2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * 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.
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
32 #include <x86/segments.h>
39 VM_SUSPEND_TRIPLEFAULT,
44 * Identifiers for architecturally defined registers.
85 VM_REG_GUEST_INTR_SHADOW,
95 #define VM_INTINFO_VECTOR(info) ((info) & 0xff)
96 #define VM_INTINFO_DEL_ERRCODE 0x800
97 #define VM_INTINFO_RSVD 0x7ffff000
98 #define VM_INTINFO_VALID 0x80000000
99 #define VM_INTINFO_TYPE 0x700
100 #define VM_INTINFO_HWINTR (0 << 8)
101 #define VM_INTINFO_NMI (2 << 8)
102 #define VM_INTINFO_HWEXCEPTION (3 << 8)
103 #define VM_INTINFO_SWINTR (4 << 8)
107 #define VM_MAX_NAMELEN 32
119 struct vm_guest_paging;
122 struct vm_eventinfo {
123 void *rptr; /* rendezvous cookie */
124 int *sptr; /* suspend cookie */
125 int *iptr; /* reqidle cookie */
128 typedef int (*vmm_init_func_t)(int ipinum);
129 typedef int (*vmm_cleanup_func_t)(void);
130 typedef void (*vmm_resume_func_t)(void);
131 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
132 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, register_t rip,
133 struct pmap *pmap, struct vm_eventinfo *info);
134 typedef void (*vmi_cleanup_func_t)(void *vmi);
135 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num,
137 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num,
139 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num,
140 struct seg_desc *desc);
141 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num,
142 struct seg_desc *desc);
143 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
144 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
145 typedef struct vmspace * (*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
146 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
147 typedef struct vlapic * (*vmi_vlapic_init)(void *vmi, int vcpu);
148 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
151 vmm_init_func_t init; /* module wide initialization */
152 vmm_cleanup_func_t cleanup;
153 vmm_resume_func_t resume;
155 vmi_init_func_t vminit; /* vm-specific initialization */
156 vmi_run_func_t vmrun;
157 vmi_cleanup_func_t vmcleanup;
158 vmi_get_register_t vmgetreg;
159 vmi_set_register_t vmsetreg;
160 vmi_get_desc_t vmgetdesc;
161 vmi_set_desc_t vmsetdesc;
162 vmi_get_cap_t vmgetcap;
163 vmi_set_cap_t vmsetcap;
164 vmi_vmspace_alloc vmspace_alloc;
165 vmi_vmspace_free vmspace_free;
166 vmi_vlapic_init vlapic_init;
167 vmi_vlapic_cleanup vlapic_cleanup;
170 extern struct vmm_ops vmm_ops_intel;
171 extern struct vmm_ops vmm_ops_amd;
173 int vm_create(const char *name, struct vm **retvm);
174 void vm_destroy(struct vm *vm);
175 int vm_reinit(struct vm *vm);
176 const char *vm_name(struct vm *vm);
179 * APIs that modify the guest memory map require all vcpus to be frozen.
181 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
182 size_t len, int prot, int flags);
183 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
184 void vm_free_memseg(struct vm *vm, int ident);
185 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
186 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
187 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
188 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
191 * APIs that inspect the guest memory map require only a *single* vcpu to
192 * be frozen. This acts like a read lock on the guest memory map since any
193 * modification requires *all* vcpus to be frozen.
195 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
196 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
197 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
198 struct vm_object **objptr);
199 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
200 int prot, void **cookie);
201 void vm_gpa_release(void *cookie);
202 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
204 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
205 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
206 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
207 struct seg_desc *ret_desc);
208 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
209 struct seg_desc *desc);
210 int vm_run(struct vm *vm, struct vm_run *vmrun);
211 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
212 int vm_inject_nmi(struct vm *vm, int vcpu);
213 int vm_nmi_pending(struct vm *vm, int vcpuid);
214 void vm_nmi_clear(struct vm *vm, int vcpuid);
215 int vm_inject_extint(struct vm *vm, int vcpu);
216 int vm_extint_pending(struct vm *vm, int vcpuid);
217 void vm_extint_clear(struct vm *vm, int vcpuid);
218 struct vlapic *vm_lapic(struct vm *vm, int cpu);
219 struct vioapic *vm_ioapic(struct vm *vm);
220 struct vhpet *vm_hpet(struct vm *vm);
221 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
222 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
223 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
224 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
225 int vm_apicid2vcpuid(struct vm *vm, int apicid);
226 int vm_activate_cpu(struct vm *vm, int vcpu);
227 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
228 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
229 void vm_exit_rendezvous(struct vm *vm, int vcpuid, uint64_t rip);
230 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
231 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
233 #ifdef _SYS__CPUSET_H_
235 * Rendezvous all vcpus specified in 'dest' and execute 'func(arg)'.
236 * The rendezvous 'func(arg)' is not allowed to do anything that will
237 * cause the thread to be put to sleep.
