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 vm_snapshot_meta;
37
38 #ifdef _KERNEL
39 SDT_PROVIDER_DECLARE(vmm);
40 #endif
41
42 enum vm_suspend_how {
43 VM_SUSPEND_NONE,
44 VM_SUSPEND_RESET,
45 VM_SUSPEND_POWEROFF,
46 VM_SUSPEND_HALT,
47 VM_SUSPEND_TRIPLEFAULT,
48 VM_SUSPEND_LAST
49 };
50
51 /*
52 * Identifiers for architecturally defined registers.
53 */
54 enum vm_reg_name {
55 VM_REG_GUEST_RAX,
56 VM_REG_GUEST_RBX,
57 VM_REG_GUEST_RCX,
58 VM_REG_GUEST_RDX,
59 VM_REG_GUEST_RSI,
60 VM_REG_GUEST_RDI,
61 VM_REG_GUEST_RBP,
62 VM_REG_GUEST_R8,
63 VM_REG_GUEST_R9,
64 VM_REG_GUEST_R10,
65 VM_REG_GUEST_R11,
66 VM_REG_GUEST_R12,
67 VM_REG_GUEST_R13,
68 VM_REG_GUEST_R14,
69 VM_REG_GUEST_R15,
70 VM_REG_GUEST_CR0,
71 VM_REG_GUEST_CR3,
72 VM_REG_GUEST_CR4,
73 VM_REG_GUEST_DR7,
74 VM_REG_GUEST_RSP,
75 VM_REG_GUEST_RIP,
76 VM_REG_GUEST_RFLAGS,
77 VM_REG_GUEST_ES,
78 VM_REG_GUEST_CS,
79 VM_REG_GUEST_SS,
80 VM_REG_GUEST_DS,
81 VM_REG_GUEST_FS,
82 VM_REG_GUEST_GS,
83 VM_REG_GUEST_LDTR,
84 VM_REG_GUEST_TR,
85 VM_REG_GUEST_IDTR,
86 VM_REG_GUEST_GDTR,
87 VM_REG_GUEST_EFER,
88 VM_REG_GUEST_CR2,
89 VM_REG_GUEST_PDPTE0,
90 VM_REG_GUEST_PDPTE1,
91 VM_REG_GUEST_PDPTE2,
92 VM_REG_GUEST_PDPTE3,
93 VM_REG_GUEST_INTR_SHADOW,
94 VM_REG_GUEST_DR0,
95 VM_REG_GUEST_DR1,
96 VM_REG_GUEST_DR2,
97 VM_REG_GUEST_DR3,
98 VM_REG_GUEST_DR6,
99 VM_REG_GUEST_ENTRY_INST_LENGTH,
100 VM_REG_LAST
101 };
102
103 enum x2apic_state {
104 X2APIC_DISABLED,
105 X2APIC_ENABLED,
106 X2APIC_STATE_LAST
107 };
108
109 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
110 #define VM_INTINFO_DEL_ERRCODE 0x800
111 #define VM_INTINFO_RSVD 0x7ffff000
112 #define VM_INTINFO_VALID 0x80000000
113 #define VM_INTINFO_TYPE 0x700
114 #define VM_INTINFO_HWINTR (0 << 8)
115 #define VM_INTINFO_NMI (2 << 8)
116 #define VM_INTINFO_HWEXCEPTION (3 << 8)
117 #define VM_INTINFO_SWINTR (4 << 8)
118
119 /*
120 * The VM name has to fit into the pathname length constraints of devfs,
121 * governed primarily by SPECNAMELEN. The length is the total number of
122 * characters in the full path, relative to the mount point and not
123 * including any leading '/' characters.
124 * A prefix and a suffix are added to the name specified by the user.
125 * The prefix is usually "vmm/" or "vmm.io/", but can be a few characters
126 * longer for future use.
127 * The suffix is a string that identifies a bootrom image or some similar
128 * image that is attached to the VM. A separator character gets added to
129 * the suffix automatically when generating the full path, so it must be
130 * accounted for, reducing the effective length by 1.
131 * The effective length of a VM name is 229 bytes for FreeBSD 13 and 37
132 * bytes for FreeBSD 12. A minimum length is set for safety and supports
133 * a SPECNAMELEN as small as 32 on old systems.
