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/cpuset.h>
35 #include <sys/sdt.h>
36 #include <x86/segments.h>
37
38 struct vm_snapshot_meta;
39
40 #ifdef _KERNEL
41 SDT_PROVIDER_DECLARE(vmm);
42 #endif
43
44 enum vm_suspend_how {
45 VM_SUSPEND_NONE,
46 VM_SUSPEND_RESET,
47 VM_SUSPEND_POWEROFF,
48 VM_SUSPEND_HALT,
49 VM_SUSPEND_TRIPLEFAULT,
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_LAST
103 };
104
105 enum x2apic_state {
106 X2APIC_DISABLED,
107 X2APIC_ENABLED,
108 X2APIC_STATE_LAST
109 };
110
111 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
112 #define VM_INTINFO_DEL_ERRCODE 0x800
113 #define VM_INTINFO_RSVD 0x7ffff000
114 #define VM_INTINFO_VALID 0x80000000
115 #define VM_INTINFO_TYPE 0x700
116 #define VM_INTINFO_HWINTR (0 << 8)
117 #define VM_INTINFO_NMI (2 << 8)
118 #define VM_INTINFO_HWEXCEPTION (3 << 8)
119 #define VM_INTINFO_SWINTR (4 << 8)
120
121 /*
122 * The VM name has to fit into the pathname length constraints of devfs,
123 * governed primarily by SPECNAMELEN. The length is the total number of
124 * characters in the full path, relative to the mount point and not
125 * including any leading '/' characters.
126 * A prefix and a suffix are added to the name specified by the user.
127 * The prefix is usually "vmm/" or "vmm.io/", but can be a few characters
128 * longer for future use.
129 * The suffix is a string that identifies a bootrom image or some similar
130 * image that is attached to the VM. A separator character gets added to
131 * the suffix automatically when generating the full path, so it must be
132 * accounted for, reducing the effective length by 1.
133 * The effective length of a VM name is 229 bytes for FreeBSD 13 and 37
134 * bytes for FreeBSD 12. A minimum length is set for safety and supports
135 * a SPECNAMELEN as small as 32 on old systems.
136 */
137 #define VM_MAX_PREFIXLEN 10
138 #define VM_MAX_SUFFIXLEN 15
139 #define VM_MIN_NAMELEN 6
140 #define VM_MAX_NAMELEN \
141 (SPECNAMELEN - VM_MAX_PREFIXLEN - VM_MAX_SUFFIXLEN - 1)
142
143 #ifdef _KERNEL
144 CTASSERT(VM_MAX_NAMELEN >= VM_MIN_NAMELEN);
145
146 struct vcpu;
147 struct vm;
148 struct vm_exception;
149 struct seg_desc;
150 struct vm_exit;
151 struct vm_run;
152 struct vhpet;
153 struct vioapic;
154 struct vlapic;
155 struct vmspace;
156 struct vm_object;
157 struct vm_guest_paging;
158 struct pmap;
159 enum snapshot_req;
160
161 struct vm_eventinfo {
162 void *rptr; /* rendezvous cookie */
163 int *sptr; /* suspend cookie */
164 int *iptr; /* reqidle cookie */
165 };
166
167 typedef int (*vmm_init_func_t)(int ipinum);
168 typedef int (*vmm_cleanup_func_t)(void);
169 typedef void (*vmm_resume_func_t)(void);
170 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
171 typedef int (*vmi_run_func_t)(void *vcpui, register_t rip,
172 struct pmap *pmap, struct vm_eventinfo *info);
173 typedef void (*vmi_cleanup_func_t)(void *vmi);
174 typedef void * (*vmi_vcpu_init_func_t)(void *vmi, struct vcpu *vcpu,
175 int vcpu_id);
176 typedef void (*vmi_vcpu_cleanup_func_t)(void *vcpui);
177 typedef int (*vmi_get_register_t)(void *vcpui, int num, uint64_t *retval);
178 typedef int (*vmi_set_register_t)(void *vcpui, int num, uint64_t val);
179 typedef int (*vmi_get_desc_t)(void *vcpui, int num, struct seg_desc *desc);
