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/sdt.h> 35 #include <x86/segments.h> 36 37 #ifdef _KERNEL 38 SDT_PROVIDER_DECLARE(vmm); 39 #endif 40 41 enum vm_suspend_how { 42 VM_SUSPEND_NONE, 43 VM_SUSPEND_RESET, 44 VM_SUSPEND_POWEROFF, 45 VM_SUSPEND_HALT, 46 VM_SUSPEND_TRIPLEFAULT, 47 VM_SUSPEND_LAST 48 }; 49 50 /* 51 * Identifiers for architecturally defined registers. 52 */ 53 enum vm_reg_name { 54 VM_REG_GUEST_RAX, 55 VM_REG_GUEST_RBX, 56 VM_REG_GUEST_RCX, 57 VM_REG_GUEST_RDX, 58 VM_REG_GUEST_RSI, 59 VM_REG_GUEST_RDI, 60 VM_REG_GUEST_RBP, 61 VM_REG_GUEST_R8, 62 VM_REG_GUEST_R9, 63 VM_REG_GUEST_R10, 64 VM_REG_GUEST_R11, 65 VM_REG_GUEST_R12, 66 VM_REG_GUEST_R13, 67 VM_REG_GUEST_R14, 68 VM_REG_GUEST_R15, 69 VM_REG_GUEST_CR0, 70 VM_REG_GUEST_CR3, 71 VM_REG_GUEST_CR4, 72 VM_REG_GUEST_DR7, 73 VM_REG_GUEST_RSP, 74 VM_REG_GUEST_RIP, 75 VM_REG_GUEST_RFLAGS, 76 VM_REG_GUEST_ES, 77 VM_REG_GUEST_CS, 78 VM_REG_GUEST_SS, 79 VM_REG_GUEST_DS, 80 VM_REG_GUEST_FS, 81 VM_REG_GUEST_GS, 82 VM_REG_GUEST_LDTR, 83 VM_REG_GUEST_TR, 84 VM_REG_GUEST_IDTR, 85 VM_REG_GUEST_GDTR, 86 VM_REG_GUEST_EFER, 87 VM_REG_GUEST_CR2, 88 VM_REG_GUEST_PDPTE0, 89 VM_REG_GUEST_PDPTE1, 90 VM_REG_GUEST_PDPTE2, 91 VM_REG_GUEST_PDPTE3, 92 VM_REG_GUEST_INTR_SHADOW, 93 VM_REG_GUEST_DR0, 94 VM_REG_GUEST_DR1, 95 VM_REG_GUEST_DR2, 96 VM_REG_GUEST_DR3, 97 VM_REG_GUEST_DR6, 98 VM_REG_LAST 99 }; 100 101 enum x2apic_state { 102 X2APIC_DISABLED, 103 X2APIC_ENABLED, 104 X2APIC_STATE_LAST 105 }; 106 107 #define VM_INTINFO_VECTOR(info) ((info) & 0xff) 108 #define VM_INTINFO_DEL_ERRCODE 0x800 109 #define VM_INTINFO_RSVD 0x7ffff000 110 #define VM_INTINFO_VALID 0x80000000 111 #define VM_INTINFO_TYPE 0x700 112 #define VM_INTINFO_HWINTR (0 << 8) 113 #define VM_INTINFO_NMI (2 << 8) 114 #define VM_INTINFO_HWEXCEPTION (3 << 8) 115 #define VM_INTINFO_SWINTR (4 << 8) 116 117 #ifdef _KERNEL 118 119 #define VM_MAX_NAMELEN 32 120 121 struct vm; 122 struct vm_exception; 123 struct seg_desc; 124 struct vm_exit; 125 struct vm_run; 126 struct vhpet; 127 struct vioapic; 128 struct vlapic; 129 struct vmspace; 130 struct vm_object; 131 struct vm_guest_paging; 132 struct pmap; 133 134 struct vm_eventinfo { 135 void *rptr; /* rendezvous cookie */ 136 int *sptr; /* suspend cookie */ 137 int *iptr; /* reqidle cookie */ 138 }; 139 140 typedef int (*vmm_init_func_t)(int ipinum); 141 typedef int (*vmm_cleanup_func_t)(void); 142 typedef void (*vmm_resume_func_t)(void); 143 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap); 144 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip, 145 struct pmap *pmap, struct vm_eventinfo *info); 146 typedef void (*vmi_cleanup_func_t)(void *vmi); 147 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 148 uint64_t *retval); 149 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 150 uint64_t val); 151 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 152 struct seg_desc *desc); 153 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 154 struct seg_desc *desc); 155 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 156 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 157 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max); 158 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace); 159 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu); 160 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 161 162 struct vmm_ops { 163 vmm_init_func_t init; /* module wide initialization */ 164 vmm_cleanup_func_t cleanup; 165 vmm_resume_func_t resume; 166 167 vmi_init_func_t vminit; /* vm-specific initialization */ 168 vmi_run_func_t vmrun; 169 vmi_cleanup_func_t vmcleanup; 170 vmi_get_register_t vmgetreg; 171 vmi_set_register_t vmsetreg; 172 vmi_get_desc_t vmgetdesc; 173 vmi_set_desc_t vmsetdesc; 174 vmi_get_cap_t vmgetcap; 175 vmi_set_cap_t vmsetcap; 176 vmi_vmspace_alloc vmspace_alloc; 177 vmi_vmspace_free vmspace_free; 178 vmi_vlapic_init vlapic_init; 179 vmi_vlapic_cleanup vlapic_cleanup; 180 }; 181 182 extern struct vmm_ops vmm_ops_intel; 183 extern struct vmm_ops vmm_ops_amd; 184 185 int vm_create(const char *name, struct vm **retvm); 186 void vm_destroy(struct vm *vm); 187 int vm_reinit(struct vm *vm); 188 const char *vm_name(struct vm *vm); 189 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores, 190 uint16_t *threads, uint16_t *maxcpus); 191 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores, 192 uint16_t threads, uint16_t maxcpus); 193 194 /* 195 * APIs that modify the guest memory map require all vcpus to be frozen. 