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 * This file and its contents are supplied under the terms of the 30 * Common Development and Distribution License ("CDDL"), version 1.0. 31 * You may only use this file in accordance with the terms of version 32 * 1.0 of the CDDL. 33 * 34 * A full copy of the text of the CDDL should have accompanied this 35 * source. A copy of the CDDL is also available via the Internet at 36 * http://www.illumos.org/license/CDDL. 37 */ 38 /* This file is dual-licensed; see usr/src/contrib/bhyve/LICENSE */ 39 40 /* 41 * Copyright 2015 Pluribus Networks Inc. 42 * Copyright 2019 Joyent, Inc. 43 * Copyright 2025 Oxide Computer Company 44 * Copyright 2021 OmniOS Community Edition (OmniOSce) Association. 45 */ 46 47 #ifndef _VMM_KERNEL_H_ 48 #define _VMM_KERNEL_H_ 49 50 #include <sys/sdt.h> 51 #include <x86/segments.h> 52 #include <sys/vmm.h> 53 #include <sys/vmm_data.h> 54 #include <sys/linker_set.h> 55 56 SDT_PROVIDER_DECLARE(vmm); 57 58 struct vm; 59 struct vm_exception; 60 struct seg_desc; 61 struct vm_exit; 62 struct vie; 63 struct vm_run; 64 struct vhpet; 65 struct vioapic; 66 struct vlapic; 67 struct vmspace; 68 struct vm_client; 69 struct vm_object; 70 struct vm_guest_paging; 71 struct vmm_data_req; 72 73 /* Return values for architecture-specific calculation of the TSC multiplier */ 74 typedef enum { 75 FR_VALID, /* valid multiplier, scaling needed */ 76 FR_SCALING_NOT_NEEDED, /* scaling not required */ 77 FR_SCALING_NOT_SUPPORTED, /* scaling not supported by platform */ 78 FR_OUT_OF_RANGE, /* freq ratio out of supported range */ 79 } freqratio_res_t; 80 81 typedef int (*vmm_init_func_t)(void); 82 typedef int (*vmm_cleanup_func_t)(void); 83 typedef void (*vmm_resume_func_t)(void); 84 typedef void * (*vmi_init_func_t)(struct vm *vm); 85 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, uint64_t rip); 86 typedef void (*vmi_cleanup_func_t)(void *vmi); 87 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 88 uint64_t *retval); 89 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 90 uint64_t val); 91 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 92 struct seg_desc *desc); 93 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 94 const struct seg_desc *desc); 95 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 96 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 97 typedef struct vlapic *(*vmi_vlapic_init)(void *vmi, int vcpu); 98 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 99 typedef void (*vmi_savectx)(void *vmi, int vcpu); 100 typedef void (*vmi_restorectx)(void *vmi, int vcpu); 101 typedef void (*vmi_pause_t)(void *vmi, int vcpu); 102 103 typedef int (*vmi_get_msr_t)(void *vmi, int vcpu, uint32_t msr, 104 uint64_t *valp); 105 typedef int (*vmi_set_msr_t)(void *vmi, int vcpu, uint32_t msr, 106 uint64_t val); 107 typedef freqratio_res_t (*vmi_freqratio_t)(uint64_t guest_hz, 108 uint64_t host_hz, uint64_t *mult); 109 110 struct vmm_ops { 111 vmm_init_func_t init; /* module wide initialization */ 112 vmm_cleanup_func_t cleanup; 113 vmm_resume_func_t resume; 114 115 vmi_init_func_t vminit; /* vm-specific initialization */ 116 vmi_run_func_t vmrun; 117 vmi_cleanup_func_t vmcleanup; 118 vmi_get_register_t vmgetreg; 119 vmi_set_register_t vmsetreg; 120 vmi_get_desc_t vmgetdesc; 121 vmi_set_desc_t vmsetdesc; 122 vmi_get_cap_t vmgetcap; 123 vmi_set_cap_t vmsetcap; 124 vmi_vlapic_init vlapic_init; 