| 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 | #ifndef _VMM_H_ |
| 30 | #define	_VMM_H_ |
| 31 | |
| 32 | #include <sys/cpuset.h> |
| 33 | #include <sys/sdt.h> |
| 34 | #include <x86/segments.h> |
| 35 | |
| 36 | struct vcpu; |
| 37 | struct vm_snapshot_meta; |
| 38 | |
| 39 | #ifdef _KERNEL |
| 40 | SDT_PROVIDER_DECLARE(vmm); |
| 41 | #endif |
| 42 | |
| 43 | enum vm_suspend_how { |
| 44 | 	VM_SUSPEND_NONE, |
| 45 | 	VM_SUSPEND_RESET, |
| 46 | 	VM_SUSPEND_POWEROFF, |
| 47 | 	VM_SUSPEND_HALT, |
| 48 | 	VM_SUSPEND_TRIPLEFAULT, |
| 49 | 	VM_SUSPEND_DESTROY, |
| 50 | 	VM_SUSPEND_LAST |
| 51 | }; |
| 52 | |
| 53 | /* |
| 54 | * Identifiers for architecturally defined registers. |
| 55 | */ |
| 56 | enum vm_reg_name { |
| 57 | 	VM_REG_GUEST_RAX, |
| 58 | 	VM_REG_GUEST_RBX, |
| 59 | 	VM_REG_GUEST_RCX, |
| 60 | 	VM_REG_GUEST_RDX, |
| 61 | 	VM_REG_GUEST_RSI, |
| 62 | 	VM_REG_GUEST_RDI, |
| 63 | 	VM_REG_GUEST_RBP, |
| 64 | 	VM_REG_GUEST_R8, |
| 65 | 	VM_REG_GUEST_R9, |
| 66 | 	VM_REG_GUEST_R10, |
| 67 | 	VM_REG_GUEST_R11, |
| 68 | 	VM_REG_GUEST_R12, |
| 69 | 	VM_REG_GUEST_R13, |
| 70 | 	VM_REG_GUEST_R14, |
| 71 | 	VM_REG_GUEST_R15, |
| 72 | 	VM_REG_GUEST_CR0, |
| 73 | 	VM_REG_GUEST_CR3, |
| 74 | 	VM_REG_GUEST_CR4, |
| 75 | 	VM_REG_GUEST_DR7, |
| 76 | 	VM_REG_GUEST_RSP, |
| 77 | 	VM_REG_GUEST_RIP, |
| 78 | 	VM_REG_GUEST_RFLAGS, |
| 79 | 	VM_REG_GUEST_ES, |
| 80 | 	VM_REG_GUEST_CS, |
| 81 | 	VM_REG_GUEST_SS, |
| 82 | 	VM_REG_GUEST_DS, |
| 83 | 	VM_REG_GUEST_FS, |
| 84 | 	VM_REG_GUEST_GS, |
| 85 | 	VM_REG_GUEST_LDTR, |
| 86 | 	VM_REG_GUEST_TR, |
| 87 | 	VM_REG_GUEST_IDTR, |
| 88 | 	VM_REG_GUEST_GDTR, |
| 89 | 	VM_REG_GUEST_EFER, |
| 90 | 	VM_REG_GUEST_CR2, |
| 91 | 	VM_REG_GUEST_PDPTE0, |
| 92 | 	VM_REG_GUEST_PDPTE1, |
| 93 | 	VM_REG_GUEST_PDPTE2, |
| 94 | 	VM_REG_GUEST_PDPTE3, |
| 95 | 	VM_REG_GUEST_INTR_SHADOW, |
| 96 | 	VM_REG_GUEST_DR0, |
| 97 | 	VM_REG_GUEST_DR1, |
| 98 | 	VM_REG_GUEST_DR2, |
| 99 | 	VM_REG_GUEST_DR3, |
| 100 | 	VM_REG_GUEST_DR6, |
| 101 | 	VM_REG_GUEST_ENTRY_INST_LENGTH, |
| 102 | 	VM_REG_GUEST_FS_BASE, |
| 103 | 	VM_REG_GUEST_GS_BASE, |
| 104 | 	VM_REG_GUEST_KGS_BASE, |
| 105 | 	VM_REG_GUEST_TPR, |
| 106 | 	VM_REG_LAST |
| 107 | }; |
| 108 | |
| 109 | enum x2apic_state { |
| 110 | 	X2APIC_DISABLED, |
| 111 | 	X2APIC_ENABLED, |
| 112 | 	X2APIC_STATE_LAST |
| 113 | }; |
| 114 | |
| 115 | #define	VM_INTINFO_VECTOR(info)	((info) & 0xff) |
| 116 | #define	VM_INTINFO_DEL_ERRCODE	0x800 |
| 117 | #define	VM_INTINFO_RSVD		0x7ffff000 |
| 118 | #define	VM_INTINFO_VALID	0x80000000 |
| 119 | #define	VM_INTINFO_TYPE		0x700 |
| 120 | #define	VM_INTINFO_HWINTR	(0 << 8) |
| 121 | #define	VM_INTINFO_NMI		(2 << 8) |
| 122 | #define	VM_INTINFO_HWEXCEPTION	(3 << 8) |
| 123 | #define	VM_INTINFO_SWINTR	(4 << 8) |
| 124 | |
| 125 | /* |
| 126 | * The VM name has to fit into the pathname length constraints of devfs, |
| 127 | * governed primarily by SPECNAMELEN. The length is the total number of |
| 128 | * characters in the full path, relative to the mount point and not |
| 129 | * including any leading '/' characters. |
| 130 | * A prefix and a suffix are added to the name specified by the user. |
| 131 | * The prefix is usually "vmm/" or "vmm.io/", but can be a few characters |
| 132 | * longer for future use. |
| 133 | * The suffix is a string that identifies a bootrom image or some similar |
| 134 | * image that is attached to the VM. A separator character gets added to |
| 135 | * the suffix automatically when generating the full path, so it must be |
| 136 | * accounted for, reducing the effective length by 1. |
