vmm.h revision 269008
1/*- 2 * Copyright (c) 2011 NetApp, Inc. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD: head/sys/amd64/include/vmm.h 269008 2014-07-23 04:28:51Z neel $ 27 */ 28 29#ifndef _VMM_H_ 30#define _VMM_H_ 31 32enum vm_suspend_how { 33 VM_SUSPEND_NONE, 34 VM_SUSPEND_RESET, 35 VM_SUSPEND_POWEROFF, 36 VM_SUSPEND_HALT, 37 VM_SUSPEND_TRIPLEFAULT, 38 VM_SUSPEND_LAST 39}; 40 41/* 42 * Identifiers for architecturally defined registers. 43 */ 44enum vm_reg_name { 45 VM_REG_GUEST_RAX, 46 VM_REG_GUEST_RBX, 47 VM_REG_GUEST_RCX, 48 VM_REG_GUEST_RDX, 49 VM_REG_GUEST_RSI, 50 VM_REG_GUEST_RDI, 51 VM_REG_GUEST_RBP, 52 VM_REG_GUEST_R8, 53 VM_REG_GUEST_R9, 54 VM_REG_GUEST_R10, 55 VM_REG_GUEST_R11, 56 VM_REG_GUEST_R12, 57 VM_REG_GUEST_R13, 58 VM_REG_GUEST_R14, 59 VM_REG_GUEST_R15, 60 VM_REG_GUEST_CR0, 61 VM_REG_GUEST_CR3, 62 VM_REG_GUEST_CR4, 63 VM_REG_GUEST_DR7, 64 VM_REG_GUEST_RSP, 65 VM_REG_GUEST_RIP, 66 VM_REG_GUEST_RFLAGS, 67 VM_REG_GUEST_ES, 68 VM_REG_GUEST_CS, 69 VM_REG_GUEST_SS, 70 VM_REG_GUEST_DS, 71 VM_REG_GUEST_FS, 72 VM_REG_GUEST_GS, 73 VM_REG_GUEST_LDTR, 74 VM_REG_GUEST_TR, 75 VM_REG_GUEST_IDTR, 76 VM_REG_GUEST_GDTR, 77 VM_REG_GUEST_EFER, 78 VM_REG_GUEST_CR2, 79 VM_REG_GUEST_PDPTE0, 80 VM_REG_GUEST_PDPTE1, 81 VM_REG_GUEST_PDPTE2, 82 VM_REG_GUEST_PDPTE3, 83 VM_REG_LAST 84}; 85 86enum x2apic_state { 87 X2APIC_DISABLED, 88 X2APIC_ENABLED, 89 X2APIC_STATE_LAST 90}; 91 92#define VM_INTINFO_VECTOR(info) ((info) & 0xff) 93#define VM_INTINFO_DEL_ERRCODE 0x800 94#define VM_INTINFO_RSVD 0x7ffff000 95#define VM_INTINFO_VALID 0x80000000 96#define VM_INTINFO_TYPE 0x700 97#define VM_INTINFO_HWINTR (0 << 8) 98#define VM_INTINFO_NMI (2 << 8) 99#define VM_INTINFO_HWEXCEPTION (3 << 8) 100#define VM_INTINFO_SWINTR (4 << 8) 101 102#ifdef _KERNEL 103 104#define VM_MAX_NAMELEN 32 105 106struct vm; 107struct vm_exception; 108struct vm_memory_segment; 109struct seg_desc; 110struct vm_exit; 111struct vm_run; 112struct vhpet; 113struct vioapic; 114struct vlapic; 115struct vmspace; 116struct vm_object; 117struct vm_guest_paging; 118struct pmap; 119 120typedef int (*vmm_init_func_t)(int ipinum); 121typedef int (*vmm_cleanup_func_t)(void); 122typedef void (*vmm_resume_func_t)(void); 123typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap); 124typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip, 125 struct pmap *pmap, void *rendezvous_cookie, 126 void *suspend_cookie); 127typedef void (*vmi_cleanup_func_t)(void *vmi); 128typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 129 uint64_t *retval); 130typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 131 uint64_t val); 132typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 133 struct seg_desc *desc); 134typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 135 struct seg_desc *desc); 136typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 137typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 138typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max); 139typedef void (*vmi_vmspace_free)(struct vmspace *vmspace); 140typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu); 141typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 142 143struct vmm_ops { 144 vmm_init_func_t init; /* module wide initialization */ 145 vmm_cleanup_func_t cleanup; 146 vmm_resume_func_t resume; 147 148 vmi_init_func_t vminit; /* vm-specific initialization */ 149 vmi_run_func_t vmrun; 150 vmi_cleanup_func_t vmcleanup; 151 vmi_get_register_t vmgetreg; 152 vmi_set_register_t vmsetreg; 153 vmi_get_desc_t vmgetdesc; 154 vmi_set_desc_t vmsetdesc; 155 vmi_get_cap_t vmgetcap; 156 vmi_set_cap_t vmsetcap; 157 vmi_vmspace_alloc vmspace_alloc; 158 vmi_vmspace_free vmspace_free; 159 vmi_vlapic_init vlapic_init; 160 vmi_vlapic_cleanup vlapic_cleanup; 161}; 162 163extern struct vmm_ops vmm_ops_intel; 164extern struct vmm_ops vmm_ops_amd; 165 166int vm_create(const char *name, struct vm **retvm); 167void vm_destroy(struct vm *vm); 168int vm_reinit(struct vm *vm); 169const char *vm_name(struct vm *vm); 170int vm_malloc(struct vm *vm, vm_paddr_t gpa, size_t len); 171int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 172int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 173void *vm_gpa_hold(struct vm *, vm_paddr_t gpa, size_t len, int