125 vmi_vlapic_cleanup vlapic_cleanup; 126 vmi_pause_t vmpause; 127 128 vmi_savectx vmsavectx; 129 vmi_restorectx vmrestorectx; 130 131 vmi_get_msr_t vmgetmsr; 132 vmi_set_msr_t vmsetmsr; 133 134 vmi_freqratio_t vmfreqratio; 135 uint32_t fr_intsize; 136 uint32_t fr_fracsize; 137 }; 138 139 extern struct vmm_ops vmm_ops_intel; 140 extern struct vmm_ops vmm_ops_amd; 141 142 int vm_create(uint64_t flags, struct vm **retvm); 143 void vm_destroy(struct vm *vm); 144 int vm_reinit(struct vm *vm, uint64_t); 145 uint16_t vm_get_maxcpus(struct vm *vm); 146 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores, 147 uint16_t *threads, uint16_t *maxcpus); 148 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores, 149 uint16_t threads, uint16_t maxcpus); 150 151 int vm_pause_instance(struct vm *); 152 int vm_resume_instance(struct vm *); 153 bool vm_is_paused(struct vm *); 154 155 /* 156 * APIs that race against hardware. 157 */ 158 int vm_track_dirty_pages(struct vm *, uint64_t, size_t, uint8_t *); 159 int vm_npt_do_operation(struct vm *, uint64_t, size_t, uint32_t, uint8_t *, 160 int *); 161 162 /* 163 * APIs that modify the guest memory map require all vcpus to be frozen. 164 */ 165 int vm_mmap_memseg(struct vm *, vm_paddr_t, int, uintptr_t, size_t, int, int); 166 int vm_munmap_memseg(struct vm *, vm_paddr_t, size_t); 167 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem); 168 void vm_free_memseg(struct vm *vm, int ident); 169 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 170 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 171 int vm_assign_pptdev(struct vm *vm, int pptfd); 172 int vm_unassign_pptdev(struct vm *vm, int pptfd); 173 174 /* 175 * APIs that inspect the guest memory map require only a *single* vcpu to 176 * be frozen. This acts like a read lock on the guest memory map since any 177 * modification requires *all* vcpus to be frozen. 178 */ 179 int vm_mmap_getnext(struct vm *, vm_paddr_t *, int *, uintptr_t *, size_t *, 180 int *, int *); 181 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem, 182 struct vm_object **objptr); 183 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm); 184 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa); 185 186 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 187 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 188 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 189 struct seg_desc *ret_desc); 190 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 191 const struct seg_desc *desc); 192 int vm_get_run_state(struct vm *vm, int vcpuid, uint32_t *state, 193 uint8_t *sipi_vec); 194 int vm_set_run_state(struct vm *vm, int vcpuid, uint32_t state, 195 uint8_t sipi_vec); 196 int vm_get_fpu(struct vm *vm, int vcpuid, void *buf, size_t len); 197 int vm_set_fpu(struct vm *vm, int vcpuid, void *buf, size_t len); 198 int vm_run(struct vm *vm, int vcpuid, const struct vm_entry *); 199 int vm_suspend(struct vm *, enum vm_suspend_how, int); 200 int vm_inject_nmi(struct vm *vm, int vcpu); 201 bool vm_nmi_pending(struct vm *vm, int vcpuid); 202 void vm_nmi_clear(struct vm *vm, int vcpuid); 203 int vm_inject_extint(struct vm *vm, int vcpu); 204 bool vm_extint_pending(struct vm *vm, int vcpuid); 205 void vm_extint_clear(struct vm *vm, int vcpuid); 206 int vm_inject_init(struct vm *vm, int vcpuid); 207 int vm_inject_sipi(struct vm *vm, int vcpuid, uint8_t vec); 208 struct vlapic *vm_lapic(struct