2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 NetApp, Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
34 #include <x86/segments.h>
41 VM_SUSPEND_TRIPLEFAULT,
46 * Identifiers for architecturally defined registers.
87 VM_REG_GUEST_INTR_SHADOW,
102 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
103 #define VM_INTINFO_DEL_ERRCODE 0x800
104 #define VM_INTINFO_RSVD 0x7ffff000
105 #define VM_INTINFO_VALID 0x80000000
106 #define VM_INTINFO_TYPE 0x700
107 #define VM_INTINFO_HWINTR (0 << 8)
108 #define VM_INTINFO_NMI (2 << 8)
109 #define VM_INTINFO_HWEXCEPTION (3 << 8)
110 #define VM_INTINFO_SWINTR (4 << 8)
114 #define VM_MAX_NAMELEN 32
126 struct vm_guest_paging;
129 struct vm_eventinfo {
130 void *rptr; /* rendezvous cookie */
131 int *sptr; /* suspend cookie */
132 int *iptr; /* reqidle cookie */
135 typedef int (*vmm_init_func_t)(int ipinum);
136 typedef int (*vmm_cleanup_func_t)(void);
137 typedef void (*vmm_resume_func_t)(void);
138 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
139 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
140 struct pmap *pmap, struct vm_eventinfo *info);
141 typedef void (*vmi_cleanup_func_t)(void *vmi);
142 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num,
144 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num,
146 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num,
147 struct seg_desc *desc);
148 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num,
149 struct seg_desc *desc);
150 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
151 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
152 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
153 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
154 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
155 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
158 vmm_init_func_t init; /* module wide initialization */
159 vmm_cleanup_func_t cleanup;
160 vmm_resume_func_t resume;
162 vmi_init_func_t vminit; /* vm-specific initialization */
163 vmi_run_func_t vmrun;
164 vmi_cleanup_func_t vmcleanup;
165 vmi_get_register_t vmgetreg;
166 vmi_set_register_t vmsetreg;
167 vmi_get_desc_t vmgetdesc;
168 vmi_set_desc_t vmsetdesc;
169 vmi_get_cap_t vmgetcap;
170 vmi_set_cap_t vmsetcap;
171 vmi_vmspace_alloc vmspace_alloc;
172 vmi_vmspace_free vmspace_free;
173 vmi_vlapic_init vlapic_init;
174 vmi_vlapic_cleanup vlapic_cleanup;
177 extern struct vmm_ops vmm_ops_intel;
178 extern struct vmm_ops vmm_ops_amd;
180 int vm_create(const char *name, struct vm **retvm);
181 void vm_destroy(struct vm *vm);
182 int vm_reinit(struct vm *vm);
183 const char *vm_name(struct vm *vm);
186 * APIs that modify the guest memory map require all vcpus to be frozen.
188 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
189 size_t len, int prot, int flags);
190 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
191 void vm_free_memseg(struct vm *vm, int ident);
192 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
193 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
194 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
195 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
198 * APIs that inspect the guest memory map require only a *single* vcpu to
199 * be frozen. This acts like a read lock on the guest memory map since any
200 * modification requires *all* vcpus to be frozen.
202 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
203 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
204 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
205 struct vm_object **objptr);
206 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
207 int prot, void **cookie);
208 void vm_gpa_release(void *cookie);
209 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
211 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
212 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
213 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
214 struct seg_desc *ret_desc);
215 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
216 struct seg_desc *desc);
217 int vm_run(struct vm *vm, struct vm_run *vmrun);
218 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
219 int vm_inject_nmi(struct vm *vm, int vcpu);
220 int vm_nmi_pending(struct vm *vm, int vcpuid);
221 void vm_nmi_clear(struct vm *vm, int vcpuid);
222 int vm_inject_extint(struct vm *vm, int vcpu);
223 int vm_extint_pending(struct vm *vm, int vcpuid);
224 void vm_extint_clear(struct vm *vm, int vcpuid);
225 struct vlapic *vm_lapic(struct vm *vm, int cpu);
226 struct vioapic *vm_ioapic(struct vm *vm);
227 struct vhpet *vm_hpet(struct vm *vm);
228 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
229 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
230 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
231 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
232 int vm_apicid2vcpuid(struct vm *vm, int apicid);
233 int vm_activate_cpu(struct vm *vm, int vcpu);
234 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
235 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
236 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
237 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
238 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
240 #ifdef _SYS__CPUSET_H_
242 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
243 * The rendezvous 'func(arg)' is not allowed to do anything that will
244 * cause the thread to be put to sleep.
