *vcpu, struct vm_eventinfo *info)
311 {
312 /*
313 * This check isn't done with atomic operations or under a lock because
314 * there's no need to. If the vcpuid bit is set, the vcpu is part of a
315 * rendezvous and the bit won't be cleared until the vcpu enters the
316 * rendezvous. On rendezvous exit, the cpuset is cleared and the vcpu
317 * will see an empty cpuset. So, the races are harmless.
318 */
319 return (CPU_ISSET(vcpu_vcpuid(vcpu), info->rptr));
320 }
321
322 static __inline int
vcpu_suspended(struct vm_eventinfo * info)323 vcpu_suspended(struct vm_eventinfo *info)
324 {
325
326 return (*info->sptr);
327 }
328
329 static __inline int
vcpu_reqidle(struct vm_eventinfo * info)330 vcpu_reqidle(struct vm_eventinfo *info)
331 {
332
333 return (*info->iptr);
334 }
335
336 int vcpu_debugged(struct vcpu *vcpu);
337
338 /*
339 * Return true if device indicated by bus/slot/func is supposed to be a
340 * pci passthrough device.
341 *
342 * Return false otherwise.
343 */
344 bool vmm_is_pptdev(int bus, int slot, int func);
345
346 void *vm_iommu_domain(struct vm *vm);
347
348 enum vcpu_state {
349 VCPU_IDLE,
350 VCPU_FROZEN,
351 VCPU_RUNNING,
352 VCPU_SLEEPING,
353 };
354
355 int vcpu_set_state(struct vcpu *vcpu, enum vcpu_state state, bool from_idle);
356 enum vcpu_state vcpu_get_state(struct vcpu *vcpu, int *hostcpu);
357
358 static int __inline
vcpu_is_running(struct vcpu * vcpu,int * hostcpu)359 vcpu_is_running(struct vcpu *vcpu, int *hostcpu)
360 {
361 return (vcpu_get_state(vcpu, hostcpu) == VCPU_RUNNING);
362 }
363
364 #ifdef _SYS_PROC_H_
365 static int __inline
vcpu_should_yield(struct vcpu * vcpu)366 vcpu_should_yield(struct vcpu *vcpu)
367 {
368 struct thread *td;
369
370 td = curthread;
371 return (td->td_ast != 0 || td->td_owepreempt != 0);
372 }
373 #endif
374
375 void *vcpu_stats(struct vcpu *vcpu);
376 void vcpu_notify_event(struct vcpu *vcpu, bool lapic_intr);
377 struct vmspace *vm_vmspace(struct vm *vm);
378 struct vm_mem *vm_mem(struct vm *vm);
379 struct vatpic *vm_atpic(struct vm *vm);
380 struct vatpit *vm_atpit(struct vm *vm);
381 struct vpmtmr *vm_pmtmr(struct vm *vm);
382 struct vrtc *vm_rtc(struct vm *vm);
383
384 /*
385 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
386 * success and non-zero on failure.
387 *
388 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
389 * this function directly because they enforce the trap-like or fault-like
390 * behavior of an exception.
391 *
392 * This function should only be called in the context of the thread that is
393 * executing this vcpu.
394 */
395 int vm_inject_exception(struct vcpu *vcpu, int vector, int err_valid,
396 uint32_t errcode, int restart_instruction);
397
398 /*
399 * This function is called after a VM-exit that occurred during exception or
400 * interrupt delivery through the IDT. The format of 'intinfo' is described
401 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
402 *
403 * If a VM-exit handler completes the event delivery successfully then it
404 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
405 * if the task switch emulation is triggered via a task gate then it should
406 * call this function with 'intinfo=0' to indicate that the external event
407 * is not pending anymore.
408 *
409 * Return value is 0 on success and non-zero on failure.
410 */
411 int vm_exit_intinfo(struct vcpu *vcpu, uint64_t intinfo);
412
413 /*
414 * This function is called before every VM-entry to retrieve a pending
415 * event that should be injected into the guest. This function combines
416 * nested events into a double or triple fault.