vm *vm, int cpu); 209 struct vioapic *vm_ioapic(struct vm *vm); 210 struct vhpet *vm_hpet(struct vm *vm); 211 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 212 int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 213 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 214 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 215 int vm_apicid2vcpuid(struct vm *vm, int apicid); 216 int vm_activate_cpu(struct vm *vm, int vcpu); 217 int vm_suspend_cpu(struct vm *vm, int vcpu); 218 int vm_resume_cpu(struct vm *vm, int vcpu); 219 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 220 struct vie *vm_vie_ctx(struct vm *vm, int vcpuid); 221 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 222 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip); 223 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip); 224 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip); 225 void vm_exit_run_state(struct vm *vm, int vcpuid, uint64_t rip); 226 int vm_service_mmio_read(struct vm *vm, int cpuid, uint64_t gpa, uint64_t *rval, 227 int rsize); 228 int vm_service_mmio_write(struct vm *vm, int cpuid, uint64_t gpa, uint64_t wval, 229 int wsize); 230 231 #ifdef _SYS__CPUSET_H_ 232 cpuset_t vm_active_cpus(struct vm *vm); 233 cpuset_t vm_debug_cpus(struct vm *vm); 234 #endif /* _SYS__CPUSET_H_ */ 235 236 bool vcpu_entry_bailout_checks(struct vm *vm, int vcpuid, uint64_t rip); 237 bool vcpu_run_state_pending(struct vm *vm, int vcpuid); 238 int vcpu_arch_reset(struct vm *vm, int vcpuid, bool init_only); 239 int vm_vcpu_barrier(struct vm *, int); 240 241 /* 242 * Return true if device indicated by bus/slot/func is supposed to be a 243 * pci passthrough device. 244 * 245 * Return false otherwise. 246 */ 247 bool vmm_is_pptdev(int bus, int slot, int func); 248 249 void *vm_iommu_domain(struct vm *vm); 250 251 enum vcpu_state { 252 VCPU_IDLE, 253 VCPU_FROZEN, 254 VCPU_RUNNING, 255 VCPU_SLEEPING, 256 }; 257 258 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 259 bool from_idle); 260 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 261 void vcpu_block_run(struct vm *, int); 262 void vcpu_unblock_run(struct vm *, int); 263 264 uint64_t vcpu_tsc_offset(struct vm *vm, int vcpuid, bool phys_adj); 265 hrtime_t vm_normalize_hrtime(struct vm *, hrtime_t); 266 hrtime_t vm_denormalize_hrtime(struct vm *, hrtime_t); 267 uint64_t vm_get_freq_multiplier(struct vm *); 268 269 static __inline bool 270 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 271 { 272 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 273 } 274 275 #ifdef _SYS_THREAD_H 276 static __inline int 277 vcpu_should_yield(struct vm *vm, int vcpu) 278 { 279 280 if (curthread->t_astflag) 281 return (1); 282 else if (CPU->cpu_runrun) 283 return (1); 284 else 285 return (0); 286 } 287 #endif /* _SYS_THREAD_H */ 288 289 typedef enum vcpu_notify { 290 VCPU_NOTIFY_NONE, 291 VCPU_NOTIFY_APIC, /* Posted intr notification (if possible) */ 292 VCPU_NOTIFY_EXIT, /* IPI to cause VM exit */ 293 } vcpu_notify_t; 294 295 void *vcpu_stats(struct vm *vm, int vcpu); 296 void vcpu_notify_event(struct vm *vm, int vcpuid); 297 void vcpu_notify_event_type(struct vm *vm, int vcpuid, vcpu_notify_t); 298 void *vm_get_cookie(struct vm *); 299 struct vmspace *vm_get_vmspace(struct vm *vm); 300 struct vm_client *vm_get_vmclient(struct vm *vm, int vcpuid); 301 struct vatpic *vm_atpic(struct vm *vm); 302 struct vatpit *vm_atpit(struct vm *vm); 303 struct vpmtmr *vm_pmtmr(struct vm *vm); 304 struct vrtc *vm_rtc(struct vm *vm); 305 306 /* 307 * Inject exception 'vector' into the guest vcpu. This function returns 0 on 308 * success and non-zero on failure. 