180 typedef int (*vmi_set_desc_t)(void *vcpui, int num, struct seg_desc *desc);
181 typedef int (*vmi_get_cap_t)(void *vcpui, int num, int *retval);
182 typedef int (*vmi_set_cap_t)(void *vcpui, int num, int val);
183 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
184 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
185 typedef struct vlapic * (*vmi_vlapic_init)(void *vcpui);
186 typedef void (*vmi_vlapic_cleanup)(struct vlapic *vlapic);
187 typedef int (*vmi_snapshot_t)(void *vmi, struct vm_snapshot_meta *meta);
188 typedef int (*vmi_snapshot_vcpu_t)(void *vcpui, struct vm_snapshot_meta *meta);
189 typedef int (*vmi_restore_tsc_t)(void *vcpui, uint64_t now);
190
191 struct vmm_ops {
192 vmm_init_func_t modinit; /* module wide initialization */
193 vmm_cleanup_func_t modcleanup;
194 vmm_resume_func_t modresume;
195
196 vmi_init_func_t init; /* vm-specific initialization */
197 vmi_run_func_t run;
198 vmi_cleanup_func_t cleanup;
199 vmi_vcpu_init_func_t vcpu_init;
200 vmi_vcpu_cleanup_func_t vcpu_cleanup;
201 vmi_get_register_t getreg;
202 vmi_set_register_t setreg;
203 vmi_get_desc_t getdesc;
204 vmi_set_desc_t setdesc;
205 vmi_get_cap_t getcap;
206 vmi_set_cap_t setcap;
207 vmi_vmspace_alloc vmspace_alloc;
208 vmi_vmspace_free vmspace_free;
209 vmi_vlapic_init vlapic_init;
210 vmi_vlapic_cleanup vlapic_cleanup;
211
212 /* checkpoint operations */
213 vmi_snapshot_t snapshot;
214 vmi_snapshot_vcpu_t vcpu_snapshot;
215 vmi_restore_tsc_t restore_tsc;
216 };
217
218 extern const struct vmm_ops vmm_ops_intel;
219 extern const struct vmm_ops vmm_ops_amd;
220
221 extern u_int vm_maxcpu; /* maximum virtual cpus */
222
223 int vm_create(const char *name, struct vm **retvm);
224 struct vcpu *vm_alloc_vcpu(struct vm *vm, int vcpuid);
225 void vm_disable_vcpu_creation(struct vm *vm);
226 void vm_slock_vcpus(struct vm *vm);
227 void vm_unlock_vcpus(struct vm *vm);
228 void vm_destroy(struct vm *vm);
229 int vm_reinit(struct vm *vm);
230 const char *vm_name(struct vm *vm);
231 uint16_t vm_get_maxcpus(struct vm *vm);
232 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
233 uint16_t *threads, uint16_t *maxcpus);
234 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
235 uint16_t threads, uint16_t maxcpus);
236
237 /*
238 * APIs that modify the guest memory map require all vcpus to be frozen.
239 */
240 void vm_slock_memsegs(struct vm *vm);
241 void vm_xlock_memsegs(struct vm *vm);
242 void vm_unlock_memsegs(struct vm *vm);
243 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
244 size_t len, int prot, int flags);
245 int vm_munmap_memseg(struct vm *vm, vm_paddr_t gpa, size_t len);
246 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
247 void vm_free_memseg(struct vm *vm, int ident);
248 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
249 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
250 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
251 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
252
253 /*
254 * APIs that inspect the guest memory map require only a *single* vcpu to
255 * be frozen. This acts like a read lock on the guest memory map since any
256 * modification requires *all* vcpus to be frozen.
