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 * This file and its contents are supplied under the terms of the 32 * Common Development and Distribution License ("CDDL"), version 1.0. 33 * You may only use this file in accordance with the terms of version 34 * 1.0 of the CDDL. 35 * 36 * A full copy of the text of the CDDL should have accompanied this 37 * source. A copy of the CDDL is also available via the Internet at 38 * http://www.illumos.org/license/CDDL. 39 * 40 * Copyright 2015 Pluribus Networks Inc. 41 * Copyright 2019 Joyent, Inc. 42 * Copyright 2022 Oxide Computer Company 43 * Copyright 2021 OmniOS Community Edition (OmniOSce) Association. 44 */ 45 46 #ifndef _VMM_KERNEL_H_ 47 #define _VMM_KERNEL_H_ 48 49 #include <sys/sdt.h> 50 #include <x86/segments.h> 51 #include <sys/vmm.h> 52 53 SDT_PROVIDER_DECLARE(vmm); 54 55 struct vm; 56 struct vm_exception; 57 struct seg_desc; 58 struct vm_exit; 59 struct vie; 60 struct vm_run; 61 struct vhpet; 62 struct vioapic; 63 struct vlapic; 64 struct vmspace; 65 struct vm_client; 66 struct vm_object; 67 struct vm_guest_paging; 68 69 typedef int (*vmm_init_func_t)(void); 70 typedef int (*vmm_cleanup_func_t)(void); 71 typedef void (*vmm_resume_func_t)(void); 72 typedef void * (*vmi_init_func_t)(struct vm *vm); 73 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, uint64_t rip); 74 typedef void (*vmi_cleanup_func_t)(void *vmi); 75 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num, 76 uint64_t *retval); 77 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num, 78 uint64_t val); 79 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num, 80 struct seg_desc *desc); 81 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num, 82 const struct seg_desc *desc); 83 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval); 84 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val); 85 typedef struct vlapic *(*vmi_vlapic_init)(void *vmi, int vcpu); 86 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic); 87 typedef void (*vmi_savectx)(void *vmi, int vcpu); 88 typedef void (*vmi_restorectx)(void *vmi, int vcpu); 89 90 struct vmm_ops { 91 vmm_init_func_t init; /* module wide initialization */ 92 vmm_cleanup_func_t cleanup; 93 vmm_resume_func_t resume; 94 95 vmi_init_func_t vminit; /* vm-specific initialization */ 96 vmi_run_func_t vmrun; 97 vmi_cleanup_func_t vmcleanup; 98 vmi_get_register_t vmgetreg; 99 vmi_set_register_t vmsetreg; 100 vmi_get_desc_t vmgetdesc; 101 vmi_set_desc_t vmsetdesc; 102 vmi_get_cap_t vmgetcap; 103 vmi_set_cap_t vmsetcap; 104 vmi_vlapic_init vlapic_init; 105 vmi_vlapic_cleanup vlapic_cleanup; 106 107 vmi_savectx vmsavectx; 108 vmi_restorectx vmrestorectx; 109 }; 110 111 extern struct vmm_ops vmm_ops_intel; 112 extern struct vmm_ops vmm_ops_amd; 113 114 int vm_create(const char *name, uint64_t flags, struct vm **retvm); 115 void vm_destroy(struct vm *vm); 116 int vm_reinit(struct vm *vm, uint64_t); 117 const char *vm_name(struct vm *vm); 118 uint16_t vm_get_maxcpus(struct vm *vm); 119 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores, 120 uint16_t *threads, uint16_t *maxcpus); 121 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores, 122 uint16_t threads, uint16_t maxcpus); 123 124 /* 125 * APIs that race against hardware. 126 */ 127 void vm_track_dirty_pages(struct vm *, uint64_t, size_t, uint8_t *); 128 129 /* 130 * APIs that modify the guest memory map require all vcpus to be frozen. 131 */ 132 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off, 133 size_t len, int prot, int flags); 134 int vm_munmap_memseg(struct vm *vm, vm_paddr_t gpa, size_t len); 135 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem); 136 void vm_free_memseg(struct vm *vm, int ident); 137 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); 138 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); 139 int vm_assign_pptdev(struct vm *vm, int pptfd); 140 int vm_unassign_pptdev(struct vm *vm, int pptfd); 141 142 /* 143 * APIs that inspect the guest memory map require only a *single* vcpu to 144 * be frozen. This acts like a read lock on the guest memory map since any 145 * modification requires *all* vcpus to be frozen. 