prot, 174 void **cookie); 175void vm_gpa_release(void *cookie); 176int vm_gpabase2memseg(struct vm *vm, vm_paddr_t gpabase, 177 struct vm_memory_segment *seg); 178int vm_get_memobj(struct vm *vm, vm_paddr_t gpa, size_t len, 179 vm_offset_t *offset, struct vm_object **object); 180boolean_t vm_mem_allocated(struct vm *vm, vm_paddr_t gpa); 181int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 182int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 183int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 184 struct seg_desc *ret_desc); 185int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 186 struct seg_desc *desc); 187int vm_run(struct vm *vm, struct vm_run *vmrun); 188int vm_suspend(struct vm *vm, enum vm_suspend_how how); 189int vm_inject_nmi(struct vm *vm, int vcpu); 190int vm_nmi_pending(struct vm *vm, int vcpuid); 191void vm_nmi_clear(struct vm *vm, int vcpuid); 192int vm_inject_extint(struct vm *vm, int vcpu); 193int vm_extint_pending(struct vm *vm, int vcpuid); 194void vm_extint_clear(struct vm *vm, int vcpuid); 195uint64_t *vm_guest_msrs(struct vm *vm, int cpu); 196struct vlapic *vm_lapic(struct vm *vm, int cpu); 197struct vioapic *vm_ioapic(struct vm *vm); 198struct vhpet *vm_hpet(struct vm *vm); 199int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 200int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 201int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 202int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 203int vm_apicid2vcpuid(struct vm *vm, int apicid); 204int vm_activate_cpu(struct vm *vm, int vcpu); 205cpuset_t vm_active_cpus(struct vm *vm); 206cpuset_t vm_suspended_cpus(struct vm *vm); 207struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 208void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 209void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip); 210void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip); 211 212/* 213 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'. 214 * The rendezvous 'func(arg)' is not allowed to do anything that will 215 * cause the thread to be put to sleep. 216 * 217 * If the rendezvous is being initiated from a vcpu context then the 218 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1. 219 * 220 * The caller cannot hold any locks when initiating the rendezvous. 221 * 222 * The implementation of this API may cause vcpus other than those specified 223 * by 'dest' to be stalled. The caller should not rely on any vcpus making 224 * forward progress when the rendezvous is in progress. 225 */ 226typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg); 227void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest, 228 vm_rendezvous_func_t func, void *arg); 229 230static __inline int 231vcpu_rendezvous_pending(void *rendezvous_cookie) 232{ 233 234 return (*(uintptr_t *)rendezvous_cookie != 0); 235} 236 237static __inline int 238vcpu_suspended(void *suspend_cookie) 239{ 240 241 return (*(int *)suspend_cookie); 242} 243 244/* 245 * Return 1 if device indicated by bus/slot/func is supposed to be a 246 * pci passthrough device. 247 * 248 * Return 0 otherwise. 249 */ 250int vmm_is_pptdev(int bus, int slot, int func); 251 252void *vm_iommu_domain(struct vm *vm); 253 254enum vcpu_state { 255 VCPU_IDLE, 256 VCPU_FROZEN, 257 VCPU_RUNNING, 258 VCPU_SLEEPING, 259}; 260 261int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 262 bool from_idle); 263enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 264 265static int __inline 266vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 267{ 268 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 269} 270 271void *vcpu_stats(struct vm *vm, int vcpu); 272void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr); 273struct vmspace *vm_get_vmspace(struct vm *vm); 274int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func); 275int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func); 276struct vatpic *vm_atpic(struct vm *vm); 277struct vatpit *vm_atpit(struct vm *vm); 278 279/* 280 * Inject exception 'vme' into the guest vcpu. This function returns 0 on 281 * success and non-zero on failure. 282 * 283 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 284 * this function directly because they enforce the trap-like or fault-like 285 * behavior of an exception. 