246 * If the rendezvous is being initiated from a vcpu context then the
247 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
249 * The caller cannot hold any locks when initiating the rendezvous.
251 * The implementation of this API may cause vcpus other than those specified
252 * by 'dest' to be stalled. The caller should not rely on any vcpus making
253 * forward progress when the rendezvous is in progress.
255 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
256 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
257 vm_rendezvous_func_t func, void *arg);
258 cpuset_t vm_active_cpus(struct vm *vm);
259 cpuset_t vm_suspended_cpus(struct vm *vm);
260 #endif /* _SYS__CPUSET_H_ */
263 vcpu_rendezvous_pending(struct vm_eventinfo *info)
266 return (*((uintptr_t *)(info->rptr)) != 0);
270 vcpu_suspended(struct vm_eventinfo *info)
273 return (*info->sptr);
277 vcpu_reqidle(struct vm_eventinfo *info)
280 return (*info->iptr);
284 * Return 1 if device indicated by bus/slot/func is supposed to be a
285 * pci passthrough device.
287 * Return 0 otherwise.
289 int vmm_is_pptdev(int bus, int slot, int func);
291 void *vm_iommu_domain(struct vm *vm);
300 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
302 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
305 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
307 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
312 vcpu_should_yield(struct vm *vm, int vcpu)
315 if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
317 else if (curthread->td_owepreempt)
324 void *vcpu_stats(struct vm *vm, int vcpu);
325 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
326 struct vmspace *vm_get_vmspace(struct vm *vm);
327 struct vatpic *vm_atpic(struct vm *vm);
328 struct vatpit *vm_atpit(struct vm *vm);
329 struct vpmtmr *vm_pmtmr(struct vm *vm);
330 struct vrtc *vm_rtc(struct vm *vm);
333 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
334 * success and non-zero on failure.
336 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
337 * this function directly because they enforce the trap-like or fault-like
338 * behavior of an exception.
340 * This function should only be called in the context of the thread that is
341 * executing this vcpu.
343 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
344 uint32_t errcode, int restart_instruction);
347 * This function is called after a VM-exit that occurred during exception or
348 * interrupt delivery through the IDT. The format of 'intinfo' is described
349 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
351 * If a VM-exit handler completes the event delivery successfully then it
352 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
353 * if the task switch emulation is triggered via a task gate then it should
354 * call this function with 'intinfo=0' to indicate that the external event
355 * is not pending anymore.
357 * Return value is 0 on success and non-zero on failure.
359 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
362 * This function is called before every VM-entry to retrieve a pending
363 * event that should be injected into the guest. This function combines
364 * nested events into a double or triple fault.
366 * Returns 0 if there are no events that need to be injected into the guest
367 * and non-zero otherwise.
369 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
371 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
373 enum vm_reg_name vm_segment_name(int seg_encoding);
383 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
384 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
385 * a copyin or PROT_WRITE for a copyout.
387 * retval is_fault Interpretation
389 * 0 1 An exception was injected into the guest
390 * EFAULT N/A Unrecoverable error
392 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
393 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
394 * 'vm_copy_teardown()' after the copy is done.
396 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
397 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
398 int num_copyinfo, int *is_fault);
399 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
401 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
402 void *kaddr, size_t len);
403 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
404 struct vm_copyinfo *copyinfo, size_t len);
406 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
409 #define VM_MAXCPU 16 /* maximum virtual cpus */
412 * Identifiers for optional vmm capabilities
418 VM_CAP_UNRESTRICTED_GUEST,
419 VM_CAP_ENABLE_INVPCID,
423 enum vm_intr_trigger {
429 * The 'access' field has the format specified in Table 21-2 of the Intel
430 * Architecture Manual vol 3b.