146 */ 147 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid, 148 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags); 149 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem, 150 struct vm_object **objptr); 151 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm); 152 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa); 153 154 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval); 155 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val); 156 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg, 157 struct seg_desc *ret_desc); 158 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg, 159 const struct seg_desc *desc); 160 int vm_get_run_state(struct vm *vm, int vcpuid, uint32_t *state, 161 uint8_t *sipi_vec); 162 int vm_set_run_state(struct vm *vm, int vcpuid, uint32_t state, 163 uint8_t sipi_vec); 164 int vm_get_fpu(struct vm *vm, int vcpuid, void *buf, size_t len); 165 int vm_set_fpu(struct vm *vm, int vcpuid, void *buf, size_t len); 166 int vm_run(struct vm *vm, int vcpuid, const struct vm_entry *); 167 int vm_suspend(struct vm *vm, enum vm_suspend_how how); 168 int vm_inject_nmi(struct vm *vm, int vcpu); 169 int vm_nmi_pending(struct vm *vm, int vcpuid); 170 void vm_nmi_clear(struct vm *vm, int vcpuid); 171 int vm_inject_extint(struct vm *vm, int vcpu); 172 int vm_extint_pending(struct vm *vm, int vcpuid); 173 void vm_extint_clear(struct vm *vm, int vcpuid); 174 int vm_inject_init(struct vm *vm, int vcpuid); 175 int vm_inject_sipi(struct vm *vm, int vcpuid, uint8_t vec); 176 struct vlapic *vm_lapic(struct vm *vm, int cpu); 177 struct vioapic *vm_ioapic(struct vm *vm); 178 struct vhpet *vm_hpet(struct vm *vm); 179 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val); 180 int vm_set_capability(struct vm *vm, int vcpu, int type, int val); 181 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state); 182 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state); 183 int vm_apicid2vcpuid(struct vm *vm, int apicid); 184 int vm_activate_cpu(struct vm *vm, int vcpu); 185 int vm_suspend_cpu(struct vm *vm, int vcpu); 186 int vm_resume_cpu(struct vm *vm, int vcpu); 187 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid); 188 struct vie *vm_vie_ctx(struct vm *vm, int vcpuid); 189 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip); 190 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip); 191 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip); 192 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip); 193 void vm_exit_run_state(struct vm *vm, int vcpuid, uint64_t rip); 194 int vm_service_mmio_read(struct vm *vm, int cpuid, uint64_t gpa, uint64_t *rval, 195 int rsize); 196 int vm_service_mmio_write(struct vm *vm, int cpuid, uint64_t gpa, uint64_t wval, 197 int wsize); 198 199 #ifdef _SYS__CPUSET_H_ 200 cpuset_t vm_active_cpus(struct vm *vm); 201 cpuset_t vm_debug_cpus(struct vm *vm); 202 cpuset_t vm_suspended_cpus(struct vm *vm); 203 #endif /* _SYS__CPUSET_H_ */ 204 205 bool vcpu_entry_bailout_checks(struct vm *vm, int vcpuid, uint64_t rip); 206 bool vcpu_run_state_pending(struct vm *vm, int vcpuid); 207 int vcpu_arch_reset(struct vm *vm, int vcpuid, bool init_only); 208 209 /* 210 * Return true if device indicated by bus/slot/func is supposed to be a 211 * pci passthrough device. 212 * 213 * Return false otherwise. 