134 */
135 #define VM_MAX_PREFIXLEN 10
136 #define VM_MAX_SUFFIXLEN 15
137 #define VM_MIN_NAMELEN 6
138 #define VM_MAX_NAMELEN \
139 (SPECNAMELEN - VM_MAX_PREFIXLEN - VM_MAX_SUFFIXLEN - 1)
140
141 #ifdef _KERNEL
142 CTASSERT(VM_MAX_NAMELEN >= VM_MIN_NAMELEN);
143
144 struct vcpu;
145 struct vm;
146 struct vm_exception;
147 struct seg_desc;
148 struct vm_exit;
149 struct vm_run;
150 struct vhpet;
151 struct vioapic;
152 struct vlapic;
153 struct vmspace;
154 struct vm_object;
155 struct vm_guest_paging;
156 struct pmap;
157 enum snapshot_req;
158
159 struct vm_eventinfo {
160 cpuset_t *rptr; /* rendezvous cookie */
161 int *sptr; /* suspend cookie */
162 int *iptr; /* reqidle cookie */
163 };
164
165 typedef int (*vmm_init_func_t)(int ipinum);
166 typedef int (*vmm_cleanup_func_t)(void);
167 typedef void (*vmm_resume_func_t)(void);
168 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
169 typedef int (*vmi_run_func_t)(void *vcpui, register_t rip,
170 struct pmap *pmap, struct vm_eventinfo *info);
171 typedef void (*vmi_cleanup_func_t)(void *vmi);
172 typedef void * (*vmi_vcpu_init_func_t)(void *vmi, struct vcpu *vcpu,
173 int vcpu_id);
174 typedef void (*vmi_vcpu_cleanup_func_t)(void *vcpui);
175 typedef int (*vmi_get_register_t)(void *vcpui, int num, uint64_t *retval);
176 typedef int (*vmi_set_register_t)(void *vcpui, int num, uint64_t val);
177 typedef int (*vmi_get_desc_t)(void *vcpui, int num, struct seg_desc *desc);
178 typedef int (*vmi_set_desc_t)(void *vcpui, int num, struct seg_desc *desc);
179 typedef int (*vmi_get_cap_t)(void *vcpui, int num, int *retval);
180 typedef int (*vmi_set_cap_t)(void *vcpui, int num, int val);
181 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
182 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
183 typedef struct vlapic * (*vmi_vlapic_init)(void *vcpui);
184 typedef void (*vmi_vlapic_cleanup)(struct vlapic *vlapic);
185 typedef int (*vmi_snapshot_vcpu_t)(void *vcpui, struct vm_snapshot_meta *meta);
186 typedef int (*vmi_restore_tsc_t)(void *vcpui, uint64_t now);
187
188 struct vmm_ops {
189 vmm_init_func_t modinit; /* module wide initialization */
190 vmm_cleanup_func_t modcleanup;
191 vmm_resume_func_t modresume;
192
193 vmi_init_func_t init; /* vm-specific initialization */
194 vmi_run_func_t run;
195 vmi_cleanup_func_t cleanup;
196 vmi_vcpu_init_func_t vcpu_init;
197 vmi_vcpu_cleanup_func_t vcpu_cleanup;
198 vmi_get_register_t getreg;
199 vmi_set_register_t setreg;
200 vmi_get_desc_t getdesc;
201 vmi_set_desc_t setdesc;
202 vmi_get_cap_t getcap;
203 vmi_set_cap_t setcap;
204 vmi_vmspace_alloc vmspace_alloc;
205 vmi_vmspace_free vmspace_free;
206 vmi_vlapic_init vlapic_init;
207 vmi_vlapic_cleanup vlapic_cleanup;
208
209 /* checkpoint operations */
210 vmi_snapshot_vcpu_t vcpu_snapshot;
211 vmi_restore_tsc_t restore_tsc;
212 };
213
214 extern const struct vmm_ops vmm_ops_intel;
215 extern const struct vmm_ops vmm_ops_amd;
216
217 extern u_int vm_maxcpu; /* maximum virtual cpus */
218
219 int vm_create(const char *name, struct vm **retvm);
220 struct vcpu *vm_alloc_vcpu(struct vm *vm, int vcpuid);
221 void vm_disable_vcpu_creation(struct vm *vm);
222 void vm_slock_vcpus(struct vm *vm);
223 void vm_unlock_vcpus(struct vm *vm);
224 void vm_destroy(struct vm *vm);
225 int vm_reinit(struct vm *vm);
226 const char *vm_name(struct vm *vm);
227 uint16_t vm_get_maxcpus(struct vm *vm);
228 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
229 uint16_t *threads, uint16_t *maxcpus);
230 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
231 uint16_t threads, uint16_t maxcpus);
232
233 /*
234 * APIs that modify the guest memory map require all vcpus to be frozen.
