vmm.h revision 266933
1169695Skan/*- 2169695Skan * Copyright (c) 2011 NetApp, Inc. 3169695Skan * All rights reserved. 4169695Skan * 5169695Skan * Redistribution and use in source and binary forms, with or without 6169695Skan * modification, are permitted provided that the following conditions 7169695Skan * are met: 8169695Skan * 1. Redistributions of source code must retain the above copyright 9169695Skan * 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 266933 2014-05-31 23:37:34Z 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_LAST 38}; 39 40#ifdef _KERNEL 41 42#define VM_MAX_NAMELEN 32 43 44struct vm; 45struct vm_exception; 46struct vm_memory_segment; 47struct seg_desc; 48struct vm_exit; 49struct vm_run; 50struct vhpet; 51struct vioapic; 52struct vlapic; 53struct vmspace; 54struct vm_object; 55struct pmap; 56 57enum vm_reg_name; 58enum x2apic_state; 59 60typedef int (*vmm_init_func_t)(int ipinum); 61typedef int (*vmm_cleanup_func_t)(void); 62typedef void (*vmm_resume_func_t)(void); 63typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap); 64typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip, 65 struct pmap *pmap, void *rendezvous_cookie, 66 void *suspend_cookie); 67typedef void (*vmi_cleanup_func_t)(void *vmi); 68typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 69 uint64_t *retval); 70typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 71 uint64_t val); 72typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 73 struct seg_desc *desc); 74typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 75 struct seg_desc *desc); 76typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 77typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 78typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max); 79typedef void (*vmi_vmspace_free)(struct vmspace *vmspace); 80typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu); 81typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 82 83struct vmm_ops { 84 vmm_init_func_t init; /* module wide initialization */ 85 vmm_cleanup_func_t cleanup; 86 vmm_resume_func_t resume; 87 88 vmi_init_func_t vminit; /* vm-specific initialization */ 89 vmi_run_func_t vmrun; 90 vmi_cleanup_func_t vmcleanup; 91 vmi_get_register_t vmgetreg; 92 vmi_set_register_t vmsetreg; 93 vmi_get_desc_t vmgetdesc; 94 vmi_set_desc_t vmsetdesc; 95 vmi_get_cap_t vmgetcap; 96 vmi_set_cap_t vmsetcap; 97 vmi_vmspace_alloc vmspace_alloc; 98 vmi_vmspace_free vmspace_free; 99 vmi_vlapic_init vlapic_init; 100 vmi_vlapic_cleanup vlapic_cleanup; 101}; 102 103extern struct vmm_ops vmm_ops_intel; 104extern struct vmm_ops vmm_ops_amd; 105 106int vm_create(const char *name, struct vm **retvm); 107void vm_destroy(struct vm *vm); 108const char *vm_name(struct vm *vm); 109int vm_malloc(struct vm *vm, vm_paddr_t gpa, size_t len); 110int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 111int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 112void *vm_gpa_hold(struct vm *, vm_paddr_t gpa, size_t len, int prot, 113 void **cookie); 114void vm_gpa_release(void *cookie); 115int vm_gpabase2memseg(struct vm *vm, vm_paddr_t gpabase, 116 struct vm_memory_segment *seg); 117int vm_get_memobj(struct vm *vm, vm_paddr_t gpa, size_t len, 118 vm_offset_t *offset, struct vm_object **object); 119boolean_t vm_mem_allocated(struct vm *vm, vm_paddr_t gpa); 120int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 121int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 