286 * 287 * This function should only be called in the context of the thread that is 288 * executing this vcpu. 289 */ 290int vm_inject_exception(struct vm *vm, int vcpuid, struct vm_exception *vme); 291 292/* 293 * This function is called after a VM-exit that occurred during exception or 294 * interrupt delivery through the IDT. The format of 'intinfo' is described 295 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. 296 * 297 * If a VM-exit handler completes the event delivery successfully then it 298 * should call vm_exit_intinfo() to extinguish the pending event. For e.g., 299 * if the task switch emulation is triggered via a task gate then it should 300 * call this function with 'intinfo=0' to indicate that the external event 301 * is not pending anymore. 302 * 303 * Return value is 0 on success and non-zero on failure. 304 */ 305int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo); 306 307/* 308 * This function is called before every VM-entry to retrieve a pending 309 * event that should be injected into the guest. This function combines 310 * nested events into a double or triple fault. 311 * 312 * Returns 0 if there are no events that need to be injected into the guest 313 * and non-zero otherwise. 314 */ 315int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info); 316 317int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2); 318 319void vm_inject_gp(struct vm *vm, int vcpuid); /* general protection fault */ 320void vm_inject_ud(struct vm *vm, int vcpuid); /* undefined instruction fault */ 321void vm_inject_ac(struct vm *vm, int vcpuid, int errcode); /* #AC */ 322void vm_inject_ss(struct vm *vm, int vcpuid, int errcode); /* #SS */ 323void vm_inject_pf(struct vm *vm, int vcpuid, int error_code, uint64_t cr2); 324 325enum vm_reg_name vm_segment_name(int seg_encoding); 326 327struct vm_copyinfo { 328 uint64_t gpa; 329 size_t len; 330 void *hva; 331 void *cookie; 332}; 333 334/* 335 * Set up 'copyinfo[]' to copy to/from guest linear address space starting 336 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for 337 * a copyin or PROT_WRITE for a copyout. 338 * 339 * Returns 0 on success. 340 * Returns 1 if an exception was injected into the guest. 341 * Returns -1 otherwise. 342 * 343 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if 344 * the return value is 0. The 'copyinfo[]' resources should be freed by calling 345 * 'vm_copy_teardown()' after the copy is done. 346 */ 347int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging, 348 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo, 349 int num_copyinfo); 350void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 351 int num_copyinfo); 352void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 353 void *kaddr, size_t len); 354void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr, 355 struct vm_copyinfo *copyinfo, size_t len); 356#endif /* KERNEL */ 357 358#define VM_MAXCPU 16 /* maximum virtual cpus */ 359 360/* 361 * Identifiers for optional vmm capabilities 362 */ 363enum vm_cap_type { 364 VM_CAP_HALT_EXIT, 365 VM_CAP_MTRAP_EXIT, 366 VM_CAP_PAUSE_EXIT, 367 VM_CAP_UNRESTRICTED_GUEST, 368 VM_CAP_ENABLE_INVPCID, 369 VM_CAP_MAX 370}; 371 372enum vm_intr_trigger { 373 EDGE_TRIGGER, 374 LEVEL_TRIGGER 375}; 376 377/* 378 * The 'access' field has the format specified in Table 21-2 of the Intel 379 * Architecture Manual vol 3b. 380 * 381 * XXX The contents of the 'access' field are architecturally defined except 382 * bit 16 - Segment Unusable. 383 */ 384struct seg_desc { 385 uint64_t base; 386 uint32_t limit; 387 uint32_t access; 388}; 389#define SEG_DESC_TYPE(access) ((access) & 0x001f) 390#define SEG_DESC_DPL(access) (((access) >> 5) & 0x3) 391#define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0) 392#define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0) 393#define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0) 394#define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0) 395 396enum vm_cpu_mode { 397 CPU_MODE_REAL, 398 CPU_MODE_PROTECTED, 399 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */ 400 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */ 401}; 402 403enum vm_paging_mode { 404 PAGING_MODE_FLAT, 405 PAGING_MODE_32, 406 PAGING_MODE_PAE, 407 PAGING_MODE_64, 408}; 409 410struct vm_guest_paging { 411 uint64_t cr3; 412 int cpl; 413 enum vm_cpu_mode cpu_mode; 414 enum vm_paging_mode paging_mode; 415}; 416 417/* 418 * The data structures 'vie' and 'vie_op' are meant to be opaque to the 419 * consumers of instruction decoding. The only reason why their contents 420 * need to be exposed is because they are part of the 'vm_exit' structure. 