309 * 310 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 311 * this function directly because they enforce the trap-like or fault-like 312 * behavior of an exception. 313 * 314 * This function should only be called in the context of the thread that is 315 * executing this vcpu. 316 */ 317 int vm_inject_exception(struct vm *vm, int vcpuid, uint8_t vector, 318 bool err_valid, uint32_t errcode, bool restart_instruction); 319 320 /* 321 * This function is called after a VM-exit that occurred during exception or 322 * interrupt delivery through the IDT. The format of 'intinfo' is described 323 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. 324 * 325 * If a VM-exit handler completes the event delivery successfully then it 326 * should call vm_exit_intinfo() to extinguish the pending event. For e.g., 327 * if the task switch emulation is triggered via a task gate then it should 328 * call this function with 'intinfo=0' to indicate that the external event 329 * is not pending anymore. 330 * 331 * Return value is 0 on success and non-zero on failure. 332 */ 333 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo); 334 335 /* 336 * This function is called before every VM-entry to retrieve a pending 337 * event that should be injected into the guest. This function combines 338 * nested events into a double or triple fault. 339 * 340 * Returns false if there are no events that need to be injected into the guest. 341 */ 342 bool vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info); 343 344 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2); 345 346 enum vm_reg_name vm_segment_name(int seg_encoding); 347 348 struct vm_copyinfo { 349 uint64_t gpa; 350 size_t len; 351 int prot; 352 void *hva; 353 void *cookie; 354 }; 355 356 /* 357 * Set up 'copyinfo[]' to copy to/from guest linear address space starting 358 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for 359 * a copyin or PROT_WRITE for a copyout. 360 * 361 * retval is_fault Interpretation 362 * 0 0 Success 363 * 0 1 An exception was injected into the guest 364 * EFAULT N/A Unrecoverable error 365 * 366 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if 367 * the return value is 0. The 'copyinfo[]' resources should be freed by calling 368 * 'vm_copy_teardown()' after the copy is done. 369 */ 370 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging, 371 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo, 372 uint_t num_copyinfo, int *is_fault); 373 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 374 uint_t num_copyinfo); 375 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 376 void *kaddr, size_t len); 377 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr, 378 struct vm_copyinfo *copyinfo, size_t len); 379 380 int vcpu_trace_exceptions(struct vm *vm, int vcpuid); 381 int vcpu_trap_wbinvd(struct vm *vm, int vcpuid); 382 383 void vm_inject_ud(struct vm *vm, int vcpuid); 384 void vm_inject_gp(struct vm *vm, int vcpuid); 385 void vm_inject_ac(struct vm *vm, int vcpuid, uint32_t errcode); 386 void vm_inject_ss(struct vm *vm, int vcpuid, uint32_t errcode); 387 void vm_inject_pf(struct vm *vm, int vcpuid, uint32_t errcode, uint64_t cr2); 388 389 /* 390 * Both SVM and VMX have complex logic for injecting events such as exceptions 391 * or interrupts into the guest. Within those two backends, the progress of 392 * event injection is tracked by event_inject_state, hopefully making it easier 393 * to reason about. 