257 */
258 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
259 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
260 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
261 struct vm_object **objptr);
262 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
263 void *vm_gpa_hold(struct vcpu *vcpu, vm_paddr_t gpa, size_t len,
264 int prot, void **cookie);
265 void *vm_gpa_hold_global(struct vm *vm, vm_paddr_t gpa, size_t len,
266 int prot, void **cookie);
267 void *vm_gpa_hold_global(struct vm *vm, vm_paddr_t gpa, size_t len,
268 int prot, void **cookie);
269 void vm_gpa_release(void *cookie);
270 bool vm_mem_allocated(struct vcpu *vcpu, vm_paddr_t gpa);
271
272 int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval);
273 int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val);
274 int vm_get_seg_desc(struct vcpu *vcpu, int reg,
275 struct seg_desc *ret_desc);
276 int vm_set_seg_desc(struct vcpu *vcpu, int reg,
277 struct seg_desc *desc);
278 int vm_run(struct vcpu *vcpu, struct vm_exit *vme_user);
279 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
280 int vm_inject_nmi(struct vcpu *vcpu);
281 int vm_nmi_pending(struct vcpu *vcpu);
282 void vm_nmi_clear(struct vcpu *vcpu);
283 int vm_inject_extint(struct vcpu *vcpu);
284 int vm_extint_pending(struct vcpu *vcpu);
285 void vm_extint_clear(struct vcpu *vcpu);
286 int vcpu_vcpuid(struct vcpu *vcpu);
287 struct vm *vcpu_vm(struct vcpu *vcpu);
288 struct vcpu *vm_vcpu(struct vm *vm, int cpu);
289 struct vlapic *vm_lapic(struct vcpu *vcpu);
290 struct vioapic *vm_ioapic(struct vm *vm);
291 struct vhpet *vm_hpet(struct vm *vm);
292 int vm_get_capability(struct vcpu *vcpu, int type, int *val);
293 int vm_set_capability(struct vcpu *vcpu, int type, int val);
294 int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state);
295 int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state);
296 int vm_apicid2vcpuid(struct vm *vm, int apicid);
297 int vm_activate_cpu(struct vcpu *vcpu);
298 int vm_suspend_cpu(struct vm *vm, struct vcpu *vcpu);
299 int vm_resume_cpu(struct vm *vm, struct vcpu *vcpu);
300 int vm_restart_instruction(struct vcpu *vcpu);
301 struct vm_exit *vm_exitinfo(struct vcpu *vcpu);
302 void vm_exit_suspended(struct vcpu *vcpu, uint64_t rip);
303 void vm_exit_debug(struct vcpu *vcpu, uint64_t rip);
304 void vm_exit_rendezvous(struct vcpu *vcpu, uint64_t rip);
305 void vm_exit_astpending(struct vcpu *vcpu, uint64_t rip);
306 void vm_exit_reqidle(struct vcpu *vcpu, uint64_t rip);
307 int vm_snapshot_req(struct vm *vm, struct vm_snapshot_meta *meta);
308 int vm_restore_time(struct vm *vm);
309
310 #ifdef _SYS__CPUSET_H_
311 /*
312 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
313 * The rendezvous 'func(arg)' is not allowed to do anything that will
314 * cause the thread to be put to sleep.
315 *
316 * The caller cannot hold any locks when initiating the rendezvous.
317 *
318 * The implementation of this API may cause vcpus other than those specified
319 * by 'dest' to be stalled. The caller should not rely on any vcpus making
320 * forward progress when the rendezvous is in progress.