| 137 | * The effective length of a VM name is 229 bytes for FreeBSD 13 and 37 |
| 138 | * bytes for FreeBSD 12. A minimum length is set for safety and supports |
| 139 | * a SPECNAMELEN as small as 32 on old systems. |
| 140 | */ |
| 141 | #define VM_MAX_PREFIXLEN 10 |
| 142 | #define VM_MAX_SUFFIXLEN 15 |
| 143 | #define VM_MIN_NAMELEN 6 |
| 144 | #define VM_MAX_NAMELEN \ |
| 145 | (SPECNAMELEN - VM_MAX_PREFIXLEN - VM_MAX_SUFFIXLEN - 1) |
| 146 | |
| 147 | #ifdef _KERNEL |
| 148 | #include <sys/kassert.h> |
| 149 | |
| 150 | CTASSERT(VM_MAX_NAMELEN >= VM_MIN_NAMELEN); |
| 151 | |
| 152 | struct vm; |
| 153 | struct vm_exception; |
| 154 | struct vm_mem; |
| 155 | struct seg_desc; |
| 156 | struct vm_exit; |
| 157 | struct vm_run; |
| 158 | struct vhpet; |
| 159 | struct vioapic; |
| 160 | struct vlapic; |
| 161 | struct vmspace; |
| 162 | struct vm_object; |
| 163 | struct vm_guest_paging; |
| 164 | struct pmap; |
| 165 | enum snapshot_req; |
| 166 | |
| 167 | struct vm_eventinfo { |
| 168 | 	cpuset_t *rptr;		/* rendezvous cookie */ |
| 169 | 	int	*sptr;		/* suspend cookie */ |
| 170 | 	int	*iptr;		/* reqidle cookie */ |
| 171 | }; |
| 172 | |
| 173 | typedef int	(*vmm_init_func_t)(int ipinum); |
| 174 | typedef int	(*vmm_cleanup_func_t)(void); |
| 175 | typedef void	(*vmm_suspend_func_t)(void); |
| 176 | typedef void	(*vmm_resume_func_t)(void); |
| 177 | typedef void *	(*vmi_init_func_t)(struct vm *vm, struct pmap *pmap); |
| 178 | typedef int	(*vmi_run_func_t)(void *vcpui, register_t rip, |
| 179 | 		 struct pmap *pmap, struct vm_eventinfo *info); |
| 180 | typedef void	(*vmi_cleanup_func_t)(void *vmi); |
| 181 | typedef void *	(*vmi_vcpu_init_func_t)(void *vmi, struct vcpu *vcpu, |
| 182 | 		 int vcpu_id); |
| 183 | typedef void	(*vmi_vcpu_cleanup_func_t)(void *vcpui); |
| 184 | typedef int	(*vmi_get_register_t)(void *vcpui, int num, uint64_t *retval); |
| 185 | typedef int	(*vmi_set_register_t)(void *vcpui, int num, uint64_t val); |
| 186 | typedef int	(*vmi_get_desc_t)(void *vcpui, int num, struct seg_desc *desc); |
| 187 | typedef int	(*vmi_set_desc_t)(void *vcpui, int num, struct seg_desc *desc); |
| 188 | typedef int	(*vmi_get_cap_t)(void *vcpui, int num, int *retval); |
| 189 | typedef int	(*vmi_set_cap_t)(void *vcpui, int num, int val); |
| 190 | typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max); |
| 191 | typedef void	(*vmi_vmspace_free)(struct vmspace *vmspace); |
| 192 | typedef struct vlapic * (*vmi_vlapic_init)(void *vcpui); |
| 193 | typedef void	(*vmi_vlapic_cleanup)(struct vlapic *vlapic); |
| 194 | typedef int	(*vmi_snapshot_vcpu_t)(void *vcpui, struct vm_snapshot_meta *meta); |
| 195 | typedef int	(*vmi_restore_tsc_t)(void *vcpui, uint64_t now); |
| 196 | |
| 197 | struct vmm_ops { |
| 198 | 	vmm_init_func_t		modinit;	/* module wide initialization */ |
| 199 | 	vmm_cleanup_func_t	modcleanup; |
| 200 | 	vmm_resume_func_t	modsuspend; |
| 201 | 	vmm_resume_func_t	modresume; |
| 202 | |
| 203 | 	vmi_init_func_t		init;		/* vm-specific initialization */ |
| 204 | 	vmi_run_func_t		run; |
| 205 | 	vmi_cleanup_func_t	cleanup; |
| 206 | 	vmi_vcpu_init_func_t	vcpu_init; |
| 207 | 	vmi_vcpu_cleanup_func_t	vcpu_cleanup; |
| 208 | 	vmi_get_register_t	getreg; |
| 209 | 	vmi_set_register_t	setreg; |
| 210 | 	vmi_get_desc_t		getdesc; |
| 211 | 	vmi_set_desc_t		setdesc; |
| 212 | 	vmi_get_cap_t		getcap; |
| 213 | 	vmi_set_cap_t		setcap; |
| 214 | 	vmi_vmspace_alloc	vmspace_alloc; |
| 215 | 	vmi_vmspace_free	vmspace_free; |
| 216 | 	vmi_vlapic_init		vlapic_init; |
| 217 | 	vmi_vlapic_cleanup	vlapic_cleanup; |
| 218 | |