235 */
236 void vm_slock_memsegs(struct vm *vm);
237 void vm_xlock_memsegs(struct vm *vm);
238 void vm_unlock_memsegs(struct vm *vm);
239 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
240 size_t len, int prot, int flags);
241 int vm_munmap_memseg(struct vm *vm, vm_paddr_t gpa, size_t len);
242 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
243 void vm_free_memseg(struct vm *vm, int ident);
244 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
245 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
246 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
247 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
248
249 /*
250 * APIs that inspect the guest memory map require only a *single* vcpu to
251 * be frozen. This acts like a read lock on the guest memory map since any
252 * modification requires *all* vcpus to be frozen.
253 */
254 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
255 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
256 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
257 struct vm_object **objptr);
258 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
259 void *vm_gpa_hold(struct vcpu *vcpu, vm_paddr_t gpa, size_t len,
260 int prot, void **cookie);
261 void *vm_gpa_hold_global(struct vm *vm, vm_paddr_t gpa, size_t len,
262 int prot, void **cookie);
263 void vm_gpa_release(void *cookie);
264 bool vm_mem_allocated(struct vcpu *vcpu, vm_paddr_t gpa);
265
266 int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval);
267 int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val);
268 int vm_get_seg_desc(struct vcpu *vcpu, int reg,
269 struct seg_desc *ret_desc);
270 int vm_set_seg_desc(struct vcpu *vcpu, int reg,
271 struct seg_desc *desc);
272 int vm_run(struct vcpu *vcpu, struct vm_exit *vme_user);
273 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
274 int vm_inject_nmi(struct vcpu *vcpu);
275 int vm_nmi_pending(struct vcpu *vcpu);
276 void vm_nmi_clear(struct vcpu *vcpu);
277 int vm_inject_extint(struct vcpu *vcpu);
278 int vm_extint_pending(struct vcpu *vcpu);
279 void vm_extint_clear(struct vcpu *vcpu);
280 int vcpu_vcpuid(struct vcpu *vcpu);
281 struct vm *vcpu_vm(struct vcpu *vcpu);
282 struct vcpu *vm_vcpu(struct vm *vm, int cpu);
283 struct vlapic *vm_lapic(struct vcpu *vcpu);
284 struct vioapic *vm_ioapic(struct vm *vm);
285 struct vhpet *vm_hpet(struct vm *vm);
286 int vm_get_capability(struct vcpu *vcpu, int type, int *val);
287 int vm_set_capability(struct vcpu *vcpu, int type, int val);
288 int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state);
289 int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state);
290 int vm_apicid2vcpuid(struct vm *vm, int apicid);
291 int vm_activate_cpu(struct vcpu *vcpu);
292 int vm_suspend_cpu(struct vm *vm, struct vcpu *vcpu);
293 int vm_resume_cpu(struct vm *vm, struct vcpu *vcpu);
294 int vm_restart_instruction(struct vcpu *vcpu);
295 struct vm_exit *vm_exitinfo(struct vcpu *vcpu);
296 void vm_exit_suspended(struct vcpu *vcpu, uint64_t rip);
297 void vm_exit_debug(struct vcpu *vcpu, uint64_t rip);
298 void vm_exit_rendezvous(struct vcpu *vcpu, uint64_t rip);
299 void vm_exit_astpending(struct vcpu *vcpu, uint64_t rip);
300 void vm_exit_reqidle(struct vcpu *vcpu, uint64_t rip);
301 int vm_snapshot_req(struct vm *vm, struct vm_snapshot_meta *meta);
302 int vm_restore_time(struct vm *vm);
303
304 #ifdef _SYS__CPUSET_H_
305 /*
306 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
307 * The rendezvous 'func(arg)' is not allowed to do anything that will
308 * cause the thread to be put to sleep.