421 */ 422struct vie_op { 423 uint8_t op_byte; /* actual opcode byte */ 424 uint8_t op_type; /* type of operation (e.g. MOV) */ 425 uint16_t op_flags; 426}; 427 428#define VIE_INST_SIZE 15 429struct vie { 430 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */ 431 uint8_t num_valid; /* size of the instruction */ 432 uint8_t num_processed; 433 434 uint8_t addrsize:4, opsize:4; /* address and operand sizes */ 435 uint8_t rex_w:1, /* REX prefix */ 436 rex_r:1, 437 rex_x:1, 438 rex_b:1, 439 rex_present:1, 440 opsize_override:1, /* Operand size override */ 441 addrsize_override:1; /* Address size override */ 442 443 uint8_t mod:2, /* ModRM byte */ 444 reg:4, 445 rm:4; 446 447 uint8_t ss:2, /* SIB byte */ 448 index:4, 449 base:4; 450 451 uint8_t disp_bytes; 452 uint8_t imm_bytes; 453 454 uint8_t scale; 455 int base_register; /* VM_REG_GUEST_xyz */ 456 int index_register; /* VM_REG_GUEST_xyz */ 457 458 int64_t displacement; /* optional addr displacement */ 459 int64_t immediate; /* optional immediate operand */ 460 461 uint8_t decoded; /* set to 1 if successfully decoded */ 462 463 struct vie_op op; /* opcode description */ 464}; 465 466enum vm_exitcode { 467 VM_EXITCODE_INOUT, 468 VM_EXITCODE_VMX, 469 VM_EXITCODE_BOGUS, 470 VM_EXITCODE_RDMSR, 471 VM_EXITCODE_WRMSR, 472 VM_EXITCODE_HLT, 473 VM_EXITCODE_MTRAP, 474 VM_EXITCODE_PAUSE, 475 VM_EXITCODE_PAGING, 476 VM_EXITCODE_INST_EMUL, 477 VM_EXITCODE_SPINUP_AP, 478 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */ 479 VM_EXITCODE_RENDEZVOUS, 480 VM_EXITCODE_IOAPIC_EOI, 481 VM_EXITCODE_SUSPENDED, 482 VM_EXITCODE_INOUT_STR, 483 VM_EXITCODE_TASK_SWITCH, 484 VM_EXITCODE_MAX 485}; 486 487struct vm_inout { 488 uint16_t bytes:3; /* 1 or 2 or 4 */ 489 uint16_t in:1; 490 uint16_t string:1; 491 uint16_t rep:1; 492 uint16_t port; 493 uint32_t eax; /* valid for out */ 494}; 495 496struct vm_inout_str { 497 struct vm_inout inout; /* must be the first element */ 498 struct vm_guest_paging paging; 499 uint64_t rflags; 500 uint64_t cr0; 501 uint64_t index; 502 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */ 503 int addrsize; 504 enum vm_reg_name seg_name; 505 struct seg_desc seg_desc; 506}; 507 508enum task_switch_reason { 509 TSR_CALL, 510 TSR_IRET, 511 TSR_JMP, 512 TSR_IDT_GATE, /* task gate in IDT */ 513}; 514 515struct vm_task_switch { 516 uint16_t tsssel; /* new TSS selector */ 517 int ext; /* task switch due to external event */ 518 uint32_t errcode; 519 int errcode_valid; /* push 'errcode' on the new stack */ 520 enum task_switch_reason reason; 521 struct vm_guest_paging paging; 522}; 523 524struct vm_exit { 525 enum vm_exitcode exitcode; 526 int inst_length; /* 0 means unknown */ 527 uint64_t rip; 528 union { 529 struct vm_inout inout; 530 struct vm_inout_str inout_str; 531 struct { 532 uint64_t gpa; 533 int fault_type; 534 } paging; 535 struct { 536 uint64_t gpa; 537 uint64_t gla; 538 int cs_d; /* CS.D */ 539 struct vm_guest_paging paging; 540 struct vie vie; 541 } inst_emul; 542 /* 543 * VMX specific payload. Used when there is no "better" 544 * exitcode to represent the VM-exit. 545 */ 546 struct { 547 int status; /* vmx inst status */ 548 /* 549 * 'exit_reason' and 'exit_qualification' are valid 550 * only if 'status' is zero. 551 */ 552 uint32_t exit_reason; 553 uint64_t exit_qualification; 554 /* 555 * 'inst_error' and 'inst_type' are valid 556 * only if 'status' is non-zero. 557 */ 558 int inst_type; 559 int inst_error; 560 } vmx; 561 struct { 562 uint32_t code; /* ecx value */ 563 uint64_t wval; 564 } msr; 565 struct { 566 int vcpu; 567 uint64_t rip; 568 } spinup_ap; 569 struct { 570 uint64_t rflags; 571 } hlt; 572 struct { 573 int vector; 574 } ioapic_eoi; 575 struct { 576 enum vm_suspend_how how; 577 } suspended; 578 struct vm_task_switch task_switch; 579 } u; 580}; 581 582#endif /* _VMM_H_ */ 583