214 */ 215 bool vmm_is_pptdev(int bus, int slot, int func); 216 217 void *vm_iommu_domain(struct vm *vm); 218 219 enum vcpu_state { 220 VCPU_IDLE, 221 VCPU_FROZEN, 222 VCPU_RUNNING, 223 VCPU_SLEEPING, 224 }; 225 226 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state, 227 bool from_idle); 228 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu); 229 void vcpu_block_run(struct vm *, int); 230 void vcpu_unblock_run(struct vm *, int); 231 232 uint64_t vcpu_tsc_offset(struct vm *vm, int vcpuid, bool phys_adj); 233 234 static __inline int 235 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu) 236 { 237 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING); 238 } 239 240 #ifdef _SYS_THREAD_H 241 static __inline int 242 vcpu_should_yield(struct vm *vm, int vcpu) 243 { 244 245 if (curthread->t_astflag) 246 return (1); 247 else if (CPU->cpu_runrun) 248 return (1); 249 else 250 return (0); 251 } 252 #endif /* _SYS_THREAD_H */ 253 254 typedef enum vcpu_notify { 255 VCPU_NOTIFY_NONE, 256 VCPU_NOTIFY_APIC, /* Posted intr notification (if possible) */ 257 VCPU_NOTIFY_EXIT, /* IPI to cause VM exit */ 258 } vcpu_notify_t; 259 260 void *vcpu_stats(struct vm *vm, int vcpu); 261 void vcpu_notify_event(struct vm *vm, int vcpuid); 262 void vcpu_notify_event_type(struct vm *vm, int vcpuid, vcpu_notify_t); 263 struct vmspace *vm_get_vmspace(struct vm *vm); 264 struct vm_client *vm_get_vmclient(struct vm *vm, int vcpuid); 265 struct vatpic *vm_atpic(struct vm *vm); 266 struct vatpit *vm_atpit(struct vm *vm); 267 struct vpmtmr *vm_pmtmr(struct vm *vm); 268 struct vrtc *vm_rtc(struct vm *vm); 269 270 /* 271 * Inject exception 'vector' into the guest vcpu. This function returns 0 on 272 * success and non-zero on failure. 273 * 274 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling 275 * this function directly because they enforce the trap-like or fault-like 276 * behavior of an exception. 277 * 278 * This function should only be called in the context of the thread that is 279 * executing this vcpu. 280 */ 281 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid, 282 uint32_t errcode, int restart_instruction); 283 284 /* 285 * This function is called after a VM-exit that occurred during exception or 286 * interrupt delivery through the IDT. The format of 'intinfo' is described 287 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. 288 * 289 * If a VM-exit handler completes the event delivery successfully then it 290 * should call vm_exit_intinfo() to extinguish the pending event. For e.g., 291 * if the task switch emulation is triggered via a task gate then it should 292 * call this function with 'intinfo=0' to indicate that the external event 293 * is not pending anymore. 294 * 295 * Return value is 0 on success and non-zero on failure. 296 */ 297 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo); 298 299 /* 300 * This function is called before every VM-entry to retrieve a pending 301 * event that should be injected into the guest. This function combines 302 * nested events into a double or triple fault. 303 * 304 * Returns 0 if there are no events that need to be injected into the guest 305 * and non-zero otherwise. 306 */ 307 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info); 308 309 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2); 310 311 enum vm_reg_name vm_segment_name(int seg_encoding); 312 313 struct vm_copyinfo { 314 uint64_t gpa; 315 size_t len; 316 int prot; 317 void *hva; 318 void *cookie; 319 }; 320 321 /* 322 * Set up 'copyinfo[]' to copy to/from guest linear address space starting 323 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for 324 * a copyin or PROT_WRITE for a copyout. 