| 219 | 	/* checkpoint operations */ |
| 220 | 	vmi_snapshot_vcpu_t	vcpu_snapshot; |
| 221 | 	vmi_restore_tsc_t	restore_tsc; |
| 222 | }; |
| 223 | |
| 224 | extern const struct vmm_ops vmm_ops_intel; |
| 225 | extern const struct vmm_ops vmm_ops_amd; |
| 226 | |
| 227 | extern u_int vm_maxcpu;			/* maximum virtual cpus */ |
| 228 | |
| 229 | int vm_create(const char *name, struct vm **retvm); |
| 230 | struct vcpu *vm_alloc_vcpu(struct vm *vm, int vcpuid); |
| 231 | void vm_disable_vcpu_creation(struct vm *vm); |
| 232 | void vm_slock_vcpus(struct vm *vm); |
| 233 | void vm_unlock_vcpus(struct vm *vm); |
| 234 | void vm_destroy(struct vm *vm); |
| 235 | int vm_reinit(struct vm *vm); |
| 236 | const char *vm_name(struct vm *vm); |
| 237 | uint16_t vm_get_maxcpus(struct vm *vm); |
| 238 | void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores, |
| 239 | uint16_t *threads, uint16_t *maxcpus); |
| 240 | int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores, |
| 241 | uint16_t threads, uint16_t maxcpus); |
| 242 | |
| 243 | int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa); |
| 244 | int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len); |
| 245 | int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func); |
| 246 | int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func); |
| 247 | |
| 248 | int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval); |
| 249 | int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val); |
| 250 | int vm_get_seg_desc(struct vcpu *vcpu, int reg, |
| 251 | 		 struct seg_desc *ret_desc); |
| 252 | int vm_set_seg_desc(struct vcpu *vcpu, int reg, |
| 253 | 		 struct seg_desc *desc); |
| 254 | int vm_run(struct vcpu *vcpu); |
| 255 | int vm_suspend(struct vm *vm, enum vm_suspend_how how); |
| 256 | int vm_inject_nmi(struct vcpu *vcpu); |
| 257 | int vm_nmi_pending(struct vcpu *vcpu); |
| 258 | void vm_nmi_clear(struct vcpu *vcpu); |
| 259 | int vm_inject_extint(struct vcpu *vcpu); |
| 260 | int vm_extint_pending(struct vcpu *vcpu); |
| 261 | void vm_extint_clear(struct vcpu *vcpu); |
| 262 | int vcpu_vcpuid(struct vcpu *vcpu); |
| 263 | struct vm *vcpu_vm(struct vcpu *vcpu); |
| 264 | struct vcpu *vm_vcpu(struct vm *vm, int cpu); |
| 265 | struct vlapic *vm_lapic(struct vcpu *vcpu); |
| 266 | struct vioapic *vm_ioapic(struct vm *vm); |
| 267 | struct vhpet *vm_hpet(struct vm *vm); |
| 268 | int vm_get_capability(struct vcpu *vcpu, int type, int *val); |
| 269 | int vm_set_capability(struct vcpu *vcpu, int type, int val); |
| 270 | int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state); |
| 271 | int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state); |
| 272 | int vm_apicid2vcpuid(struct vm *vm, int apicid); |
| 273 | int vm_activate_cpu(struct vcpu *vcpu); |
| 274 | int vm_suspend_cpu(struct vm *vm, struct vcpu *vcpu); |
| 275 | int vm_resume_cpu(struct vm *vm, struct vcpu *vcpu); |
| 276 | int vm_restart_instruction(struct vcpu *vcpu); |
| 277 | struct vm_exit *vm_exitinfo(struct vcpu *vcpu); |
| 278 | cpuset_t *vm_exitinfo_cpuset(struct vcpu *vcpu); |
| 279 | void vm_exit_suspended(struct vcpu *vcpu, uint64_t rip); |
| 280 | void vm_exit_debug(struct vcpu *vcpu, uint64_t rip); |
| 281 | void vm_exit_rendezvous(struct vcpu *vcpu, uint64_t rip); |
| 282 | void vm_exit_astpending(struct vcpu *vcpu, uint64_t rip); |
| 283 | void vm_exit_reqidle(struct vcpu *vcpu, uint64_t rip); |
| 284 | int vm_snapshot_req(struct vm *vm, struct vm_snapshot_meta *meta); |
| 285 | int vm_restore_time(struct vm *vm); |
| 286 | |
| 287 | #ifdef _SYS__CPUSET_H_ |