309 *
310 * The caller cannot hold any locks when initiating the rendezvous.
311 *
312 * The implementation of this API may cause vcpus other than those specified
313 * by 'dest' to be stalled. The caller should not rely on any vcpus making
314 * forward progress when the rendezvous is in progress.
315 */
316 typedef void (*vm_rendezvous_func_t)(struct vcpu *vcpu, void *arg);
317 int vm_smp_rendezvous(struct vcpu *vcpu, cpuset_t dest,
318 vm_rendezvous_func_t func, void *arg);
319
320 cpuset_t vm_active_cpus(struct vm *vm);
321 cpuset_t vm_debug_cpus(struct vm *vm);
322 cpuset_t vm_suspended_cpus(struct vm *vm);
323 cpuset_t vm_start_cpus(struct vm *vm, const cpuset_t *tostart);
324 void vm_await_start(struct vm *vm, const cpuset_t *waiting);
325 #endif /* _SYS__CPUSET_H_ */
326
327 static __inline int
vcpu_rendezvous_pending(struct vcpu * vcpu,struct vm_eventinfo * info)328 vcpu_rendezvous_pending(struct vcpu *vcpu, struct vm_eventinfo *info)
329 {
330 /*
331 * This check isn't done with atomic operations or under a lock because
332 * there's no need to. If the vcpuid bit is set, the vcpu is part of a
333 * rendezvous and the bit won't be cleared until the vcpu enters the
334 * rendezvous. On rendezvous exit, the cpuset is cleared and the vcpu
335 * will see an empty cpuset. So, the races are harmless.
336 */
337 return (CPU_ISSET(vcpu_vcpuid(vcpu), info->rptr));
338 }
339
340 static __inline int
vcpu_suspended(struct vm_eventinfo * info)341 vcpu_suspended(struct vm_eventinfo *info)
342 {
343
344 return (*info->sptr);
345 }
346
347 static __inline int
vcpu_reqidle(struct vm_eventinfo * info)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
vcpu_is_running(struct vcpu * vcpu,int * hostcpu)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
vcpu_should_yield(struct vcpu * vcpu)384 vcpu_should_yield(struct vcpu *vcpu)
385 {
386
387 if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
388 return (1);
389 else if (curthread->td_owepreempt)
390 return (1);
391 else
392 return (0);
393 }
394 #endif
395
396 void *vcpu_stats(struct vcpu *vcpu);
397 void vcpu_notify_event(struct vcpu *vcpu, bool lapic_intr);
398 struct vmspace *vm_get_vmspace(struct vm *vm);
399 struct vatpic *vm_atpic(struct vm *vm);
400 struct vatpit *vm_atpit(struct vm *vm);
401 struct vpmtmr *vm_pmtmr(struct vm *vm);
402 struct vrtc *vm_rtc(struct vm *vm);
403
404 /*
405 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
406 * success and non-zero on failure.
407 *
408 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
409 * this function directly because they enforce the trap-like or fault-like
410 * behavior of an exception.
411 *
412 * This function should only be called in the context of the thread that is
413 * executing this vcpu.
414 */
415 int vm_inject_exception(struct vcpu *vcpu, int vector, int err_valid,
416 uint32_t errcode, int restart_instruction);
417
418 /*
419 * This function is called after a VM-exit that occurred during exception or
420 * interrupt delivery through the IDT. The format of 'intinfo' is described
421 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
422 *
423 * If a VM-exit handler completes the event delivery successfully then it
424 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
425 * if the task switch emulation is triggered via a task gate then it should
426 * call this function with 'intinfo=0' to indicate that the external event
427 * is not pending anymore.
428 *
429 * Return value is 0 on success and non-zero on failure.
430 */
431 int vm_exit_intinfo(struct vcpu *vcpu, uint64_t intinfo);
432
433 /*
434 * This function is called before every VM-entry to retrieve a pending
435 * event that should be injected into the guest. This function combines
436 * nested events into a double or triple fault.