196 */ 197 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off, 198 size_t len, int prot, int flags); 199 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem); 200 void vm_free_memseg(struct vm *vm, int ident); 201 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 202 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 203 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func); 204 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func); 205 206 /* 207 * APIs that inspect the guest memory map require only a *single* vcpu to 208 * be frozen. This acts like a read lock on the guest memory map since any 209 * modification requires *all* vcpus to be frozen. 210 */ 211 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid, 212 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags); 213 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem, 214 struct vm_object **objptr); 215 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm); 216 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len, 217 int prot, void **cookie); 218 void vm_gpa_release(void *cookie); 219 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa); 220 221 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 222 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 223 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 224 struct seg_desc *ret_desc); 225 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 226 struct seg_desc *desc); 227 int vm_run(struct vm *vm, struct vm_run *vmrun); 228 int vm_suspend(struct vm *vm, enum vm_suspend_how how); 229 int vm_inject_nmi(struct vm *vm, int vcpu); 230 int vm_nmi_pending(struct vm *vm, int vcpuid); 231 void vm_nmi_clear(struct vm *vm, int vcpuid); 232 int vm_inject_extint(struct vm *vm, int vcpu); 233 int vm_extint_pending(struct vm *vm, int vcpuid); 234 void vm_extint_clear(struct vm *vm, int vcpuid); 235 struct vlapic *vm_lapic(struct vm *vm, int cpu); 236 struct vioapic *vm_ioapic(struct vm *vm); 237 struct vhpet *vm_hpet(struct vm *vm); 238 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 239 int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 240 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 241 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 242 int vm_apicid2vcpuid(struct vm *vm, int apicid); 243 int vm_activate_cpu(struct vm *vm, int vcpu); 244 int vm_suspend_cpu(struct vm *vm, int vcpu); 245 int vm_resume_cpu(struct vm *vm, int vcpu); 246 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 247 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 248 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip); 249 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip); 250 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip); 251 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip); 252 253 #ifdef _SYS__CPUSET_H_ 254 /* 255 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'. 256 * The rendezvous 'func(arg)' is not allowed to do anything that will 257 * cause the thread to be put to sleep. 258 * 259 * If the rendezvous is being initiated from a vcpu context then the 260 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1. 261 * 262 * The caller cannot hold any locks when initiating the rendezvous. 263 * 264 * The implementation of this API may cause vcpus other than those specified 265 * by 'dest' to be stalled. The caller should not rely on any vcpus making 266 * forward progress when the rendezvous is in progress. 