| 288 | /* |
| 289 | * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'. |
| 290 | * The rendezvous 'func(arg)' is not allowed to do anything that will |
| 291 | * cause the thread to be put to sleep. |
| 292 | * |
| 293 | * The caller cannot hold any locks when initiating the rendezvous. |
| 294 | * |
| 295 | * The implementation of this API may cause vcpus other than those specified |
| 296 | * by 'dest' to be stalled. The caller should not rely on any vcpus making |
| 297 | * forward progress when the rendezvous is in progress. |
| 298 | */ |
| 299 | typedef void (*vm_rendezvous_func_t)(struct vcpu *vcpu, void *arg); |
| 300 | int vm_smp_rendezvous(struct vcpu *vcpu, cpuset_t dest, |
| 301 | vm_rendezvous_func_t func, void *arg); |
| 302 | |
| 303 | cpuset_t vm_active_cpus(struct vm *vm); |
| 304 | cpuset_t vm_debug_cpus(struct vm *vm); |
| 305 | cpuset_t vm_suspended_cpus(struct vm *vm); |
| 306 | cpuset_t vm_start_cpus(struct vm *vm, const cpuset_t *tostart); |
| 307 | void vm_await_start(struct vm *vm, const cpuset_t *waiting); |
| 308 | #endif	/* _SYS__CPUSET_H_ */ |
| 309 | |
| 310 | static __inline int |
| 311 | vcpu_rendezvous_pending(struct vcpu *vcpu, struct vm_eventinfo *info) |
| 312 | { |
| 313 | 	/* |
| 314 | 	 * This check isn't done with atomic operations or under a lock because |
| 315 | 	 * there's no need to. If the vcpuid bit is set, the vcpu is part of a |
| 316 | 	 * rendezvous and the bit won't be cleared until the vcpu enters the |
| 317 | 	 * rendezvous. On rendezvous exit, the cpuset is cleared and the vcpu |
| 318 | 	 * will see an empty cpuset. So, the races are harmless. |
| 319 | 	 */ |
| 320 | 	return (CPU_ISSET(vcpu_vcpuid(vcpu), info->rptr)); |
| 321 | } |
| 322 | |
| 323 | static __inline int |
| 324 | vcpu_suspended(struct vm_eventinfo *info) |
| 325 | { |
| 326 | |
| 327 | 	return (*info->sptr); |
| 328 | } |
| 329 | |
| 330 | static __inline int |
| 331 | vcpu_reqidle(struct vm_eventinfo *info) |
| 332 | { |
| 333 | |
| 334 | 	return (*info->iptr); |
| 335 | } |
| 336 | |
| 337 | int vcpu_debugged(struct vcpu *vcpu); |
| 338 | |
| 339 | /* |
| 340 | * Return true if device indicated by bus/slot/func is supposed to be a |
| 341 | * pci passthrough device. |
| 342 | * |
| 343 | * Return false otherwise. |
| 344 | */ |
| 345 | bool vmm_is_pptdev(int bus, int slot, int func); |
| 346 | |
| 347 | void *vm_iommu_domain(struct vm *vm); |
| 348 | |
| 349 | enum vcpu_state { |
| 350 | 	VCPU_IDLE, |
| 351 | 	VCPU_FROZEN, |
| 352 | 	VCPU_RUNNING, |
| 353 | 	VCPU_SLEEPING, |
| 354 | }; |
| 355 | |
| 356 | int vcpu_set_state(struct vcpu *vcpu, enum vcpu_state state, bool from_idle); |
| 357 | enum vcpu_state vcpu_get_state(struct vcpu *vcpu, int *hostcpu); |
| 358 | |
| 359 | static int __inline |
| 360 | vcpu_is_running(struct vcpu *vcpu, int *hostcpu) |
| 361 | { |
| 362 | 	return (vcpu_get_state(vcpu, hostcpu) == VCPU_RUNNING); |
| 363 | } |
| 364 | |
| 365 | #ifdef _SYS_PROC_H_ |
| 366 | static int __inline |
| 367 | vcpu_should_yield(struct vcpu *vcpu) |
| 368 | { |
| 369 | 	struct thread *td; |
| 370 | |
| 371 | 	td = curthread; |
| 372 | 	return (td->td_ast != 0 || td->td_owepreempt != 0); |
| 373 | } |
| 374 | #endif |
| 375 | |
| 376 | void *vcpu_stats(struct vcpu *vcpu); |
| 377 | void vcpu_notify_event(struct vcpu *vcpu, bool lapic_intr); |
| 378 | struct vmspace *vm_vmspace(struct vm *vm); |
| 379 | struct vm_mem *vm_mem(struct vm *vm); |
| 380 | struct vatpic *vm_atpic(struct vm *vm); |
| 381 | struct vatpit *vm_atpit(struct vm *vm); |
| 382 | struct vpmtmr *vm_pmtmr(struct vm *vm); |