325 * 326 * retval is_fault Interpretation 327 * 0 0 Success 328 * 0 1 An exception was injected into the guest 329 * EFAULT N/A Unrecoverable error 330 * 331 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if 332 * the return value is 0. The 'copyinfo[]' resources should be freed by calling 333 * 'vm_copy_teardown()' after the copy is done. 334 */ 335 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging, 336 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo, 337 uint_t num_copyinfo, int *is_fault); 338 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 339 uint_t num_copyinfo); 340 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo, 341 void *kaddr, size_t len); 342 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr, 343 struct vm_copyinfo *copyinfo, size_t len); 344 345 int vcpu_trace_exceptions(struct vm *vm, int vcpuid); 346 347 /* APIs to inject faults into the guest */ 348 void vm_inject_fault(struct vm *vm, int vcpuid, int vector, int errcode_valid, 349 int errcode); 350 351 void vm_inject_ud(struct vm *vm, int vcpuid); 352 void vm_inject_gp(struct vm *vm, int vcpuid); 353 void vm_inject_ac(struct vm *vm, int vcpuid, int errcode); 354 void vm_inject_ss(struct vm *vm, int vcpuid, int errcode); 355 void vm_inject_pf(struct vm *vm, int vcpuid, int errcode, uint64_t cr2); 356 357 /* 358 * Both SVM and VMX have complex logic for injecting events such as exceptions 359 * or interrupts into the guest. Within those two backends, the progress of 360 * event injection is tracked by event_inject_state, hopefully making it easier 361 * to reason about. 362 */ 363 enum event_inject_state { 364 EIS_CAN_INJECT = 0, /* exception/interrupt can be injected */ 365 EIS_EV_EXISTING = 1, /* blocked by existing event */ 366 EIS_EV_INJECTED = 2, /* blocked by injected event */ 367 EIS_GI_BLOCK = 3, /* blocked by guest interruptability */ 368 369 /* 370 * Flag to request an immediate exit from VM context after event 371 * injection in order to perform more processing 372 */ 373 EIS_REQ_EXIT = (1 << 15), 374 }; 375 376 /* Possible result codes for MSR access emulation */ 377 typedef enum vm_msr_result { 378 VMR_OK = 0, /* succesfully emulated */ 379 VMR_GP = 1, /* #GP should be injected */ 380 VMR_UNHANLDED = 2, /* handle in userspace, kernel cannot emulate */ 381 } vm_msr_result_t; 382 383 void vmm_sol_glue_init(void); 384 void vmm_sol_glue_cleanup(void); 385 386 int vmm_mod_load(void); 387 int vmm_mod_unload(void); 388 389 void vmm_call_trap(uint64_t); 390 391 /* 392 * Because of tangled headers, this is not exposed directly via the vmm_drv 393 * interface, but rather mirrored as vmm_drv_iop_cb_t in vmm_drv.h. 394 */ 395 typedef int (*ioport_handler_t)(void *, bool, uint16_t, uint8_t, uint32_t *); 396 397 int vm_ioport_access(struct vm *vm, int vcpuid, bool in, uint16_t port, 398 uint8_t bytes, uint32_t *val); 399 400 int vm_ioport_attach(struct vm *vm, uint16_t port, ioport_handler_t func, 401 void *arg, void **cookie); 402 int vm_ioport_detach(struct vm *vm, void **cookie, ioport_handler_t *old_func, 403 void **old_arg); 404 405 int vm_ioport_hook(struct vm *, uint16_t, ioport_handler_t, void *, void **); 406 void vm_ioport_unhook(struct vm *, void **); 407 408 enum vcpu_ustate { 409 VU_INIT = 0, /* initialized but has not yet attempted to run */ 410 VU_RUN, /* running in guest context */ 411 VU_IDLE, /* idle (HLTed, wait-for-SIPI, etc) */ 412 VU_EMU_KERN, /* emulation performed in-kernel */ 413 VU_EMU_USER, /* emulation performed in userspace */ 414 VU_SCHED, /* off-cpu for interrupt, preempt, lock contention */ 415 VU_MAX 416 }; 417 418 void vcpu_ustate_change(struct vm *, int, enum vcpu_ustate); 419 420 typedef struct vmm_kstats { 421 kstat_named_t vk_name; 422 } vmm_kstats_t; 423 424 typedef struct vmm_vcpu_kstats { 425 kstat_named_t vvk_vcpu; 426 kstat_named_t vvk_time_init; 427 kstat_named_t vvk_time_run; 428 kstat_named_t vvk_time_idle; 429 kstat_named_t vvk_time_emu_kern; 430 kstat_named_t vvk_time_emu_user; 431 kstat_named_t vvk_time_sched; 432 } vmm_vcpu_kstats_t; 433 434 #define VMM_KSTAT_CLASS "misc" 435 436 int vmm_kstat_update_vcpu(struct kstat *, int); 437 438 #endif /* _VMM_KERNEL_H_ */ 439