417 *
418 * Returns 0 if there are no events that need to be injected into the guest
419 * and non-zero otherwise.
420 */
421 int vm_entry_intinfo(struct vcpu *vcpu, uint64_t *info);
422
423 int vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2);
424
425 /*
426 * Function used to keep track of the guest's TSC offset. The
427 * offset is used by the virtualization extensions to provide a consistent
428 * value for the Time Stamp Counter to the guest.
429 */
430 void vm_set_tsc_offset(struct vcpu *vcpu, uint64_t offset);
431
432 enum vm_reg_name vm_segment_name(int seg_encoding);
433
434 struct vm_copyinfo {
435 uint64_t gpa;
436 size_t len;
437 void *hva;
438 void *cookie;
439 };
440
441 /*
442 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
443 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
444 * a copyin or PROT_WRITE for a copyout.
445 *
446 * retval is_fault Interpretation
447 * 0 0 Success
448 * 0 1 An exception was injected into the guest
449 * EFAULT N/A Unrecoverable error
450 *
451 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
452 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
453 * 'vm_copy_teardown()' after the copy is done.
454 */
455 int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging,
456 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
457 int num_copyinfo, int *is_fault);
458 void vm_copy_teardown(struct vm_copyinfo *copyinfo, int num_copyinfo);
459 void vm_copyin(struct vm_copyinfo *copyinfo, void *kaddr, size_t len);
460 void vm_copyout(const void *kaddr, struct vm_copyinfo *copyinfo, size_t len);
461
462 int vcpu_trace_exceptions(struct vcpu *vcpu);
463 int vcpu_trap_wbinvd(struct vcpu *vcpu);
464 #endif /* KERNEL */
465
466 /*
467 * Identifiers for optional vmm capabilities
468 */
469 enum vm_cap_type {
470 VM_CAP_HALT_EXIT,
471 VM_CAP_MTRAP_EXIT,
472 VM_CAP_PAUSE_EXIT,
473 VM_CAP_UNRESTRICTED_GUEST,
474 VM_CAP_ENABLE_INVPCID,
475 VM_CAP_BPT_EXIT,
476 VM_CAP_RDPID,
477 VM_CAP_RDTSCP,
478 VM_CAP_IPI_EXIT,
479 VM_CAP_MASK_HWINTR,
480 VM_CAP_RFLAGS_TF,
481 VM_CAP_MAX
482 };
483
484 enum vm_intr_trigger {
485 EDGE_TRIGGER,
486 LEVEL_TRIGGER
487 };
488
489 /*
490 * The 'access' field has the format specified in Table 21-2 of the Intel
491 * Architecture Manual vol 3b.
492 *
493 * XXX The contents of the 'access' field are architecturally defined except
494 * bit 16 - Segment Unusable.
495 */
496 struct seg_desc {
497 uint64_t base;
498 uint32_t limit;
499 uint32_t access;
500 };
501 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
502 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
503 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
504 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
505 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
506 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
507
508 enum vm_cpu_mode {
509 CPU_MODE_REAL,
510 CPU_MODE_PROTECTED,
511 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
512 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
513 };
514
515 enum vm_paging_mode {
516 PAGING_MODE_FLAT,
517 PAGING_MODE_32,
518 PAGING_MODE_PAE,
519 PAGING_MODE_64,
520 PAGING_MODE_64_LA57,
521 };
522
523 struct vm_guest_paging {
524 uint64_t cr3;
525 int cpl;
526 enum vm_cpu_mode cpu_mode;
527 enum vm_paging_mode paging_mode;
528 };
529
530 /*
531 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
532 * consumers of instruction decoding. The only reason why their contents
533 * need to be exposed is because they are part of the 'vm_exit' structure.