321 */
322 typedef void (*vm_rendezvous_func_t)(struct vcpu *vcpu, void *arg);
323 int vm_smp_rendezvous(struct vcpu *vcpu, cpuset_t dest,
324 vm_rendezvous_func_t func, void *arg);
325
326 cpuset_t vm_active_cpus(struct vm *vm);
327 cpuset_t vm_debug_cpus(struct vm *vm);
328 cpuset_t vm_suspended_cpus(struct vm *vm);
329 cpuset_t vm_start_cpus(struct vm *vm, const cpuset_t *tostart);
330 void vm_await_start(struct vm *vm, const cpuset_t *waiting);
331 #endif /* _SYS__CPUSET_H_ */
332
333 static __inline int
334 vcpu_rendezvous_pending(struct vm_eventinfo *info)
335 {
336
337 return (*((uintptr_t *)(info->rptr)) != 0);
338 }
339
340 static __inline int
341 vcpu_suspended(struct vm_eventinfo *info)
342 {
343
344 return (*info->sptr);
345 }
346
347 static __inline int
348 vcpu_reqidle(struct vm_eventinfo *info)
349 {
350
351 return (*info->iptr);
352 }
353
354 int vcpu_debugged(struct vcpu *vcpu);
355
356 /*
357 * Return true if device indicated by bus/slot/func is supposed to be a
358 * pci passthrough device.
359 *
360 * Return false otherwise.
361 */
362 bool vmm_is_pptdev(int bus, int slot, int func);
363
364 void *vm_iommu_domain(struct vm *vm);
365
366 enum vcpu_state {
367 VCPU_IDLE,
368 VCPU_FROZEN,
369 VCPU_RUNNING,
370 VCPU_SLEEPING,
371 };
372
373 int vcpu_set_state(struct vcpu *vcpu, enum vcpu_state state, bool from_idle);
374 enum vcpu_state vcpu_get_state(struct vcpu *vcpu, int *hostcpu);
375
376 static int __inline
377 vcpu_is_running(struct vcpu *vcpu, int *hostcpu)
378 {
379 return (vcpu_get_state(vcpu, hostcpu) == VCPU_RUNNING);
380 }
381
382 #ifdef _SYS_PROC_H_
383 static int __inline
384 vcpu_should_yield(struct vcpu *vcpu)
385 {
386 struct thread *td;
387
388 td = curthread;
389 return (td->td_ast != 0 || td->td_owepreempt != 0);
390 }
391 #endif
392
393 void *vcpu_stats(struct vcpu *vcpu);
394 void vcpu_notify_event(struct vcpu *vcpu, bool lapic_intr);
395 struct vmspace *vm_get_vmspace(struct vm *vm);
396 struct vatpic *vm_atpic(struct vm *vm);
397 struct vatpit *vm_atpit(struct vm *vm);
398 struct vpmtmr *vm_pmtmr(struct vm *vm);
399 struct vrtc *vm_rtc(struct vm *vm);
400
401 /*
402 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
403 * success and non-zero on failure.
404 *
405 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
406 * this function directly because they enforce the trap-like or fault-like
407 * behavior of an exception.
408 *
409 * This function should only be called in the context of the thread that is
410 * executing this vcpu.
411 */
412 int vm_inject_exception(struct vcpu *vcpu, int vector, int err_valid,
413 uint32_t errcode, int restart_instruction);
414
415 /*
416 * This function is called after a VM-exit that occurred during exception or
417 * interrupt delivery through the IDT. The format of 'intinfo' is described
418 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
419 *
420 * If a VM-exit handler completes the event delivery successfully then it
421 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
422 * if the task switch emulation is triggered via a task gate then it should
423 * call this function with 'intinfo=0' to indicate that the external event
424 * is not pending anymore.
425 *
426 * Return value is 0 on success and non-zero on failure.
427 */
428 int vm_exit_intinfo(struct vcpu *vcpu, uint64_t intinfo);
429
430 /*
431 * This function is called before every VM-entry to retrieve a pending
432 * event that should be injected into the guest. This function combines
433 * nested events into a double or triple fault.
434 *
435 * Returns 0 if there are no events that need to be injected into the guest
436 * and non-zero otherwise.
437 */
438 int vm_entry_intinfo(struct vcpu *vcpu, uint64_t *info);
439
440 int vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2);
441
442 /*
443 * Function used to keep track of the guest's TSC offset. The
444 * offset is used by the virutalization extensions to provide a consistent
445 * value for the Time Stamp Counter to the guest.