394 */ 395 enum event_inject_state { 396 EIS_CAN_INJECT = 0, /* exception/interrupt can be injected */ 397 EIS_EV_EXISTING = 1, /* blocked by existing event */ 398 EIS_EV_INJECTED = 2, /* blocked by injected event */ 399 EIS_GI_BLOCK = 3, /* blocked by guest interruptability */ 400 401 /* 402 * Flag to request an immediate exit from VM context after event 403 * injection in order to perform more processing 404 */ 405 EIS_REQ_EXIT = (1 << 15), 406 }; 407 408 /* Possible result codes for MSR access emulation */ 409 typedef enum vm_msr_result { 410 VMR_OK = 0, /* succesfully emulated */ 411 VMR_GP = 1, /* #GP should be injected */ 412 VMR_UNHANLDED = 2, /* handle in userspace, kernel cannot emulate */ 413 } vm_msr_result_t; 414 415 enum vm_cpuid_capability { 416 VCC_NONE, 417 VCC_NO_EXECUTE, 418 VCC_FFXSR, 419 VCC_TCE, 420 VCC_LAST 421 }; 422 423 /* Possible flags and entry count limit definited in sys/vmm.h */ 424 typedef struct vcpu_cpuid_config { 425 uint32_t vcc_flags; 426 uint32_t vcc_nent; 427 struct vcpu_cpuid_entry *vcc_entries; 428 } vcpu_cpuid_config_t; 429 430 vcpu_cpuid_config_t *vm_cpuid_config(struct vm *, int); 431 int vm_get_cpuid(struct vm *, int, vcpu_cpuid_config_t *); 432 int vm_set_cpuid(struct vm *, int, const vcpu_cpuid_config_t *); 433 void vcpu_emulate_cpuid(struct vm *, int, uint64_t *, uint64_t *, uint64_t *, 434 uint64_t *); 435 void legacy_emulate_cpuid(struct vm *, int, uint32_t *, uint32_t *, uint32_t *, 436 uint32_t *); 437 void vcpu_cpuid_init(vcpu_cpuid_config_t *); 438 void vcpu_cpuid_cleanup(vcpu_cpuid_config_t *); 439 440 bool vm_cpuid_capability(struct vm *, int, enum vm_cpuid_capability); 441 bool validate_guest_xcr0(uint64_t, uint64_t); 442 443 void vmm_sol_glue_init(void); 444 void vmm_sol_glue_cleanup(void); 445 446 void *vmm_contig_alloc(size_t); 447 void vmm_contig_free(void *, size_t); 448 449 int vmm_mod_load(void); 450 int vmm_mod_unload(void); 451 452 bool vmm_check_iommu(void); 453 454 void vmm_call_trap(uint64_t); 455 456 uint64_t vmm_host_tsc_delta(void); 457 458 /* 459 * Because of tangled headers, this is not exposed directly via the vmm_drv 460 * interface, but rather mirrored as vmm_drv_iop_cb_t in vmm_drv.h. 461 */ 462 typedef int (*ioport_handler_t)(void *, bool, uint16_t, uint8_t, uint32_t *); 463 464 int vm_ioport_access(struct vm *vm, int vcpuid, bool in, uint16_t port, 465 uint8_t bytes, uint32_t *val); 466 467 int vm_ioport_attach(struct vm *vm, uint16_t port, ioport_handler_t func, 468 void *arg, void **cookie); 469 int vm_ioport_detach(struct vm *vm, void **cookie, ioport_handler_t *old_func, 470 void **old_arg); 471 472 int vm_ioport_hook(struct vm *, uint16_t, ioport_handler_t, void *, void **); 473 void vm_ioport_unhook(struct vm *, void **); 474 475 enum vcpu_ustate { 476 VU_INIT = 0, /* initialized but has not yet attempted to run */ 477 VU_RUN, /* running in guest context */ 478 VU_IDLE, /* idle (HLTed, wait-for-SIPI, etc) */ 479 VU_EMU_KERN, /* emulation performed in-kernel */ 480 VU_EMU_USER, /* emulation performed in userspace */ 481 VU_SCHED, /* off-cpu for interrupt, preempt, lock contention */ 482 VU_MAX 483 }; 484 485 void