vmm.h revision 268889
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 268889 2014-07-19 20:59:08Z 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 pmap; 118 119typedef int (*vmm_init_func_t)(int ipinum); 120typedef int (*vmm_cleanup_func_t)(void); 121typedef void (*vmm_resume_func_t)(void); 122typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap); 123typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip, 124 struct pmap *pmap, void *rendezvous_cookie, 125 void *suspend_cookie); 126typedef void (*vmi_cleanup_func_t)(void *vmi); 127typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 128 uint64_t *retval); 129typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 130 uint64_t val); 131typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 132 struct seg_desc *desc); 133typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 134 struct seg_desc *desc); 135typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 136typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 137typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max); 138typedef void (*vmi_vmspace_free)(struct vmspace *vmspace); 139typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu); 140typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 141 142struct vmm_ops { 143 vmm_init_func_t init; /* module wide initialization */ 144 vmm_cleanup_func_t cleanup; 145 vmm_resume_func_t resume; 146 147 vmi_init_func_t vminit; /* vm-specific initialization */ 148 vmi_run_func_t vmrun; 149 vmi_cleanup_func_t vmcleanup; 150 vmi_get_register_t vmgetreg; 151 vmi_set_register_t vmsetreg; 152 vmi_get_desc_t vmgetdesc; 153 vmi_set_desc_t vmsetdesc; 154 vmi_get_cap_t vmgetcap; 155 vmi_set_cap_t vmsetcap; 156 vmi_vmspace_alloc vmspace_alloc; 157 vmi_vmspace_free vmspace_free; 158 vmi_vlapic_init vlapic_init; 159 vmi_vlapic_cleanup vlapic_cleanup; 160}; 161 162extern struct vmm_ops vmm_ops_intel; 163extern struct vmm_ops vmm_ops_amd; 164 165int vm_create(const char *name, struct vm **retvm); 166void vm_destroy(struct vm *vm); 167int vm_reinit(struct vm *vm); 168const char *vm_name(struct vm *vm); 169int vm_malloc(struct vm *vm, vm_paddr_t gpa, size_t len); 170int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 171int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 172void *vm_gpa_hold(struct vm *, vm_paddr_t gpa, size_t len, int prot, 173 void **cookie); 174void vm_gpa_release(void *cookie); 175int vm_gpabase2memseg(struct vm *vm, vm_paddr_t gpabase, 176 struct vm_memory_segment *seg); 177int vm_get_memobj(struct vm *vm, vm_paddr_t gpa, size_t len, 178 vm_offset_t *offset, struct vm_object **object); 179boolean_t vm_mem_allocated(struct vm *vm, vm_paddr_t gpa); 180int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 181int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 182int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 183 struct seg_desc *ret_desc); 184int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 185 struct seg_desc *desc); 186int vm_run(struct vm *vm, struct vm_run *vmrun); 187int vm_suspend(struct vm *vm, enum vm_suspend_how how); 188int vm_inject_nmi(struct vm *vm, int vcpu); 189int vm_nmi_pending(struct vm *vm, int vcpuid); 190void vm_nmi_clear(struct vm *vm, int vcpuid); 191int vm_inject_extint(struct vm *vm, int vcpu); 192int vm_extint_pending(struct vm *vm, int vcpuid); 193void vm_extint_clear(struct vm *vm, int vcpuid); 194uint64_t *vm_guest_msrs(struct vm *vm, int cpu); 195struct vlapic *vm_lapic(struct vm *vm, int cpu); 196struct vioapic *vm_ioapic(struct vm *vm); 197struct vhpet *vm_hpet(struct vm *vm); 198int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 199int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 200int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 201int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 202int vm_apicid2vcpuid(struct vm *vm, int apicid); 203int vm_activate_cpu(struct vm *vm, int vcpu); 204cpuset_t vm_active_cpus(struct vm *vm); 205cpuset_t vm_suspended_cpus(struct vm *vm); 206struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 207void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 208void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip); 209void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip); 210 211/* 212 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'. 213 * The rendezvous 'func(arg)' is not allowed to do anything that will 214 * cause the thread to be put to sleep. 215 * 216 * If the rendezvous is being initiated from a vcpu context then the 217 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1. 218 * 219 * The caller cannot hold any locks when initiating the rendezvous. 