437 *
438 * Returns 0 if there are no events that need to be injected into the guest
439 * and non-zero otherwise.
440 */
441 int vm_entry_intinfo(struct vcpu *vcpu, uint64_t *info);
442
443 int vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2);
444
445 /*
446 * Function used to keep track of the guest's TSC offset. The
447 * offset is used by the virutalization extensions to provide a consistent
448 * value for the Time Stamp Counter to the guest.
449 */
450 void vm_set_tsc_offset(struct vcpu *vcpu, uint64_t offset);
451
452 enum vm_reg_name vm_segment_name(int seg_encoding);
453
454 struct vm_copyinfo {
455 uint64_t gpa;
456 size_t len;
457 void *hva;
458 void *cookie;
459 };
460
461 /*
462 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
463 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
464 * a copyin or PROT_WRITE for a copyout.
465 *
466 * retval is_fault Interpretation
467 * 0 0 Success
468 * 0 1 An exception was injected into the guest
469 * EFAULT N/A Unrecoverable error
470 *
471 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
472 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
473 * 'vm_copy_teardown()' after the copy is done.
474 */
475 int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging,
476 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
477 int num_copyinfo, int *is_fault);
478 void vm_copy_teardown(struct vm_copyinfo *copyinfo, int num_copyinfo);
479 void vm_copyin(struct vm_copyinfo *copyinfo, void *kaddr, size_t len);
480 void vm_copyout(const void *kaddr, struct vm_copyinfo *copyinfo, size_t len);
481
482 int vcpu_trace_exceptions(struct vcpu *vcpu);
483 int vcpu_trap_wbinvd(struct vcpu *vcpu);
484 #endif /* KERNEL */
485
486 /*
487 * Identifiers for optional vmm capabilities
488 */
489 enum vm_cap_type {
490 VM_CAP_HALT_EXIT,
491 VM_CAP_MTRAP_EXIT,
492 VM_CAP_PAUSE_EXIT,
493 VM_CAP_UNRESTRICTED_GUEST,
494 VM_CAP_ENABLE_INVPCID,
495 VM_CAP_BPT_EXIT,
496 VM_CAP_RDPID,
497 VM_CAP_RDTSCP,
498 VM_CAP_IPI_EXIT,
499 VM_CAP_MASK_HWINTR,
500 VM_CAP_MAX
501 };
502
503 enum vm_intr_trigger {
504 EDGE_TRIGGER,
505 LEVEL_TRIGGER
506 };
507
508 /*
509 * The 'access' field has the format specified in Table 21-2 of the Intel
510 * Architecture Manual vol 3b.
511 *
512 * XXX The contents of the 'access' field are architecturally defined except
513 * bit 16 - Segment Unusable.
514 */
515 struct seg_desc {
516 uint64_t base;
517 uint32_t limit;
518 uint32_t access;
519 };
520 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
521 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
522 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
523 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
524 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
525 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
526
527 enum vm_cpu_mode {
528 CPU_MODE_REAL,
529 CPU_MODE_PROTECTED,
530 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
531 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
532 };
533
534 enum vm_paging_mode {
535 PAGING_MODE_FLAT,
536 PAGING_MODE_32,
537 PAGING_MODE_PAE,
538 PAGING_MODE_64,
539 PAGING_MODE_64_LA57,
540 };
541
542 struct vm_guest_paging {
543 uint64_t cr3;
544 int cpl;
545 enum vm_cpu_mode cpu_mode;
546 enum vm_paging_mode paging_mode;
547 };
548
549 /*
550 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
551 * consumers of instruction decoding. The only reason why their contents
552 * need to be exposed is because they are part of the 'vm_exit' structure.