446 */
447 void vm_set_tsc_offset(struct vcpu *vcpu, uint64_t offset);
448
449 enum vm_reg_name vm_segment_name(int seg_encoding);
450
451 struct vm_copyinfo {
452 uint64_t gpa;
453 size_t len;
454 void *hva;
455 void *cookie;
456 };
457
458 /*
459 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
460 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
461 * a copyin or PROT_WRITE for a copyout.
462 *
463 * retval is_fault Interpretation
464 * 0 0 Success
465 * 0 1 An exception was injected into the guest
466 * EFAULT N/A Unrecoverable error
467 *
468 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
469 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
470 * 'vm_copy_teardown()' after the copy is done.
471 */
472 int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging,
473 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
474 int num_copyinfo, int *is_fault);
475 void vm_copy_teardown(struct vm_copyinfo *copyinfo, int num_copyinfo);
476 void vm_copyin(struct vm_copyinfo *copyinfo, void *kaddr, size_t len);
477 void vm_copyout(const void *kaddr, struct vm_copyinfo *copyinfo, size_t len);
478
479 int vcpu_trace_exceptions(struct vcpu *vcpu);
480 int vcpu_trap_wbinvd(struct vcpu *vcpu);
481 #endif /* KERNEL */
482
483 /*
484 * Identifiers for optional vmm capabilities
485 */
486 enum vm_cap_type {
487 VM_CAP_HALT_EXIT,
488 VM_CAP_MTRAP_EXIT,
489 VM_CAP_PAUSE_EXIT,
490 VM_CAP_UNRESTRICTED_GUEST,
491 VM_CAP_ENABLE_INVPCID,
492 VM_CAP_BPT_EXIT,
493 VM_CAP_RDPID,
494 VM_CAP_RDTSCP,
495 VM_CAP_IPI_EXIT,
496 VM_CAP_MAX
497 };
498
499 enum vm_intr_trigger {
500 EDGE_TRIGGER,
501 LEVEL_TRIGGER
502 };
503
504 /*
505 * The 'access' field has the format specified in Table 21-2 of the Intel
506 * Architecture Manual vol 3b.
507 *
508 * XXX The contents of the 'access' field are architecturally defined except
509 * bit 16 - Segment Unusable.
510 */
511 struct seg_desc {
512 uint64_t base;
513 uint32_t limit;
514 uint32_t access;
515 };
516 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
517 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
518 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
519 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
520 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
521 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
522
523 enum vm_cpu_mode {
524 CPU_MODE_REAL,
525 CPU_MODE_PROTECTED,
526 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
527 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
528 };
529
530 enum vm_paging_mode {
531 PAGING_MODE_FLAT,
532 PAGING_MODE_32,
533 PAGING_MODE_PAE,
534 PAGING_MODE_64,
535 PAGING_MODE_64_LA57,
536 };
537
538 struct vm_guest_paging {
539 uint64_t cr3;
540 int cpl;
541 enum vm_cpu_mode cpu_mode;
542 enum vm_paging_mode paging_mode;
543 };
544
545 /*
546 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
547 * consumers of instruction decoding. The only reason why their contents
548 * need to be exposed is because they are part of the 'vm_exit' structure.