220 * 221 * The implementation of this API may cause vcpus other than those specified 222 * by 'dest' to be stalled. The caller should not rely on any vcpus making 223 * forward progress when the rendezvous is in progress. 224 */ 225typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg); 226void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest, 227 vm_rendezvous_func_t func, void *arg); 228 229static __inline int 230vcpu_rendezvous_pending(void *rendezvous_cookie) 231{ 232 233 return (*(uintptr_t *)rendezvous_cookie != 0); 234} 235 236static __inline int 237vcpu_suspended(void *suspend_cookie) 238{ 239 240 return (*(int *)suspend_cookie); 241} 242 243/* 244 * Return 1 if device indicated by bus/slot/func is supposed to be a 245 * pci passthrough device. 246 * 247 * Return 0 otherwise. 248 */ 249int vmm_is_pptdev(int bus, int slot, int func); 250 251void *vm_iommu_domain(struct vm *vm); 252 253enum vcpu_state { 254 VCPU_IDLE, 255 VCPU_FROZEN, 256 VCPU_RUNNING, 257 VCPU_SLEEPING, 258}; 259 260int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 261 bool from_idle); 262enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 263 264static int __inline 265vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 266{ 267 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 268} 269 270void *vcpu_stats(struct vm *vm, int vcpu); 271void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr); 272struct vmspace *vm_get_vmspace(struct vm *vm); 273int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func); 274int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func); 275struct vatpic *vm_atpic(struct vm *vm); 276struct vatpit *vm_atpit(struct vm *vm); 277 278/* 279 * Inject exception 'vme' into the guest vcpu. This function returns 0 on 280 * success and non-zero on failure. 281 * 282 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 283 * this function directly because they enforce the trap-like or fault-like 284 * behavior of an exception. 285 * 286 * This function should only be called in the context of the thread that is 287 * executing this vcpu. 288 */ 289int vm_inject_exception(struct vm *vm, int vcpuid, struct vm_exception *vme); 290 291/* 292 * This function is called after a VM-exit that occurred during exception or 293 * interrupt delivery through the IDT. The format of 'intinfo' is described 294 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. 295 * 296 * If a VM-exit handler completes the event delivery successfully then it 297 * should call vm_exit_intinfo() to extinguish the pending event. For e.g., 298 * if the task switch emulation is triggered via a task gate then it should 299 * call this function with 'intinfo=0' to indicate that the external event 300 * is not pending anymore. 301 * 302 * Return value is 0 on success and non-zero on failure. 303 */ 304int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo); 305 306/* 307 * This function is called before every VM-entry to retrieve a pending 308 * event that should be injected into the guest. This function combines 309 * nested events into a double or triple fault. 310 * 311 * Returns 0 if there are no events that need to be injected into the guest 312 * and non-zero otherwise. 313 */ 314int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info); 315 316int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2); 317 318void vm_inject_gp(struct vm *vm, int vcpuid); /* general protection fault */ 319void vm_inject_ud(struct vm *vm, int vcpuid); /* undefined instruction fault */ 320void vm_inject_pf(struct vm *vm, int vcpuid, int error_code, uint64_t cr2); 321 322enum vm_reg_name vm_segment_name(int seg_encoding); 323 324#endif /* KERNEL */ 325 326#define VM_MAXCPU 16 /* maximum virtual cpus */ 327 328/* 329 * Identifiers for optional vmm capabilities 330 */ 331enum vm_cap_type { 332 VM_CAP_HALT_EXIT, 333 VM_CAP_MTRAP_EXIT, 334 VM_CAP_PAUSE_EXIT, 335 VM_CAP_UNRESTRICTED_GUEST, 336 VM_CAP_ENABLE_INVPCID, 337 VM_CAP_MAX 338}; 339 340enum vm_intr_trigger { 341 EDGE_TRIGGER, 342 LEVEL_TRIGGER 343}; 344 345/* 346 * The 'access' field has the format specified in Table 21-2 of the Intel 347 * Architecture Manual vol 3b. 348 * 349 * XXX The contents of the 'access' field are architecturally defined except 350 * bit 16 - Segment Unusable. 