553 */
554 struct vie_op {
555 uint8_t op_byte; /* actual opcode byte */
556 uint8_t op_type; /* type of operation (e.g. MOV) */
557 uint16_t op_flags;
558 };
559 _Static_assert(sizeof(struct vie_op) == 4, "ABI");
560 _Static_assert(_Alignof(struct vie_op) == 2, "ABI");
561
562 #define VIE_INST_SIZE 15
563 struct vie {
564 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
565 uint8_t num_valid; /* size of the instruction */
566
567 /* The following fields are all zeroed upon restart. */
568 #define vie_startzero num_processed
569 uint8_t num_processed;
570
571 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
572 uint8_t rex_w:1, /* REX prefix */
573 rex_r:1,
574 rex_x:1,
575 rex_b:1,
576 rex_present:1,
577 repz_present:1, /* REP/REPE/REPZ prefix */
578 repnz_present:1, /* REPNE/REPNZ prefix */
579 opsize_override:1, /* Operand size override */
580 addrsize_override:1, /* Address size override */
581 segment_override:1; /* Segment override */
582
583 uint8_t mod:2, /* ModRM byte */
584 reg:4,
585 rm:4;
586
587 uint8_t ss:2, /* SIB byte */
588 vex_present:1, /* VEX prefixed */
589 vex_l:1, /* L bit */
590 index:4, /* SIB byte */
591 base:4; /* SIB byte */
592
593 uint8_t disp_bytes;
594 uint8_t imm_bytes;
595
596 uint8_t scale;
597
598 uint8_t vex_reg:4, /* vvvv: first source register specifier */
599 vex_pp:2, /* pp */
600 _sparebits:2;
601
602 uint8_t _sparebytes[2];
603
604 int base_register; /* VM_REG_GUEST_xyz */
605 int index_register; /* VM_REG_GUEST_xyz */
606 int segment_register; /* VM_REG_GUEST_xyz */
607
608 int64_t displacement; /* optional addr displacement */
609 int64_t immediate; /* optional immediate operand */
610
611 uint8_t decoded; /* set to 1 if successfully decoded */
612
613 uint8_t _sparebyte;
614
615 struct vie_op op; /* opcode description */
616 };
617 _Static_assert(sizeof(struct vie) == 64, "ABI");
618 _Static_assert(__offsetof(struct vie, disp_bytes) == 22, "ABI");
619 _Static_assert(__offsetof(struct vie, scale) == 24, "ABI");
620 _Static_assert(__offsetof(struct vie, base_register) == 28, "ABI");
621
622 enum vm_exitcode {
623 VM_EXITCODE_INOUT,
624 VM_EXITCODE_VMX,
625 VM_EXITCODE_BOGUS,
626 VM_EXITCODE_RDMSR,
627 VM_EXITCODE_WRMSR,
628 VM_EXITCODE_HLT,
629 VM_EXITCODE_MTRAP,
630 VM_EXITCODE_PAUSE,
631 VM_EXITCODE_PAGING,
632 VM_EXITCODE_INST_EMUL,
633 VM_EXITCODE_SPINUP_AP,
634 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
635 VM_EXITCODE_RENDEZVOUS,
636 VM_EXITCODE_IOAPIC_EOI,
637 VM_EXITCODE_SUSPENDED,
638 VM_EXITCODE_INOUT_STR,
639 VM_EXITCODE_TASK_SWITCH,
640 VM_EXITCODE_MONITOR,
641 VM_EXITCODE_MWAIT,
642 VM_EXITCODE_SVM,
643 VM_EXITCODE_REQIDLE,
644 VM_EXITCODE_DEBUG,
645 VM_EXITCODE_VMINSN,
646 VM_EXITCODE_BPT,
647 VM_EXITCODE_IPI,
648 VM_EXITCODE_MAX
649 };
650
651 struct vm_inout {
652 uint16_t bytes:3; /* 1 or 2 or 4 */
653 uint16_t in:1;
654 uint16_t string:1;
655 uint16_t rep:1;
656 uint16_t port;
657 uint32_t eax; /* valid for out */
658 };
659
660 struct vm_inout_str {
661 struct vm_inout inout; /* must be the first element */
662 struct vm_guest_paging paging;
663 uint64_t rflags;
664 uint64_t cr0;
665 uint64_t index;
666 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
667 int addrsize;
668 enum vm_reg_name seg_name;
669 struct seg_desc seg_desc;
670 };
671
672 enum task_switch_reason {
673 TSR_CALL,
674 TSR_IRET,
675 TSR_JMP,
676 TSR_IDT_GATE, /* task gate in IDT */
677 };
678
679 struct vm_task_switch {
680 uint16_t tsssel; /* new TSS selector */
681 int ext; /* task switch due to external event */
682 uint32_t errcode;
683 int errcode_valid; /* push 'errcode' on the new stack */
684 enum task_switch_reason reason;
685 struct vm_guest_paging paging;
686 };
687
688 struct vm_exit {
689 enum vm_exitcode exitcode;
690 int inst_length; /* 0 means unknown */
691 uint64_t rip;
692 union {
693 struct vm_inout inout;
694 struct vm_inout_str inout_str;
695 struct {
696 uint64_t gpa;
697 int fault_type;
698 } paging;
699 struct {
700 uint64_t gpa;
701 uint64_t gla;
702 uint64_t cs_base;
703 int cs_d; /* CS.D */
704 struct vm_guest_paging paging;
705 struct vie vie;
706 } inst_emul;
707 /*
708 * VMX specific payload. Used when there is no "better"
709 * exitcode to represent the VM-exit.