| 383 | struct vrtc *vm_rtc(struct vm *vm); |
| 384 | |
| 385 | /* |
| 386 | * Inject exception 'vector' into the guest vcpu. This function returns 0 on |
| 387 | * success and non-zero on failure. |
| 388 | * |
| 389 | * Wrapper functions like 'vm_inject_gp()' should be preferred to calling |
| 390 | * this function directly because they enforce the trap-like or fault-like |
| 391 | * behavior of an exception. |
| 392 | * |
| 393 | * This function should only be called in the context of the thread that is |
| 394 | * executing this vcpu. |
| 395 | */ |
| 396 | int vm_inject_exception(struct vcpu *vcpu, int vector, int err_valid, |
| 397 | uint32_t errcode, int restart_instruction); |
| 398 | |
| 399 | /* |
| 400 | * This function is called after a VM-exit that occurred during exception or |
| 401 | * interrupt delivery through the IDT. The format of 'intinfo' is described |
| 402 | * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2. |
| 403 | * |
| 404 | * If a VM-exit handler completes the event delivery successfully then it |
| 405 | * should call vm_exit_intinfo() to extinguish the pending event. For e.g., |
| 406 | * if the task switch emulation is triggered via a task gate then it should |
| 407 | * call this function with 'intinfo=0' to indicate that the external event |
| 408 | * is not pending anymore. |
| 409 | * |
| 410 | * Return value is 0 on success and non-zero on failure. |
| 411 | */ |
| 412 | int vm_exit_intinfo(struct vcpu *vcpu, uint64_t intinfo); |
| 413 | |
| 414 | /* |
| 415 | * This function is called before every VM-entry to retrieve a pending |
| 416 | * event that should be injected into the guest. This function combines |
| 417 | * nested events into a double or triple fault. |
| 418 | * |
| 419 | * Returns 0 if there are no events that need to be injected into the guest |
| 420 | * and non-zero otherwise. |
| 421 | */ |
| 422 | int vm_entry_intinfo(struct vcpu *vcpu, uint64_t *info); |
| 423 | |
| 424 | int vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2); |
| 425 | |
| 426 | /* |
| 427 | * Function used to keep track of the guest's TSC offset. The |
| 428 | * offset is used by the virtualization extensions to provide a consistent |
| 429 | * value for the Time Stamp Counter to the guest. |
| 430 | */ |
| 431 | void vm_set_tsc_offset(struct vcpu *vcpu, uint64_t offset); |
| 432 | |
| 433 | enum vm_reg_name vm_segment_name(int seg_encoding); |
| 434 | |
| 435 | struct vm_copyinfo { |
| 436 | 	uint64_t	gpa; |
| 437 | 	size_t		len; |
| 438 | 	void		*hva; |
| 439 | 	void		*cookie; |
| 440 | }; |
| 441 | |
| 442 | /* |
| 443 | * Set up 'copyinfo[]' to copy to/from guest linear address space starting |
| 444 | * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for |
| 445 | * a copyin or PROT_WRITE for a copyout. |
| 446 | * |
| 447 | * retval	is_fault	Interpretation |
| 448 | * 0		 0		Success |
| 449 | * 0		 1		An exception was injected into the guest |
| 450 | * EFAULT	 N/A		Unrecoverable error |
| 451 | * |
| 452 | * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if |
| 453 | * the return value is 0. The 'copyinfo[]' resources should be freed by calling |
| 454 | * 'vm_copy_teardown()' after the copy is done. |
| 455 | */ |
| 456 | int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging, |
| 457 | uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo, |
| 458 | int num_copyinfo, int *is_fault); |
| 459 | void vm_copy_teardown(struct vm_copyinfo *copyinfo, int num_copyinfo); |
| 460 | void vm_copyin(struct vm_copyinfo *copyinfo, void *kaddr, size_t len); |