122int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 123 struct seg_desc *ret_desc); 124int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 125 struct seg_desc *desc); 126int vm_run(struct vm *vm, struct vm_run *vmrun); 127int vm_suspend(struct vm *vm, enum vm_suspend_how how); 128int vm_inject_nmi(struct vm *vm, int vcpu); 129int vm_nmi_pending(struct vm *vm, int vcpuid); 130void vm_nmi_clear(struct vm *vm, int vcpuid); 131int vm_inject_extint(struct vm *vm, int vcpu); 132int vm_extint_pending(struct vm *vm, int vcpuid); 133void vm_extint_clear(struct vm *vm, int vcpuid); 134uint64_t *vm_guest_msrs(struct vm *vm, int cpu); 135struct vlapic *vm_lapic(struct vm *vm, int cpu); 136struct vioapic *vm_ioapic(struct vm *vm); 137struct vhpet *vm_hpet(struct vm *vm); 138int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 139int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 140int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 141int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 142int vm_apicid2vcpuid(struct vm *vm, int apicid); 143int vm_activate_cpu(struct vm *vm, int vcpu); 144cpuset_t vm_active_cpus(struct vm *vm); 145cpuset_t vm_suspended_cpus(struct vm *vm); 146struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 147void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 148 149/* 150 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'. 151 * The rendezvous 'func(arg)' is not allowed to do anything that will 152 * cause the thread to be put to sleep. 153 * 154 * If the rendezvous is being initiated from a vcpu context then the 155 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1. 156 * 157 * The caller cannot hold any locks when initiating the rendezvous. 158 * 159 * The implementation of this API may cause vcpus other than those specified 160 * by 'dest' to be stalled. The caller should not rely on any vcpus making 161 * forward progress when the rendezvous is in progress. 162 */ 163typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg); 164void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest, 165 vm_rendezvous_func_t func, void *arg); 166 167static __inline int 168vcpu_rendezvous_pending(void *rendezvous_cookie) 169{ 170 171 return (*(uintptr_t *)rendezvous_cookie != 0); 172} 173 174static __inline int 175vcpu_suspended(void *suspend_cookie) 176{ 177 178 return (*(int *)suspend_cookie); 179} 180 181/* 182 * Return 1 if device indicated by bus/slot/func is supposed to be a 183 * pci passthrough device. 184 * 185 * Return 0 otherwise. 186 */ 187int vmm_is_pptdev(int bus, int slot, int func); 188 189void *vm_iommu_domain(struct vm *vm); 190 191enum vcpu_state { 192 VCPU_IDLE, 193 VCPU_FROZEN, 194 VCPU_RUNNING, 195 VCPU_SLEEPING, 196}; 197 198int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 199 bool from_idle); 200enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 201 202static int __inline 203vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 204{ 205 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 206} 207 208void *vcpu_stats(struct vm *vm, int vcpu); 209void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr); 210struct vmspace *vm_get_vmspace(struct vm *vm); 211int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func); 212int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func); 213struct vatpic *vm_atpic(struct vm *vm); 214struct vatpit *vm_atpit(struct vm *vm); 215 216/* 217 * Inject exception 'vme' into the guest vcpu. This function returns 0 on 218 * success and non-zero on failure. 