351 */ 352struct seg_desc { 353 uint64_t base; 354 uint32_t limit; 355 uint32_t access; 356}; 357#define SEG_DESC_TYPE(access) ((access) & 0x001f) 358#define SEG_DESC_DPL(access) (((access) >> 5) & 0x3) 359#define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0) 360#define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0) 361#define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0) 362#define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0) 363 364enum vm_cpu_mode { 365 CPU_MODE_REAL, 366 CPU_MODE_PROTECTED, 367 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */ 368 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */ 369}; 370 371enum vm_paging_mode { 372 PAGING_MODE_FLAT, 373 PAGING_MODE_32, 374 PAGING_MODE_PAE, 375 PAGING_MODE_64, 376}; 377 378struct vm_guest_paging { 379 uint64_t cr3; 380 int cpl; 381 enum vm_cpu_mode cpu_mode; 382 enum vm_paging_mode paging_mode; 383}; 384 385/* 386 * The data structures 'vie' and 'vie_op' are meant to be opaque to the 387 * consumers of instruction decoding. The only reason why their contents 388 * need to be exposed is because they are part of the 'vm_exit' structure. 389 */ 390struct vie_op { 391 uint8_t op_byte; /* actual opcode byte */ 392 uint8_t op_type; /* type of operation (e.g. MOV) */ 393 uint16_t op_flags; 394}; 395 396#define VIE_INST_SIZE 15 397struct vie { 398 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */ 399 uint8_t num_valid; /* size of the instruction */ 400 uint8_t num_processed; 401 402 uint8_t addrsize:4, opsize:4; /* address and operand sizes */ 403 uint8_t rex_w:1, /* REX prefix */ 404 rex_r:1, 405 rex_x:1, 406 rex_b:1, 407 rex_present:1, 408 opsize_override:1, /* Operand size override */ 409 addrsize_override:1; /* Address size override */ 410 411 uint8_t mod:2, /* ModRM byte */ 412 reg:4, 413 rm:4; 414 415 uint8_t ss:2, /* SIB byte */ 416 index:4, 417 base:4; 418 419 uint8_t disp_bytes; 420 uint8_t imm_bytes; 421 422 uint8_t scale; 423 int base_register; /* VM_REG_GUEST_xyz */ 424 int index_register; /* VM_REG_GUEST_xyz */ 425 426 int64_t displacement; /* optional addr displacement */ 427 int64_t immediate; /* optional immediate operand */ 428 429 uint8_t decoded; /* set to 1 if successfully decoded */ 430 431 struct vie_op op; /* opcode description */ 432}; 433 434enum vm_exitcode { 435 VM_EXITCODE_INOUT, 436 VM_EXITCODE_VMX, 437 VM_EXITCODE_BOGUS, 438 VM_EXITCODE_RDMSR, 439 VM_EXITCODE_WRMSR, 440 VM_EXITCODE_HLT, 441 VM_EXITCODE_MTRAP, 442 VM_EXITCODE_PAUSE, 443 VM_EXITCODE_PAGING, 444 VM_EXITCODE_INST_EMUL, 445 VM_EXITCODE_SPINUP_AP, 446 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */ 447 VM_EXITCODE_RENDEZVOUS, 448 VM_EXITCODE_IOAPIC_EOI, 449 VM_EXITCODE_SUSPENDED, 450 VM_EXITCODE_INOUT_STR, 451 VM_EXITCODE_TASK_SWITCH, 452 VM_EXITCODE_MAX 453}; 454 455struct vm_inout { 456 uint16_t bytes:3; /* 1 or 2 or 4 */ 457 uint16_t in:1; 458 uint16_t string:1; 459 uint16_t rep:1; 460 uint16_t port; 461 uint32_t eax; /* valid for out */ 462}; 463 464struct vm_inout_str { 465 struct vm_inout inout; /* must be the first element */ 466 struct vm_guest_paging paging; 467 uint64_t rflags; 468 uint64_t cr0; 469 uint64_t index; 470 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */ 471 int addrsize; 472 enum vm_reg_name seg_name; 473 struct seg_desc seg_desc; 474}; 475 476enum task_switch_reason { 477 TSR_CALL, 478 TSR_IRET, 479 TSR_JMP, 480 TSR_IDT_GATE, /* task gate in IDT */ 481}; 482 483struct vm_task_switch { 484 uint16_t tsssel; /* new TSS selector */ 485 int ext; /* task switch due to external event */ 486 uint32_t errcode; 487 int errcode_valid; /* push 'errcode' on the new stack */ 488 enum task_switch_reason reason; 489 struct vm_guest_paging paging; 490}; 491 492struct vm_exit { 493 enum vm_exitcode exitcode; 494 int inst_length; /* 0 means unknown */ 495 uint64_t rip; 496 union { 497 struct vm_inout inout; 498 struct vm_inout_str inout_str; 499 struct { 500 uint64_t gpa; 501 int fault_type; 502 } paging; 503 struct { 504 uint64_t gpa; 505 uint64_t gla; 506 int cs_d; /* CS.D */ 507 struct vm_guest_paging paging; 508 struct vie vie; 509 } inst_emul; 510 /* 511 * VMX specific payload. Used when there is no "better" 512 * exitcode to represent the VM-exit. 513 */ 514 struct { 515 int status; /* vmx inst status */ 516 /* 517 * 'exit_reason' and 'exit_qualification' are valid 518 * only if 'status' is zero. 519 */ 520 uint32_t exit_reason; 521 uint64_t exit_qualification; 522 /* 523 * 'inst_error' and 'inst_type' are valid 524 * only if 'status' is non-zero. 525 */ 526 int inst_type; 527 int inst_error; 528 } vmx; 529 struct { 530 uint32_t code; /* ecx value */ 531 uint64_t wval; 532 } msr; 533 struct { 534 int vcpu; 535 uint64_t rip; 536 } spinup_ap; 537 struct { 538 uint64_t rflags; 539 } hlt; 540 struct { 541 int vector; 542 } ioapic_eoi; 543 struct { 544 enum vm_suspend_how how; 545 } suspended; 546 struct vm_task_switch task_switch; 547 } u; 548}; 549 550#endif /* _VMM_H_ */ 551