267 */ 268 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg); 269 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest, 270 vm_rendezvous_func_t func, void *arg); 271 cpuset_t vm_active_cpus(struct vm *vm); 272 cpuset_t vm_debug_cpus(struct vm *vm); 273 cpuset_t vm_suspended_cpus(struct vm *vm); 274 #endif /* _SYS__CPUSET_H_ */ 275 276 static __inline int 277 vcpu_rendezvous_pending(struct vm_eventinfo *info) 278 { 279 280 return (*((uintptr_t *)(info->rptr)) != 0); 281 } 282 283 static __inline int 284 vcpu_suspended(struct vm_eventinfo *info) 285 { 286 287 return (*info->sptr); 288 } 289 290 static __inline int 291 vcpu_reqidle(struct vm_eventinfo *info) 292 { 293 294 return (*info->iptr); 295 } 296 297 int vcpu_debugged(struct vm *vm, int vcpuid); 298 299 /* 300 * Return 1 if device indicated by bus/slot/func is supposed to be a 301 * pci passthrough device. 302 * 303 * Return 0 otherwise. 304 */ 305 int vmm_is_pptdev(int bus, int slot, int func); 306 307 void *vm_iommu_domain(struct vm *vm); 308 309 enum vcpu_state { 310 VCPU_IDLE, 311 VCPU_FROZEN, 312 VCPU_RUNNING, 313 VCPU_SLEEPING, 314 }; 315 316 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 317 bool from_idle); 318 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 319 320 static int __inline 321 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 322 { 323 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 324 } 325 326 #ifdef _SYS_PROC_H_ 327 static int __inline 328 vcpu_should_yield(struct vm *vm, int vcpu) 329 { 330 331 if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED)) 332 return (1); 333 else if (curthread->td_owepreempt) 334 return (1); 335 else 336 return (0); 337 } 338 #endif 339 340 void *vcpu_stats(struct vm *vm, int vcpu); 341 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr); 342 struct vmspace *vm_get_vmspace(struct vm *vm); 343 struct vatpic *vm_atpic(struct vm *vm); 344 struct vatpit *vm_atpit(struct vm *vm); 345 struct vpmtmr *vm_pmtmr(struct vm *vm); 346 struct vrtc *vm_rtc(struct vm *vm); 347 348 /* 349 * Inject exception 'vector' into the guest vcpu. This function returns 0 on 350 * success and non-zero on failure. 351 * 352 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 353 * this function directly because they enforce the trap-like or fault-like 354 * behavior of an exception. 355 * 356 * This function should only be called in the context of the thread that is 357 * executing this vcpu. 358 */ 359 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid, 360 uint32_t errcode, int restart_instruction); 361 362 /* 363 * This function is called after a VM-exit that occurred during exception or 364 * interrupt delivery through the IDT. The format of 'intinfo' is described 365 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. 366 * 367 * If a VM-exit handler completes the event delivery successfully then it 368 * should call vm_exit_intinfo() to extinguish the pending event. For e.g., 369 * if the task switch emulation is triggered via a task gate then it should 370 * call this function with 'intinfo=0' to indicate that the external event 371 * is not pending anymore. 372 * 373 * Return value is 0 on success and non-zero on failure. 374 */ 375 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo); 376 377 /* 378 * This function is called before every VM-entry to retrieve a pending 379 * event that should be injected into the guest. This function combines 380 * nested events into a double or triple fault. 381 * 382 * Returns 0 if there are no events that need to be injected into the guest 383 * and non-zero otherwise. 384 */ 385 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info); 386 387 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2); 388 389 enum vm_reg_name vm_segment_name(int seg_encoding); 390 391 struct vm_copyinfo { 392 uint64_t gpa; 393 size_t len; 394 void *hva; 395 void *cookie; 396 }; 397 398 /* 399 * Set up 'copyinfo[]' to copy to/from guest linear address space starting 400 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for 401 * a copyin or PROT_WRITE for a copyout. 