| 461 | void vm_copyout(const void *kaddr, struct vm_copyinfo *copyinfo, size_t len); |
| 462 | |
| 463 | int vcpu_trace_exceptions(struct vcpu *vcpu); |
| 464 | int vcpu_trap_wbinvd(struct vcpu *vcpu); |
| 465 | #endif	/* KERNEL */ |
| 466 | |
| 467 | /* |
| 468 | * Identifiers for optional vmm capabilities |
| 469 | */ |
| 470 | enum vm_cap_type { |
| 471 | 	VM_CAP_HALT_EXIT, |
| 472 | 	VM_CAP_MTRAP_EXIT, |
| 473 | 	VM_CAP_PAUSE_EXIT, |
| 474 | 	VM_CAP_UNRESTRICTED_GUEST, |
| 475 | 	VM_CAP_ENABLE_INVPCID, |
| 476 | 	VM_CAP_BPT_EXIT, |
| 477 | 	VM_CAP_RDPID, |
| 478 | 	VM_CAP_RDTSCP, |
| 479 | 	VM_CAP_IPI_EXIT, |
| 480 | 	VM_CAP_MASK_HWINTR, |
| 481 | 	VM_CAP_RFLAGS_TF, |
| 482 | 	VM_CAP_MAX |
| 483 | }; |
| 484 | |
| 485 | enum vm_intr_trigger { |
| 486 | 	EDGE_TRIGGER, |
| 487 | 	LEVEL_TRIGGER |
| 488 | }; |
| 489 | |
| 490 | /* |
| 491 | * The 'access' field has the format specified in Table 21-2 of the Intel |
| 492 | * Architecture Manual vol 3b. |
| 493 | * |
| 494 | * XXX The contents of the 'access' field are architecturally defined except |
| 495 | * bit 16 - Segment Unusable. |
| 496 | */ |
| 497 | struct seg_desc { |
| 498 | 	uint64_t	base; |
| 499 | 	uint32_t	limit; |
| 500 | 	uint32_t	access; |
| 501 | }; |
| 502 | #define	SEG_DESC_TYPE(access)		((access) & 0x001f) |
| 503 | #define	SEG_DESC_DPL(access)		(((access) >> 5) & 0x3) |
| 504 | #define	SEG_DESC_PRESENT(access)	(((access) & 0x0080) ? 1 : 0) |
| 505 | #define	SEG_DESC_DEF32(access)		(((access) & 0x4000) ? 1 : 0) |
| 506 | #define	SEG_DESC_GRANULARITY(access)	(((access) & 0x8000) ? 1 : 0) |
| 507 | #define	SEG_DESC_UNUSABLE(access)	(((access) & 0x10000) ? 1 : 0) |
| 508 | |
| 509 | enum vm_cpu_mode { |
| 510 | 	CPU_MODE_REAL, |
| 511 | 	CPU_MODE_PROTECTED, |
| 512 | 	CPU_MODE_COMPATIBILITY,		/* IA-32E mode (CS.L = 0) */ |
| 513 | 	CPU_MODE_64BIT,			/* IA-32E mode (CS.L = 1) */ |
| 514 | }; |
| 515 | |
| 516 | enum vm_paging_mode { |
| 517 | 	PAGING_MODE_FLAT, |
| 518 | 	PAGING_MODE_32, |
| 519 | 	PAGING_MODE_PAE, |
| 520 | 	PAGING_MODE_64, |
| 521 | 	PAGING_MODE_64_LA57, |
| 522 | }; |
| 523 | |
| 524 | struct vm_guest_paging { |
| 525 | 	uint64_t	cr3; |
| 526 | 	int		cpl; |
| 527 | 	enum vm_cpu_mode cpu_mode; |
| 528 | 	enum vm_paging_mode paging_mode; |
| 529 | }; |
| 530 | |
| 531 | /* |
| 532 | * The data structures 'vie' and 'vie_op' are meant to be opaque to the |
| 533 | * consumers of instruction decoding. The only reason why their contents |
| 534 | * need to be exposed is because they are part of the 'vm_exit' structure. |
| 535 | */ |
| 536 | struct vie_op { |
| 537 | 	uint8_t		op_byte;	/* actual opcode byte */ |
| 538 | 	uint8_t		op_type;	/* type of operation (e.g. MOV) */ |
| 539 | 	uint16_t	op_flags; |
| 540 | }; |
| 541 | _Static_assert(sizeof(struct vie_op) == 4, "ABI"); |
| 542 | _Static_assert(_Alignof(struct vie_op) == 2, "ABI"); |
| 543 | |
| 544 | #define	VIE_INST_SIZE	15 |
| 545 | struct vie { |
| 546 | 	uint8_t		inst[VIE_INST_SIZE];	/* instruction bytes */ |
| 547 | 	uint8_t		num_valid;		/* size of the instruction */ |
| 548 | |
| 549 | /* The following fields are all zeroed upon restart. */ |
| 550 | #define	vie_startzero	num_processed |
| 551 | 	uint8_t		num_processed; |
| 552 | |
| 553 | 	uint8_t		addrsize:4, opsize:4;	/* address and operand sizes */ |
| 554 | 	uint8_t		rex_w:1,		/* REX prefix */ |
| 555 | 			rex_r:1, |
| 556 | 			rex_x:1, |
| 557 | 			rex_b:1, |
| 558 | 			rex_present:1, |
| 559 | 			repz_present:1,		/* REP/REPE/REPZ prefix */ |
| 560 | 			repnz_present:1,	/* REPNE/REPNZ prefix */ |
| 561 | 			opsize_override:1,	/* Operand size override */ |