710 */
711 struct {
712 int status; /* vmx inst status */
713 /*
714 * 'exit_reason' and 'exit_qualification' are valid
715 * only if 'status' is zero.
716 */
717 uint32_t exit_reason;
718 uint64_t exit_qualification;
719 /*
720 * 'inst_error' and 'inst_type' are valid
721 * only if 'status' is non-zero.
722 */
723 int inst_type;
724 int inst_error;
725 } vmx;
726 /*
727 * SVM specific payload.
728 */
729 struct {
730 uint64_t exitcode;
731 uint64_t exitinfo1;
732 uint64_t exitinfo2;
733 } svm;
734 struct {
735 int inst_length;
736 } bpt;
737 struct {
738 uint32_t code; /* ecx value */
739 uint64_t wval;
740 } msr;
741 struct {
742 int vcpu;
743 uint64_t rip;
744 } spinup_ap;
745 struct {
746 uint64_t rflags;
747 uint64_t intr_status;
748 } hlt;
749 struct {
750 int vector;
751 } ioapic_eoi;
752 struct {
753 enum vm_suspend_how how;
754 } suspended;
755 struct {
756 uint32_t mode;
757 uint8_t vector;
758 cpuset_t dmask;
759 } ipi;
760 struct vm_task_switch task_switch;
761 } u;
762 };
763
764 /* APIs to inject faults into the guest */
765 #ifdef _KERNEL
766 void vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid,
767 int errcode);
768
769 static __inline void
vm_inject_ud(struct vcpu * vcpu)770 vm_inject_ud(struct vcpu *vcpu)
771 {
772 vm_inject_fault(vcpu, IDT_UD, 0, 0);
773 }
774
775 static __inline void
vm_inject_gp(struct vcpu * vcpu)776 vm_inject_gp(struct vcpu *vcpu)
777 {
778 vm_inject_fault(vcpu, IDT_GP, 1, 0);
779 }
780
781 static __inline void
vm_inject_ac(struct vcpu * vcpu,int errcode)782 vm_inject_ac(struct vcpu *vcpu, int errcode)
783 {
784 vm_inject_fault(vcpu, IDT_AC, 1, errcode);
785 }
786
787 static __inline void
vm_inject_ss(struct vcpu * vcpu,int errcode)788 vm_inject_ss(struct vcpu *vcpu, int errcode)
789 {
790 vm_inject_fault(vcpu, IDT_SS, 1, errcode);
791 }
792
793 void vm_inject_pf(struct vcpu *vcpu, int error_code, uint64_t cr2);
794 #else
795 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
796 int errcode);
797
798 static __inline void
vm_inject_ud(void * vm,int vcpuid)799 vm_inject_ud(void *vm, int vcpuid)
800 {
801 vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
802 }
803
804 static __inline void
vm_inject_gp(void * vm,int vcpuid)805 vm_inject_gp(void *vm, int vcpuid)
806 {
807 vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
808 }
809
810 static __inline void
vm_inject_ac(void * vm,int vcpuid,int errcode)811 vm_inject_ac(void *vm, int vcpuid, int errcode)
812 {
813 vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
814 }
815
816 static __inline void
vm_inject_ss(void * vm,int vcpuid,int errcode)817 vm_inject_ss(void *vm, int vcpuid, int errcode)
818 {
819 vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
820 }
821
822 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
823 #endif
824
825 #endif /* _VMM_H_ */
826