432 * XXX The contents of the 'access' field are architecturally defined except
433 * bit 16 - Segment Unusable.
440 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
441 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
442 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
443 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
444 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
445 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
450 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
451 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
454 enum vm_paging_mode {
461 struct vm_guest_paging {
464 enum vm_cpu_mode cpu_mode;
465 enum vm_paging_mode paging_mode;
469 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
470 * consumers of instruction decoding. The only reason why their contents
471 * need to be exposed is because they are part of the 'vm_exit' structure.
474 uint8_t op_byte; /* actual opcode byte */
475 uint8_t op_type; /* type of operation (e.g. MOV) */
479 #define VIE_INST_SIZE 15
481 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
482 uint8_t num_valid; /* size of the instruction */
483 uint8_t num_processed;
485 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
486 uint8_t rex_w:1, /* REX prefix */
491 repz_present:1, /* REP/REPE/REPZ prefix */
492 repnz_present:1, /* REPNE/REPNZ prefix */
493 opsize_override:1, /* Operand size override */
494 addrsize_override:1, /* Address size override */
495 segment_override:1; /* Segment override */
497 uint8_t mod:2, /* ModRM byte */
501 uint8_t ss:2, /* SIB byte */
509 int base_register; /* VM_REG_GUEST_xyz */
510 int index_register; /* VM_REG_GUEST_xyz */
511 int segment_register; /* VM_REG_GUEST_xyz */
513 int64_t displacement; /* optional addr displacement */
514 int64_t immediate; /* optional immediate operand */
516 uint8_t decoded; /* set to 1 if successfully decoded */
518 struct vie_op op; /* opcode description */
531 VM_EXITCODE_INST_EMUL,
532 VM_EXITCODE_SPINUP_AP,
533 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
534 VM_EXITCODE_RENDEZVOUS,
535 VM_EXITCODE_IOAPIC_EOI,
536 VM_EXITCODE_SUSPENDED,
537 VM_EXITCODE_INOUT_STR,
538 VM_EXITCODE_TASK_SWITCH,
547 uint16_t bytes:3; /* 1 or 2 or 4 */
552 uint32_t eax; /* valid for out */
555 struct vm_inout_str {
556 struct vm_inout inout; /* must be the first element */
557 struct vm_guest_paging paging;
561 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
563 enum vm_reg_name seg_name;
564 struct seg_desc seg_desc;
567 enum task_switch_reason {
571 TSR_IDT_GATE, /* task gate in IDT */
574 struct vm_task_switch {
575 uint16_t tsssel; /* new TSS selector */
576 int ext; /* task switch due to external event */
578 int errcode_valid; /* push 'errcode' on the new stack */
579 enum task_switch_reason reason;
580 struct vm_guest_paging paging;
584 enum vm_exitcode exitcode;
585 int inst_length; /* 0 means unknown */
588 struct vm_inout inout;
589 struct vm_inout_str inout_str;
599 struct vm_guest_paging paging;
603 * VMX specific payload. Used when there is no "better"
604 * exitcode to represent the VM-exit.
607 int status; /* vmx inst status */
609 * 'exit_reason' and 'exit_qualification' are valid
610 * only if 'status' is zero.
612 uint32_t exit_reason;
613 uint64_t exit_qualification;
615 * 'inst_error' and 'inst_type' are valid
616 * only if 'status' is non-zero.
622 * SVM specific payload.
630 uint32_t code; /* ecx value */
644 enum vm_suspend_how how;
646 struct vm_task_switch task_switch;
650 /* APIs to inject faults into the guest */
651 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
655 vm_inject_ud(void *vm, int vcpuid)
657 vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
661 vm_inject_gp(void *vm, int vcpuid)
663 vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
667 vm_inject_ac(void *vm, int vcpuid, int errcode)
669 vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
673 vm_inject_ss(void *vm, int vcpuid, int errcode)
675 vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
678 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
680 int vm_restart_instruction(void *vm, int vcpuid);