402 * 403 * retval is_fault Interpretation 404 * 0 0 Success 405 * 0 1 An exception was injected into the guest 406 * EFAULT N/A Unrecoverable error 407 * 408 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if 409 * the return value is 0. The 'copyinfo[]' resources should be freed by calling 410 * 'vm_copy_teardown()' after the copy is done. 411 */ 412 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging, 413 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo, 414 int num_copyinfo, int *is_fault); 415 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 416 int num_copyinfo); 417 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 418 void *kaddr, size_t len); 419 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr, 420 struct vm_copyinfo *copyinfo, size_t len); 421 422 int vcpu_trace_exceptions(struct vm *vm, int vcpuid); 423 #endif /* KERNEL */ 424 425 #define VM_MAXCPU 16 /* maximum virtual cpus */ 426 427 /* 428 * Identifiers for optional vmm capabilities 429 */ 430 enum vm_cap_type { 431 VM_CAP_HALT_EXIT, 432 VM_CAP_MTRAP_EXIT, 433 VM_CAP_PAUSE_EXIT, 434 VM_CAP_UNRESTRICTED_GUEST, 435 VM_CAP_ENABLE_INVPCID, 436 VM_CAP_MAX 437 }; 438 439 enum vm_intr_trigger { 440 EDGE_TRIGGER, 441 LEVEL_TRIGGER 442 }; 443 444 /* 445 * The 'access' field has the format specified in Table 21-2 of the Intel 446 * Architecture Manual vol 3b. 447 * 448 * XXX The contents of the 'access' field are architecturally defined except 449 * bit 16 - Segment Unusable. 450 */ 451 struct seg_desc { 452 uint64_t base; 453 uint32_t limit; 454 uint32_t access; 455 }; 456 #define SEG_DESC_TYPE(access) ((access) & 0x001f) 457 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3) 458 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0) 459 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0) 460 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0) 461 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0) 462 463 enum vm_cpu_mode { 464 CPU_MODE_REAL, 465 CPU_MODE_PROTECTED, 466 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */ 467 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */ 468 }; 469 470 enum vm_paging_mode { 471 PAGING_MODE_FLAT, 472 PAGING_MODE_32, 473 PAGING_MODE_PAE, 474 PAGING_MODE_64, 475 }; 476 477 struct vm_guest_paging { 478 uint64_t cr3; 479 int cpl; 480 enum vm_cpu_mode cpu_mode; 481 enum vm_paging_mode paging_mode; 482 }; 483 484 /* 485 * The data structures 'vie' and 'vie_op' are meant to be opaque to the 486 * consumers of instruction decoding. The only reason why their contents 487 * need to be exposed is because they are part of the 'vm_exit' structure. 488 */ 489 struct vie_op { 490 uint8_t op_byte; /* actual opcode byte */ 491 uint8_t op_type; /* type of operation (e.g. MOV) */ 492 uint16_t op_flags; 493 }; 494 495 #define VIE_INST_SIZE 15 496 struct vie { 497 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */ 498 uint8_t num_valid; /* size of the instruction */ 499 uint8_t num_processed; 500 501 uint8_t addrsize:4, opsize:4; /* address and operand sizes */ 502 uint8_t rex_w:1, /* REX prefix */ 503 rex_r:1, 504 rex_x:1, 505 rex_b:1, 506 rex_present:1, 507 repz_present:1, /* REP/REPE/REPZ prefix */ 508 repnz_present:1, /* REPNE/REPNZ prefix */ 509 opsize_override:1, /* Operand size override */ 510 addrsize_override:1, /* Address size override */ 511 segment_override:1; /* Segment override */ 512 513 uint8_t mod:2, /* ModRM byte */ 514 reg:4, 515 rm:4; 516 517 uint8_t ss:2, /* SIB byte */ 518 index:4, 519 base:4; 520 521 uint8_t disp_bytes; 522 uint8_t imm_bytes; 523 524 uint8_t scale; 525 int base_register; /* VM_REG_GUEST_xyz */ 526 int index_register; /* VM_REG_GUEST_xyz */ 527 int segment_register; /* VM_REG_GUEST_xyz */ 528 529 int64_t displacement; /* optional addr displacement */ 530 int64_t immediate; /* optional immediate operand */ 531 532 