549 */
550 struct vie_op {
551 uint8_t op_byte; /* actual opcode byte */
552 uint8_t op_type; /* type of operation (e.g. MOV) */
553 uint16_t op_flags;
554 };
555 _Static_assert(sizeof(struct vie_op) == 4, "ABI");
556 _Static_assert(_Alignof(struct vie_op) == 2, "ABI");
557
558 #define VIE_INST_SIZE 15
559 struct vie {
560 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
561 uint8_t num_valid; /* size of the instruction */
562
563 /* The following fields are all zeroed upon restart. */
564 #define vie_startzero num_processed
565 uint8_t num_processed;
566
567 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
568 uint8_t rex_w:1, /* REX prefix */
569 rex_r:1,
570 rex_x:1,
571 rex_b:1,
572 rex_present:1,
573 repz_present:1, /* REP/REPE/REPZ prefix */
574 repnz_present:1, /* REPNE/REPNZ prefix */
575 opsize_override:1, /* Operand size override */
576 addrsize_override:1, /* Address size override */
577 segment_override:1; /* Segment override */
578
579 uint8_t mod:2, /* ModRM byte */
580 reg:4,
581 rm:4;
582
583 uint8_t ss:2, /* SIB byte */
584 vex_present:1, /* VEX prefixed */
585 vex_l:1, /* L bit */
586 index:4, /* SIB byte */
587 base:4; /* SIB byte */
588
589 uint8_t disp_bytes;
590 uint8_t imm_bytes;
591
592 uint8_t scale;
593
594 uint8_t vex_reg:4, /* vvvv: first source register specifier */
595 vex_pp:2, /* pp */
596 _sparebits:2;
597
598 uint8_t _sparebytes[2];
599
600 int base_register; /* VM_REG_GUEST_xyz */
601 int index_register; /* VM_REG_GUEST_xyz */
602 int segment_register; /* VM_REG_GUEST_xyz */
603
604 int64_t displacement; /* optional addr displacement */
605 int64_t immediate; /* optional immediate operand */
606
607 uint8_t decoded; /* set to 1 if successfully decoded */
608
609 uint8_t _sparebyte;
610
611 struct vie_op op; /* opcode description */
612 };
613 _Static_assert(sizeof(struct vie) == 64, "ABI");
614 _Static_assert(__offsetof(struct vie, disp_bytes) == 22, "ABI");
615 _Static_assert(__offsetof(struct vie, scale) == 24, "ABI");
616 _Static_assert(__offsetof(struct vie, base_register) == 28, "ABI");
617
618 enum vm_exitcode {
619 VM_EXITCODE_INOUT,
620 VM_EXITCODE_VMX,
621 VM_EXITCODE_BOGUS,
622 VM_EXITCODE_RDMSR,
623 VM_EXITCODE_WRMSR,
624 VM_EXITCODE_HLT,
625 VM_EXITCODE_MTRAP,
626 VM_EXITCODE_PAUSE,
627 VM_EXITCODE_PAGING,
628 VM_EXITCODE_INST_EMUL,
629 VM_EXITCODE_SPINUP_AP,
630 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
631 VM_EXITCODE_RENDEZVOUS,
632 VM_EXITCODE_IOAPIC_EOI,
633 VM_EXITCODE_SUSPENDED,
634 VM_EXITCODE_INOUT_STR,
635 VM_EXITCODE_TASK_SWITCH,
636 VM_EXITCODE_MONITOR,
637 VM_EXITCODE_MWAIT,
638 VM_EXITCODE_SVM,
639 VM_EXITCODE_REQIDLE,
640 VM_EXITCODE_DEBUG,
641 VM_EXITCODE_VMINSN,
642 VM_EXITCODE_BPT,
643 VM_EXITCODE_IPI,
644 VM_EXITCODE_MAX
645 };
646
647 struct vm_inout {
648 uint16_t bytes:3; /* 1 or 2 or 4 */
649 uint16_t in:1;
650 uint16_t string:1;
651 uint16_t rep:1;
652 uint16_t port;
653 uint32_t eax; /* valid for out */
654 };
655
656 struct vm_inout_str {
657 struct vm_inout inout; /* must be the first element */
658 struct vm_guest_paging paging;
659 uint64_t rflags;
660 uint64_t cr0;
661 uint64_t index;
662 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
663 int addrsize;
664 enum vm_reg_name seg_name;
665 struct seg_desc seg_desc;
666 };
667
668 enum task_switch_reason {
669 TSR_CALL,
670 TSR_IRET,
671 TSR_JMP,
672 TSR_IDT_GATE, /* task gate in IDT */
673 };
674
675 struct vm_task_switch {
676 uint16_t tsssel; /* new TSS selector */
677 int ext; /* task switch due to external event */
678 uint32_t errcode;
679 int errcode_valid; /* push 'errcode' on the new stack */
680 enum task_switch_reason reason;
681 struct vm_guest_paging paging;
682 };
683
684 struct vm_exit {
685 enum vm_exitcode exitcode;
686 int inst_length; /* 0 means unknown */
687 uint64_t rip;
688 union {
689 struct vm_inout inout;
690 struct vm_inout_str inout_str;
691 struct {
692 uint64_t gpa;
693 int fault_type;
694 } paging;
695 struct {
696 uint64_t gpa;
697 uint64_t gla;
698 uint64_t cs_base;
699 int cs_d; /* CS.D */
700 struct vm_guest_paging paging;
701 struct vie vie;
702 } inst_emul;
703 /*
704 * VMX specific payload. Used when there is no "better"
705 * exitcode to represent the VM-exit.