219 * 220 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 221 * this function directly because they enforce the trap-like or fault-like 222 * behavior of an exception. 223 * 224 * This function should only be called in the context of the thread that is 225 * executing this vcpu. 226 */ 227int vm_inject_exception(struct vm *vm, int vcpuid, struct vm_exception *vme); 228 229/* 230 * Returns 0 if there is no exception pending for this vcpu. Returns 1 if an 231 * exception is pending and also updates 'vme'. The pending exception is 232 * cleared when this function returns. 233 * 234 * This function should only be called in the context of the thread that is 235 * executing this vcpu. 236 */ 237int vm_exception_pending(struct vm *vm, int vcpuid, struct vm_exception *vme); 238 239void vm_inject_gp(struct vm *vm, int vcpuid); /* general protection fault */ 240void vm_inject_ud(struct vm *vm, int vcpuid); /* undefined instruction fault */ 241void vm_inject_pf(struct vm *vm, int vcpuid, int error_code, uint64_t cr2); 242 243enum vm_reg_name vm_segment_name(int seg_encoding); 244 245#endif /* KERNEL */ 246 247#define VM_MAXCPU 16 /* maximum virtual cpus */ 248 249/* 250 * Identifiers for architecturally defined registers. 251 */ 252enum vm_reg_name { 253 VM_REG_GUEST_RAX, 254 VM_REG_GUEST_RBX, 255 VM_REG_GUEST_RCX, 256 VM_REG_GUEST_RDX, 257 VM_REG_GUEST_RSI, 258 VM_REG_GUEST_RDI, 259 VM_REG_GUEST_RBP, 260 VM_REG_GUEST_R8, 261 VM_REG_GUEST_R9, 262 VM_REG_GUEST_R10, 263 VM_REG_GUEST_R11, 264 VM_REG_GUEST_R12, 265 VM_REG_GUEST_R13, 266 VM_REG_GUEST_R14, 267 VM_REG_GUEST_R15, 268 VM_REG_GUEST_CR0, 269 VM_REG_GUEST_CR3, 270 VM_REG_GUEST_CR4, 271 VM_REG_GUEST_DR7, 272 VM_REG_GUEST_RSP, 273 VM_REG_GUEST_RIP, 274 VM_REG_GUEST_RFLAGS, 275 VM_REG_GUEST_ES, 276 VM_REG_GUEST_CS, 277 VM_REG_GUEST_SS, 278 VM_REG_GUEST_DS, 279 VM_REG_GUEST_FS, 280 VM_REG_GUEST_GS, 281 VM_REG_GUEST_LDTR, 282 VM_REG_GUEST_TR, 283 VM_REG_GUEST_IDTR, 284 VM_REG_GUEST_GDTR, 285 VM_REG_GUEST_EFER, 286 VM_REG_GUEST_CR2, 287 VM_REG_LAST 288}; 289 290/* 291 * Identifiers for optional vmm capabilities 292 */ 293enum vm_cap_type { 294 VM_CAP_HALT_EXIT, 295 VM_CAP_MTRAP_EXIT, 296 VM_CAP_PAUSE_EXIT, 297 VM_CAP_UNRESTRICTED_GUEST, 298 VM_CAP_ENABLE_INVPCID, 299 VM_CAP_MAX 300}; 301 302enum x2apic_state { 303 X2APIC_DISABLED, 304 X2APIC_ENABLED, 305 X2APIC_STATE_LAST 306}; 307 308enum vm_intr_trigger { 309 EDGE_TRIGGER, 310 LEVEL_TRIGGER 311}; 312 313/* 314 * The 'access' field has the format specified in Table 21-2 of the Intel 315 * Architecture Manual vol 3b. 316 * 317 * XXX The contents of the 'access' field are architecturally defined except 318 * bit 16 - Segment Unusable. 319 */ 320struct seg_desc { 321 uint64_t base; 322 uint32_t limit; 323 uint32_t access; 324}; 325#define SEG_DESC_TYPE(desc) ((desc)->access & 0x001f) 326#define SEG_DESC_PRESENT(desc) ((desc)->access & 0x0080) 327#define SEG_DESC_DEF32(desc) ((desc)->access & 0x4000) 328#define SEG_DESC_GRANULARITY(desc) ((desc)->access & 0x8000) 329#define SEG_DESC_UNUSABLE(desc) ((desc)->access & 0x10000) 330 331enum vm_cpu_mode { 332 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */ 333 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */ 334}; 335 336enum vm_paging_mode { 337 PAGING_MODE_FLAT, 338 PAGING_MODE_32, 339 PAGING_MODE_PAE, 340 PAGING_MODE_64, 341}; 342 343struct vm_guest_paging { 344 uint64_t cr3; 345 int cpl; 346 enum vm_cpu_mode cpu_mode; 347 enum vm_paging_mode paging_mode; 348}; 349 350/* 351 * The data structures 'vie' and 'vie_op' are meant to be opaque to the 352 * consumers of instruction decoding. The only reason why their contents 353 * need to be exposed is because they are part of the 'vm_exit' structure. 