| 562 | 			addrsize_override:1,	/* Address size override */ |
| 563 | 			segment_override:1;	/* Segment override */ |
| 564 | |
| 565 | 	uint8_t		mod:2,			/* ModRM byte */ |
| 566 | 			reg:4, |
| 567 | 			rm:4; |
| 568 | |
| 569 | 	uint8_t		ss:2,			/* SIB byte */ |
| 570 | 			vex_present:1,		/* VEX prefixed */ |
| 571 | 			vex_l:1,		/* L bit */ |
| 572 | 			index:4,		/* SIB byte */ |
| 573 | 			base:4;			/* SIB byte */ |
| 574 | |
| 575 | 	uint8_t		disp_bytes; |
| 576 | 	uint8_t		imm_bytes; |
| 577 | |
| 578 | 	uint8_t		scale; |
| 579 | |
| 580 | 	uint8_t		vex_reg:4,		/* vvvv: first source register specifier */ |
| 581 | 			vex_pp:2,		/* pp */ |
| 582 | 			_sparebits:2; |
| 583 | |
| 584 | 	uint8_t		_sparebytes[2]; |
| 585 | |
| 586 | 	int		base_register;		/* VM_REG_GUEST_xyz */ |
| 587 | 	int		index_register;		/* VM_REG_GUEST_xyz */ |
| 588 | 	int		segment_register;	/* VM_REG_GUEST_xyz */ |
| 589 | |
| 590 | 	int64_t		displacement;		/* optional addr displacement */ |
| 591 | 	int64_t		immediate;		/* optional immediate operand */ |
| 592 | |
| 593 | 	uint8_t		decoded;	/* set to 1 if successfully decoded */ |
| 594 | |
| 595 | 	uint8_t		_sparebyte; |
| 596 | |
| 597 | 	struct vie_op	op;			/* opcode description */ |
| 598 | }; |
| 599 | _Static_assert(sizeof(struct vie) == 64, "ABI"); |
| 600 | _Static_assert(__offsetof(struct vie, disp_bytes) == 22, "ABI"); |
| 601 | _Static_assert(__offsetof(struct vie, scale) == 24, "ABI"); |
| 602 | _Static_assert(__offsetof(struct vie, base_register) == 28, "ABI"); |
| 603 | |
| 604 | enum vm_exitcode { |
| 605 | 	VM_EXITCODE_INOUT, |
| 606 | 	VM_EXITCODE_VMX, |
| 607 | 	VM_EXITCODE_BOGUS, |
| 608 | 	VM_EXITCODE_RDMSR, |
| 609 | 	VM_EXITCODE_WRMSR, |
| 610 | 	VM_EXITCODE_HLT, |
| 611 | 	VM_EXITCODE_MTRAP, |
| 612 | 	VM_EXITCODE_PAUSE, |
| 613 | 	VM_EXITCODE_PAGING, |
| 614 | 	VM_EXITCODE_INST_EMUL, |
| 615 | 	VM_EXITCODE_SPINUP_AP, |
| 616 | 	VM_EXITCODE_DEPRECATED1,	/* used to be SPINDOWN_CPU */ |
| 617 | 	VM_EXITCODE_RENDEZVOUS, |
| 618 | 	VM_EXITCODE_IOAPIC_EOI, |
| 619 | 	VM_EXITCODE_SUSPENDED, |
| 620 | 	VM_EXITCODE_INOUT_STR, |
| 621 | 	VM_EXITCODE_TASK_SWITCH, |
| 622 | 	VM_EXITCODE_MONITOR, |
| 623 | 	VM_EXITCODE_MWAIT, |
| 624 | 	VM_EXITCODE_SVM, |
| 625 | 	VM_EXITCODE_REQIDLE, |
| 626 | 	VM_EXITCODE_DEBUG, |
| 627 | 	VM_EXITCODE_VMINSN, |
| 628 | 	VM_EXITCODE_BPT, |
| 629 | 	VM_EXITCODE_IPI, |
| 630 | 	VM_EXITCODE_DB, |
| 631 | 	VM_EXITCODE_MAX |
| 632 | }; |
| 633 | |
| 634 | struct vm_inout { |
| 635 | 	uint16_t	bytes:3;	/* 1 or 2 or 4 */ |
| 636 | 	uint16_t	in:1; |
| 637 | 	uint16_t	string:1; |
| 638 | 	uint16_t	rep:1; |
| 639 | 	uint16_t	port; |
| 640 | 	uint32_t	eax;		/* valid for out */ |
| 641 | }; |
| 642 | |
| 643 | struct vm_inout_str { |
| 644 | 	struct vm_inout	inout;		/* must be the first element */ |
| 645 | 	struct vm_guest_paging paging; |
| 646 | 	uint64_t	rflags; |
| 647 | 	uint64_t	cr0; |
| 648 | 	uint64_t	index; |
| 649 | 	uint64_t	count;		/* rep=1 (%rcx), rep=0 (1) */ |
| 650 | 	int		addrsize; |
| 651 | 	enum vm_reg_name seg_name; |
| 652 | 	struct seg_desc seg_desc; |
| 653 | 	int		cs_d; |
| 654 | 	uint64_t	cs_base; |
| 655 | }; |
| 656 | |
| 657 | enum task_switch_reason { |
| 658 | 	TSR_CALL, |
| 659 | 	TSR_IRET, |
| 660 | 	TSR_JMP, |
| 661 | 	TSR_IDT_GATE,	/* task gate in IDT */ |
| 662 | }; |
| 663 | |
| 664 | struct vm_task_switch { |
| 665 | 	uint16_t	tsssel;		/* new TSS selector */ |
| 666 | 	int		ext;		/* task switch due to external event */ |
| 667 | 	uint32_t	errcode; |
| 668 | 	int		errcode_valid;	/* push 'errcode' on the new stack */ |