706 */
707 struct {
708 int status; /* vmx inst status */
709 /*
710 * 'exit_reason' and 'exit_qualification' are valid
711 * only if 'status' is zero.
712 */
713 uint32_t exit_reason;
714 uint64_t exit_qualification;
715 /*
716 * 'inst_error' and 'inst_type' are valid
717 * only if 'status' is non-zero.
718 */
719 int inst_type;
720 int inst_error;
721 } vmx;
722 /*
723 * SVM specific payload.
724 */
725 struct {
726 uint64_t exitcode;
727 uint64_t exitinfo1;
728 uint64_t exitinfo2;
729 } svm;
730 struct {
731 int inst_length;
732 } bpt;
733 struct {
734 uint32_t code; /* ecx value */
735 uint64_t wval;
736 } msr;
737 struct {
738 int vcpu;
739 uint64_t rip;
740 } spinup_ap;
741 struct {
742 uint64_t rflags;
743 uint64_t intr_status;
744 } hlt;
745 struct {
746 int vector;
747 } ioapic_eoi;
748 struct {
749 enum vm_suspend_how how;
750 } suspended;
751 struct {
752 uint32_t mode;
753 uint8_t vector;
754 cpuset_t dmask;
755 } ipi;
756 struct vm_task_switch task_switch;
757 } u;
758 };
759
760 /* APIs to inject faults into the guest */
761 #ifdef _KERNEL
762 void vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid,
763 int errcode);
764
765 static __inline void
766 vm_inject_ud(struct vcpu *vcpu)
767 {
768 vm_inject_fault(vcpu, IDT_UD, 0, 0);
769 }
770
771 static __inline void
772 vm_inject_gp(struct vcpu *vcpu)
773 {
774 vm_inject_fault(vcpu, IDT_GP, 1, 0);
775 }
776
777 static __inline void
778 vm_inject_ac(struct vcpu *vcpu, int errcode)
779 {
780 vm_inject_fault(vcpu, IDT_AC, 1, errcode);
781 }
782
783 static __inline void
784 vm_inject_ss(struct vcpu *vcpu, int errcode)
785 {
786 vm_inject_fault(vcpu, IDT_SS, 1, errcode);
787 }
788
789 void vm_inject_pf(struct vcpu *vcpu, int error_code, uint64_t cr2);
790 #else
791 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
792 int errcode);
793
794 static __inline void
795 vm_inject_ud(void *vm, int vcpuid)
796 {
797 vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
798 }
799
800 static __inline void
801 vm_inject_gp(void *vm, int vcpuid)
802 {
803 vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
804 }
805
806 static __inline void
807 vm_inject_ac(void *vm, int vcpuid, int errcode)
808 {
809 vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
810 }
811
812 static __inline void
813 vm_inject_ss(void *vm, int vcpuid, int errcode)
814 {
815 vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
816 }
817
818 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
819 #endif
820
821 #endif /* _VMM_H_ */
Cache object: 59ee3d88807e40521ee0fced34b291ab
|