uint8_t decoded; /* set to 1 if successfully decoded */ 533 534 struct vie_op op; /* opcode description */ 535 }; 536 537 enum vm_exitcode { 538 VM_EXITCODE_INOUT, 539 VM_EXITCODE_VMX, 540 VM_EXITCODE_BOGUS, 541 VM_EXITCODE_RDMSR, 542 VM_EXITCODE_WRMSR, 543 VM_EXITCODE_HLT, 544 VM_EXITCODE_MTRAP, 545 VM_EXITCODE_PAUSE, 546 VM_EXITCODE_PAGING, 547 VM_EXITCODE_INST_EMUL, 548 VM_EXITCODE_SPINUP_AP, 549 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */ 550 VM_EXITCODE_RENDEZVOUS, 551 VM_EXITCODE_IOAPIC_EOI, 552 VM_EXITCODE_SUSPENDED, 553 VM_EXITCODE_INOUT_STR, 554 VM_EXITCODE_TASK_SWITCH, 555 VM_EXITCODE_MONITOR, 556 VM_EXITCODE_MWAIT, 557 VM_EXITCODE_SVM, 558 VM_EXITCODE_REQIDLE, 559 VM_EXITCODE_DEBUG, 560 VM_EXITCODE_VMINSN, 561 VM_EXITCODE_MAX 562 }; 563 564 struct vm_inout { 565 uint16_t bytes:3; /* 1 or 2 or 4 */ 566 uint16_t in:1; 567 uint16_t string:1; 568 uint16_t rep:1; 569 uint16_t port; 570 uint32_t eax; /* valid for out */ 571 }; 572 573 struct vm_inout_str { 574 struct vm_inout inout; /* must be the first element */ 575 struct vm_guest_paging paging; 576 uint64_t rflags; 577 uint64_t cr0; 578 uint64_t index; 579 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */ 580 int addrsize; 581 enum vm_reg_name seg_name; 582 struct seg_desc seg_desc; 583 }; 584 585 enum task_switch_reason { 586 TSR_CALL, 587 TSR_IRET, 588 TSR_JMP, 589 TSR_IDT_GATE, /* task gate in IDT */ 590 }; 591 592 struct vm_task_switch { 593 uint16_t tsssel; /* new TSS selector */ 594 int ext; /* task switch due to external event */ 595 uint32_t errcode; 596 int errcode_valid; /* push 'errcode' on the new stack */ 597 enum task_switch_reason reason; 598 struct vm_guest_paging paging; 599 }; 600 601 struct vm_exit { 602 enum vm_exitcode exitcode; 603 int inst_length; /* 0 means unknown */ 604 uint64_t rip; 605 union { 606 struct vm_inout inout; 607 struct vm_inout_str inout_str; 608 struct { 609 uint64_t gpa; 610 int fault_type; 611 } paging; 612 struct { 613 uint64_t gpa; 614 uint64_t gla; 615 uint64_t cs_base; 616 int cs_d; /* CS.D */ 617 struct vm_guest_paging paging; 618 struct vie vie; 619 } inst_emul; 620 /* 621 * VMX specific payload. Used when there is no "better" 622 * exitcode to represent the VM-exit. 623 */ 624 struct { 625 int status; /* vmx inst status */ 626 /* 627 * 'exit_reason' and 'exit_qualification' are valid 628 * only if 'status' is zero. 629 */ 630 uint32_t exit_reason; 631 uint64_t exit_qualification; 632 /* 633 * 'inst_error' and 'inst_type' are valid 634 * only if 'status' is non-zero. 635 */ 636 int inst_type; 637 int inst_error; 638 } vmx; 639 /* 640 * SVM specific payload. 641 */ 642 struct { 643 uint64_t exitcode; 644 uint64_t exitinfo1; 645 uint64_t exitinfo2; 646 } svm; 647 struct { 648 uint32_t code; /* ecx value */ 649 uint64_t wval; 650 } msr; 651 struct { 652 int vcpu; 653 uint64_t rip; 654 } spinup_ap; 655 struct { 656 uint64_t rflags; 657 uint64_t intr_status; 658 } hlt; 659 struct { 660 int vector; 661 } ioapic_eoi; 662 struct { 663 enum vm_suspend_how how; 664 } suspended; 665 struct vm_task_switch task_switch; 666 } u; 667 }; 668 669 /* APIs to inject faults into the guest */ 670 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid, 671 int errcode); 672 673 static __inline void 674 vm_inject_ud(void *vm, int vcpuid) 675 { 676 vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0); 677 } 678 679 static __inline void 680 vm_inject_gp(void *vm, int vcpuid) 681 { 682 vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0); 683 } 684 685 static __inline void 686 vm_inject_ac(void *vm, int vcpuid, int errcode) 687 { 688 vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode); 689 } 690 691 static __inline void 692 vm_inject_ss(void *vm, int vcpuid, int errcode) 693 { 694 vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode); 695 } 696 697 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2); 698 699 int vm_restart_instruction(void *vm, int vcpuid); 700 701 #endif /* _VMM_H_ */ 702