534 */
535 struct vie_op {
536 uint8_t op_byte; /* actual opcode byte */
537 uint8_t op_type; /* type of operation (e.g. MOV) */
538 uint16_t op_flags;
539 };
540 _Static_assert(sizeof(struct vie_op) == 4, "ABI");
541 _Static_assert(_Alignof(struct vie_op) == 2, "ABI");
542
543 #define VIE_INST_SIZE 15
544 struct vie {
545 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
546 uint8_t num_valid; /* size of the instruction */
547
548 /* The following fields are all zeroed upon restart. */
549 #define vie_startzero num_processed
550 uint8_t num_processed;
551
552 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
553 uint8_t rex_w:1, /* REX prefix */
554 rex_r:1,
555 rex_x:1,
556 rex_b:1,
557 rex_present:1,
558 repz_present:1, /* REP/REPE/REPZ prefix */
559 repnz_present:1, /* REPNE/REPNZ prefix */
560 opsize_override:1, /* Operand size override */
561 addrsize_override:1, /* Address size override */
562 segment_override:1; /* Segment override */
563
564 uint8_t mod:2, /* ModRM byte */
565 reg:4,
566 rm:4;
567
568 uint8_t ss:2, /* SIB byte */
569 vex_present:1, /* VEX prefixed */
570 vex_l:1, /* L bit */
571 index:4, /* SIB byte */
572 base:4; /* SIB byte */
573
574 uint8_t disp_bytes;
575 uint8_t imm_bytes;
576
577 uint8_t scale;
578
579 uint8_t vex_reg:4, /* vvvv: first source register specifier */
580 vex_pp:2, /* pp */
581 _sparebits:2;
582
583 uint8_t _sparebytes[2];
584
585 int base_register; /* VM_REG_GUEST_xyz */
586 int index_register; /* VM_REG_GUEST_xyz */
587 int segment_register; /* VM_REG_GUEST_xyz */
588
589 int64_t displacement; /* optional addr displacement */
590 int64_t immediate; /* optional immediate operand */
591
592 uint8_t decoded; /* set to 1 if successfully decoded */
593
594 uint8_t _sparebyte;
595
596 struct vie_op op; /* opcode description */
597 };
598 _Static_assert(sizeof(struct vie) == 64, "ABI");
599 _Static_assert(__offsetof(struct vie, disp_bytes) == 22, "ABI");
600 _Static_assert(__offsetof(struct vie, scale) == 24, "ABI");
601 _Static_assert(__offsetof(struct vie, base_register) == 28, "ABI");
602
603 enum vm_exitcode {
604 VM_EXITCODE_INOUT,
605 VM_EXITCODE_VMX,
606 VM_EXITCODE_BOGUS,
607 VM_EXITCODE_RDMSR,
608 VM_EXITCODE_WRMSR,
609 VM_EXITCODE_HLT,
610 VM_EXITCODE_MTRAP,
611 VM_EXITCODE_PAUSE,
612 VM_EXITCODE_PAGING,
613 VM_EXITCODE_INST_EMUL,
614 VM_EXITCODE_SPINUP_AP,
615 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
616 VM_EXITCODE_RENDEZVOUS,
617 VM_EXITCODE_IOAPIC_EOI,
618 VM_EXITCODE_SUSPENDED,
619 VM_EXITCODE_INOUT_STR,
620 VM_EXITCODE_TASK_SWITCH,
621 VM_EXITCODE_MONITOR,
622 VM_EXITCODE_MWAIT,
623 VM_EXITCODE_SVM,
624 VM_EXITCODE_REQIDLE,
625 VM_EXITCODE_DEBUG,
626 VM_EXITCODE_VMINSN,
627 VM_EXITCODE_BPT,
628 VM_EXITCODE_IPI,
629 VM_EXITCODE_DB,
630 VM_EXITCODE_MAX
631 };
632
633 struct vm_inout {
634 uint16_t bytes:3; /* 1 or 2 or 4 */
635 uint16_t in:1;
636 uint16_t string:1;
637 uint16_t rep:1;
638 uint16_t port;
639 uint32_t eax; /* valid for out */
640 };
641
642 struct vm_inout_str {
643 struct vm_inout inout; /* must be the first element */
644 struct vm_guest_paging paging;
645 uint64_t rflags;
646 uint64_t cr0;
647 uint64_t index;
648 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
649 int addrsize;
650 enum vm_reg_name seg_name;
651 struct seg_desc seg_desc;
652 };
653
654 enum task_switch_reason {
655 TSR_CALL,
656 TSR_IRET,
657 TSR_JMP,
658 TSR_IDT_GATE, /* task gate in IDT */
659 };
660
661 struct vm_task_switch {
662 uint16_t tsssel; /* new TSS selector */
663 int ext; /* task switch due to external event */
664 uint32_t errcode;
665 int errcode_valid; /* push 'errcode' on the new stack */
666 enum task_switch_reason reason;
667 struct vm_guest_paging paging;
668 };
669
670 struct vm_exit {
671 enum vm_exitcode exitcode;
672 int inst_length; /* 0 means unknown */
673 uint64_t rip;
674 union {
675 struct vm_inout inout;
676 struct vm_inout_str inout_str;
677 struct {
678 uint64_t gpa;
679 int fault_type;
680 } paging;
681 struct {
682 uint64_t gpa;
683 uint64_t gla;
684 uint64_t cs_base;
685 int cs_d; /* CS.D */
686 struct vm_guest_paging paging;
687 struct vie vie;
688 } inst_emul;
689 /*
690 * VMX specific payload. Used when there is no "better"
691 * exitcode to represent the VM-exit.