vcpu_ustate_change(struct vm *, int, enum vcpu_ustate); 486 487 typedef struct vmm_kstats { 488 kstat_named_t vk_name; 489 } vmm_kstats_t; 490 491 typedef struct vmm_vcpu_kstats { 492 kstat_named_t vvk_vcpu; 493 kstat_named_t vvk_time_init; 494 kstat_named_t vvk_time_run; 495 kstat_named_t vvk_time_idle; 496 kstat_named_t vvk_time_emu_kern; 497 kstat_named_t vvk_time_emu_user; 498 kstat_named_t vvk_time_sched; 499 } vmm_vcpu_kstats_t; 500 501 #define VMM_KSTAT_CLASS "misc" 502 503 int vmm_kstat_update_vcpu(struct kstat *, int); 504 505 typedef struct vmm_data_req { 506 uint16_t vdr_class; 507 uint16_t vdr_version; 508 uint32_t vdr_flags; 509 uint32_t vdr_len; 510 void *vdr_data; 511 uint32_t *vdr_result_len; 512 int vdr_vcpuid; 513 } vmm_data_req_t; 514 515 typedef int (*vmm_data_writef_t)(void *, const vmm_data_req_t *); 516 typedef int (*vmm_data_readf_t)(void *, const vmm_data_req_t *); 517 typedef int (*vmm_data_vcpu_writef_t)(struct vm *, int, const vmm_data_req_t *); 518 typedef int (*vmm_data_vcpu_readf_t)(struct vm *, int, const vmm_data_req_t *); 519 520 typedef struct vmm_data_version_entry { 521 uint16_t vdve_class; 522 uint16_t vdve_version; 523 524 /* 525 * If these handlers accept/emit a single item of a fixed length, it 526 * should be specified in vdve_len_expect. The vmm-data logic will then 527 * ensure that requests possess at least that specified length before 528 * calling into the defined handlers. 529 */ 530 uint16_t vdve_len_expect; 531 532 /* 533 * For handlers which deal with (potentially) multiple items of a fixed 534 * length, vdve_len_per_item is used to hint (via the VDC_VERSION class) 535 * to userspace what that item size is. Although not strictly mutually 536 * exclusive with vdve_len_expect, it is nonsensical to set them both. 537 */ 538 uint16_t vdve_len_per_item; 539 540 /* 541 * A vmm-data handler is expected to provide read/write functions which 542 * are either VM-wide (via vdve_readf and vdve_writef) or per-vCPU 543 * (via vdve_vcpu_readf and vdve_vcpu_writef). Providing both is not 544 * allowed (but is not currently checked at compile time). 545 */ 546 547 /* VM-wide handlers */ 548 vmm_data_readf_t vdve_readf; 549 vmm_data_writef_t vdve_writef; 550 551 /* Per-vCPU handlers */ 552 vmm_data_vcpu_readf_t vdve_vcpu_readf; 553 vmm_data_vcpu_writef_t vdve_vcpu_writef; 554 555 /* 556 * The vdve_vcpu_readf/writef handlers can rely on vcpuid to be within 557 * the [0, VM_MAXCPU) bounds. If they also can handle vcpuid == -1 (for 558 * VM-wide data), then they can opt into such cases by setting 559 * vdve_vcpu_wildcard to true. 560 * 561 * At a later time, it would make sense to improve the logic so a 562 * vmm-data class could define both the VM-wide and per-vCPU handlers, 563 * letting the incoming vcpuid determine which would be called. Until 564 * then, vdve_vcpu_wildcard is the stopgap. 565 */ 566 bool vdve_vcpu_wildcard; 567 } vmm_data_version_entry_t; 568 569 #define VMM_DATA_VERSION(sym) SET_ENTRY(vmm_data_version_entries, sym) 570 571 int vmm_data_read(struct vm *, const vmm_data_req_t *); 572 int vmm_data_write(struct vm *, const vmm_data_req_t *); 573 574 /* 575 * TSC Scaling 576 */ 577 uint64_t vmm_calc_freq_multiplier(uint64_t guest_hz, uint64_t host_hz, 578 uint32_t frac); 579 580 /* represents a multiplier for a guest in which no scaling is required */ 581 #define VM_TSCM_NOSCALE 0 582 583 #endif /* _VMM_KERNEL_H_ */ 584