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