| 669 | 	enum task_switch_reason reason; |
| 670 | 	struct vm_guest_paging paging; |
| 671 | }; |
| 672 | |
| 673 | struct vm_exit { |
| 674 | 	enum vm_exitcode	exitcode; |
| 675 | 	int			inst_length;	/* 0 means unknown */ |
| 676 | 	uint64_t		rip; |
| 677 | 	union { |
| 678 | 		struct vm_inout	inout; |
| 679 | 		struct vm_inout_str inout_str; |
| 680 | 		struct { |
| 681 | 			uint64_t	gpa; |
| 682 | 			int		fault_type; |
| 683 | 		} paging; |
| 684 | 		struct { |
| 685 | 			uint64_t	gpa; |
| 686 | 			uint64_t	gla; |
| 687 | 			uint64_t	cs_base; |
| 688 | 			int		cs_d;		/* CS.D */ |
| 689 | 			struct vm_guest_paging paging; |
| 690 | 			struct vie	vie; |
| 691 | 		} inst_emul; |
| 692 | 		/* |
| 693 | 		 * VMX specific payload. Used when there is no "better" |
| 694 | 		 * exitcode to represent the VM-exit. |
| 695 | 		 */ |
| 696 | 		struct { |
| 697 | 			int		status;		/* vmx inst status */ |
| 698 | 			/* |
| 699 | 			 * 'exit_reason' and 'exit_qualification' are valid |
| 700 | 			 * only if 'status' is zero. |
| 701 | 			 */ |
| 702 | 			uint32_t	exit_reason; |
| 703 | 			uint64_t	exit_qualification; |
| 704 | 			/* |
| 705 | 			 * 'inst_error' and 'inst_type' are valid |
| 706 | 			 * only if 'status' is non-zero. |
| 707 | 			 */ |
| 708 | 			int		inst_type; |
| 709 | 			int		inst_error; |
| 710 | 		} vmx; |
| 711 | 		/* |
| 712 | 		 * SVM specific payload. |
| 713 | 		 */ |
| 714 | 		struct { |
| 715 | 			uint64_t	exitcode; |
| 716 | 			uint64_t	exitinfo1; |
| 717 | 			uint64_t	exitinfo2; |
| 718 | 		} svm; |
| 719 | 		struct { |
| 720 | 			int		inst_length; |
| 721 | 		} bpt; |
| 722 | 		struct { |
| 723 | 			int		trace_trap; |
| 724 | 			int		pushf_intercept; |
| 725 | 			int		tf_shadow_val; |
| 726 | 			struct		vm_guest_paging paging; |
| 727 | 		} dbg; |
| 728 | 		struct { |
| 729 | 			uint32_t	code;		/* ecx value */ |
| 730 | 			uint64_t	wval; |
| 731 | 		} msr; |
| 732 | 		struct { |
| 733 | 			int		vcpu; |
| 734 | 			uint64_t	rip; |
| 735 | 		} spinup_ap; |
| 736 | 		struct { |
| 737 | 			uint64_t	rflags; |
| 738 | 			uint64_t	intr_status; |
| 739 | 		} hlt; |
| 740 | 		struct { |
| 741 | 			int		vector; |
| 742 | 		} ioapic_eoi; |
| 743 | 		struct { |
| 744 | 			enum vm_suspend_how how; |
| 745 | 		} suspended; |
| 746 | 		struct { |
| 747 | 			/* |
| 748 | 			 * The destination vCPU mask is saved in vcpu->cpuset |
| 749 | 			 * and is copied out to userspace separately to avoid |
| 750 | 			 * ABI concerns. |
| 751 | 			 */ |
| 752 | 			uint32_t mode; |
| 753 | 			uint8_t vector; |
| 754 | 		} ipi; |
| 755 | 		struct vm_task_switch task_switch; |
| 756 | 	} u; |
| 757 | }; |
| 758 | |
| 759 | /* APIs to inject faults into the guest */ |
| 760 | void vm_inject_fault(struct vcpu *vcpu, int vector, int errcode_valid, |
| 761 | int errcode); |
| 762 | |
| 763 | static __inline void |
| 764 | vm_inject_ud(struct vcpu *vcpu) |
| 765 | { |
| 766 | 	vm_inject_fault(vcpu, IDT_UD, 0, 0); |
| 767 | } |
| 768 | |
| 769 | static __inline void |
| 770 | vm_inject_gp(struct vcpu *vcpu) |
| 771 | { |
| 772 | 	vm_inject_fault(vcpu, IDT_GP, 1, 0); |
| 773 | } |
| 774 | |
| 775 | static __inline void |
| 776 | vm_inject_ac(struct vcpu *vcpu, int errcode) |
| 777 | { |
| 778 | 	vm_inject_fault(vcpu, IDT_AC, 1, errcode); |
| 779 | } |
| 780 | |
| 781 | static __inline void |
| 782 | vm_inject_ss(struct vcpu *vcpu, int errcode) |
| 783 | { |
| 784 | 	vm_inject_fault(vcpu, IDT_SS, 1, errcode); |
| 785 | } |
| 786 | |
| 787 | void vm_inject_pf(struct vcpu *vcpu, int error_code, uint64_t cr2); |
| 788 | |
| 789 | #endif	/* _VMM_H_ */ |