692 */
693 struct {
694 int status; /* vmx inst status */
695 /*
696 * 'exit_reason' and 'exit_qualification' are valid
697 * only if 'status' is zero.
698 */
699 uint32_t exit_reason;
700 uint64_t exit_qualification;
701 /*
702 * 'inst_error' and 'inst_type' are valid
703 * only if 'status' is non-zero.
704 */
705 int inst_type;
706 int inst_error;
707 } vmx;
708 /*
709 * SVM specific payload.
710 */
711 struct {
712 uint64_t exitcode;
713 uint64_t exitinfo1;
714 uint64_t exitinfo2;
715 } svm;
716 struct {
717 int inst_length;
718 } bpt;
719 struct {
720 int trace_trap;
721 int pushf_intercept;
722 int tf_shadow_val;
723 struct vm_guest_paging paging;
724 } dbg;
725 struct {
726 uint32_t code; /* ecx value */
727 uint64_t wval;
728 } msr;
729 struct {
730 int vcpu;
731 uint64_t rip;
732 } spinup_ap;
733 struct {
734 uint64_t rflags;
735 uint64_t intr_status;
736 } hlt;
737 struct {
738 int vector;
739 } ioapic_eoi;
740 struct {
741 enum vm_suspend_how how;
742 } suspended;
743 struct {
744 /*
745 * The destination vCPU mask is saved in vcpu->cpuset
746 * and is copied out to userspace separately to avoid
747 * ABI concerns.
748 */
749 uint32_t mode;
750 uint8_t vector;
751 } ipi;
752 struct vm_task_switch task_switch;
753 } u;
754 };
755
756 /* APIs to inject faults into the guest */
757 void vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid,
758 int errcode);
759
760 static __inline void
vm_inject_ud(struct vcpu * vcpu)761 vm_inject_ud(struct vcpu *vcpu)
762 {
763 vm_inject_fault(vcpu, IDT_UD, 0, 0);
764 }
765
766 static __inline void
vm_inject_gp(struct vcpu * vcpu)767 vm_inject_gp(struct vcpu *vcpu)
768 {
769 vm_inject_fault(vcpu, IDT_GP, 1, 0);
770 }
771
772 static __inline void
vm_inject_ac(struct vcpu * vcpu,int errcode)773 vm_inject_ac(struct vcpu *vcpu, int errcode)
774 {
775 vm_inject_fault(vcpu, IDT_AC, 1, errcode);
776 }
777
778 static __inline void
vm_inject_ss(struct vcpu * vcpu,int errcode)779 vm_inject_ss(struct vcpu *vcpu, int errcode)
780 {
781 vm_inject_fault(vcpu, IDT_SS, 1, errcode);
782 }
783
784 void vm_inject_pf(struct vcpu *vcpu, int error_code, uint64_t cr2);
785
786 #endif /* _VMM_H_ */
787