354 */ 355struct vie_op { 356 uint8_t op_byte; /* actual opcode byte */ 357 uint8_t op_type; /* type of operation (e.g. MOV) */ 358 uint16_t op_flags; 359}; 360 361#define VIE_INST_SIZE 15 362struct vie { 363 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */ 364 uint8_t num_valid; /* size of the instruction */ 365 uint8_t num_processed; 366 367 uint8_t rex_w:1, /* REX prefix */ 368 rex_r:1, 369 rex_x:1, 370 rex_b:1, 371 rex_present:1; 372 373 uint8_t mod:2, /* ModRM byte */ 374 reg:4, 375 rm:4; 376 377 uint8_t ss:2, /* SIB byte */ 378 index:4, 379 base:4; 380 381 uint8_t disp_bytes; 382 uint8_t imm_bytes; 383 384 uint8_t scale; 385 int base_register; /* VM_REG_GUEST_xyz */ 386 int index_register; /* VM_REG_GUEST_xyz */ 387 388 int64_t displacement; /* optional addr displacement */ 389 int64_t immediate; /* optional immediate operand */ 390 391 uint8_t decoded; /* set to 1 if successfully decoded */ 392 393 struct vie_op op; /* opcode description */ 394}; 395 396enum vm_exitcode { 397 VM_EXITCODE_INOUT, 398 VM_EXITCODE_VMX, 399 VM_EXITCODE_BOGUS, 400 VM_EXITCODE_RDMSR, 401 VM_EXITCODE_WRMSR, 402 VM_EXITCODE_HLT, 403 VM_EXITCODE_MTRAP, 404 VM_EXITCODE_PAUSE, 405 VM_EXITCODE_PAGING, 406 VM_EXITCODE_INST_EMUL, 407 VM_EXITCODE_SPINUP_AP, 408 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */ 409 VM_EXITCODE_RENDEZVOUS, 410 VM_EXITCODE_IOAPIC_EOI, 411 VM_EXITCODE_SUSPENDED, 412 VM_EXITCODE_INOUT_STR, 413 VM_EXITCODE_MAX 414}; 415 416struct vm_inout { 417 uint16_t bytes:3; /* 1 or 2 or 4 */ 418 uint16_t in:1; 419 uint16_t string:1; 420 uint16_t rep:1; 421 uint16_t port; 422 uint32_t eax; /* valid for out */ 423}; 424 425struct vm_inout_str { 426 struct vm_inout inout; /* must be the first element */ 427 struct vm_guest_paging paging; 428 uint64_t rflags; 429 uint64_t cr0; 430 uint64_t index; 431 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */ 432 int addrsize; 433 enum vm_reg_name seg_name; 434 struct seg_desc seg_desc; 435}; 436 437struct vm_exit { 438 enum vm_exitcode exitcode; 439 int inst_length; /* 0 means unknown */ 440 uint64_t rip; 441 union { 442 struct vm_inout inout; 443 struct vm_inout_str inout_str; 444 struct { 445 uint64_t gpa; 446 int fault_type; 447 } paging; 448 struct { 449 uint64_t gpa; 450 uint64_t gla; 451 struct vm_guest_paging paging; 452 struct vie vie; 453 } inst_emul; 454 /* 455 * VMX specific payload. Used when there is no "better" 456 * exitcode to represent the VM-exit. 457 */ 458 struct { 459 int status; /* vmx inst status */ 460 /* 461 * 'exit_reason' and 'exit_qualification' are valid 462 * only if 'status' is zero. 463 */ 464 uint32_t exit_reason; 465 uint64_t exit_qualification; 466 /* 467 * 'inst_error' and 'inst_type' are valid 468 * only if 'status' is non-zero. 469 */ 470 int inst_type; 471 int inst_error; 472 } vmx; 473 struct { 474 uint32_t code; /* ecx value */ 475 uint64_t wval; 476 } msr; 477 struct { 478 int vcpu; 479 uint64_t rip; 480 } spinup_ap; 481 struct { 482 uint64_t rflags; 483 } hlt; 484 struct { 485 int vector; 486 } ioapic_eoi; 487 struct { 488 enum vm_suspend_how how; 489 } suspended; 490 } u; 491}; 492 493#endif /* _VMM_H_ */ 494