239 * If the rendezvous is being initiated from a vcpu context then the
240 * 'vcpuid' must refer to that vcpu, otherwise it should be set to -1.
242 * The caller cannot hold any locks when initiating the rendezvous.
244 * The implementation of this API may cause vcpus other than those specified
245 * by 'dest' to be stalled. The caller should not rely on any vcpus making
246 * forward progress when the rendezvous is in progress.
248 typedef void (*vm_rendezvous_func_t)(struct vm *vm, int vcpuid, void *arg);
249 void vm_smp_rendezvous(struct vm *vm, int vcpuid, cpuset_t dest,
250 vm_rendezvous_func_t func, void *arg);
251 cpuset_t vm_active_cpus(struct vm *vm);
252 cpuset_t vm_suspended_cpus(struct vm *vm);
253 #endif /* _SYS__CPUSET_H_ */
256 vcpu_rendezvous_pending(struct vm_eventinfo *info)
259 return (*((uintptr_t *)(info->rptr)) != 0);
263 vcpu_suspended(struct vm_eventinfo *info)
266 return (*info->sptr);
270 vcpu_reqidle(struct vm_eventinfo *info)
273 return (*info->iptr);
277 * Return 1 if device indicated by bus/slot/func is supposed to be a
278 * pci passthrough device.
280 * Return 0 otherwise.
282 int vmm_is_pptdev(int bus, int slot, int func);
284 void *vm_iommu_domain(struct vm *vm);
293 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
295 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
298 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
300 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
305 vcpu_should_yield(struct vm *vm, int vcpu)
308 if (curthread->td_flags & (TDF_ASTPENDING | TDF_NEEDRESCHED))
310 else if (curthread->td_owepreempt)
317 void *vcpu_stats(struct vm *vm, int vcpu);
318 void vcpu_notify_event(struct vm *vm, int vcpuid, bool lapic_intr);
319 struct vmspace *vm_get_vmspace(struct vm *vm);
320 struct vatpic *vm_atpic(struct vm *vm);
321 struct vatpit *vm_atpit(struct vm *vm);
322 struct vpmtmr *vm_pmtmr(struct vm *vm);
323 struct vrtc *vm_rtc(struct vm *vm);
326 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
327 * success and non-zero on failure.
329 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
330 * this function directly because they enforce the trap-like or fault-like
331 * behavior of an exception.
333 * This function should only be called in the context of the thread that is
334 * executing this vcpu.
336 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
337 uint32_t errcode, int restart_instruction);
340 * This function is called after a VM-exit that occurred during exception or
341 * interrupt delivery through the IDT. The format of 'intinfo' is described
342 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
344 * If a VM-exit handler completes the event delivery successfully then it
345 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
346 * if the task switch emulation is triggered via a task gate then it should
347 * call this function with 'intinfo=0' to indicate that the external event
348 * is not pending anymore.
350 * Return value is 0 on success and non-zero on failure.
352 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
355 * This function is called before every VM-entry to retrieve a pending
356 * event that should be injected into the guest. This function combines
357 * nested events into a double or triple fault.
359 * Returns 0 if there are no events that need to be injected into the guest
360 * and non-zero otherwise.
362 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
364 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
366 enum vm_reg_name vm_segment_name(int seg_encoding);
376 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
377 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
378 * a copyin or PROT_WRITE for a copyout.
380 * retval is_fault Interpretation
382 * 0 1 An exception was injected into the guest
383 * EFAULT N/A Unrecoverable error
385 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
386 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
387 * 'vm_copy_teardown()' after the copy is done.
389 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
390 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
391 int num_copyinfo, int *is_fault);
392 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
394 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
395 void *kaddr, size_t len);
396 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
397 struct vm_copyinfo *copyinfo, size_t len);
399 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
402 #define VM_MAXCPU 16 /* maximum virtual cpus */
405 * Identifiers for optional vmm capabilities
411 VM_CAP_UNRESTRICTED_GUEST,
412 VM_CAP_ENABLE_INVPCID,
416 enum vm_intr_trigger {
422 * The 'access' field has the format specified in Table 21-2 of the Intel
423 * Architecture Manual vol 3b.
425 * XXX The contents of the 'access' field are architecturally defined except
426 * bit 16 - Segment Unusable.
433 #define SEG_DESC_TYPE(access) ((access) & 0x001f)
434 #define SEG_DESC_DPL(access) (((access) >> 5) & 0x3)
435 #define SEG_DESC_PRESENT(access) (((access) & 0x0080) ? 1 : 0)
436 #define SEG_DESC_DEF32(access) (((access) & 0x4000) ? 1 : 0)
437 #define SEG_DESC_GRANULARITY(access) (((access) & 0x8000) ? 1 : 0)
438 #define SEG_DESC_UNUSABLE(access) (((access) & 0x10000) ? 1 : 0)
443 CPU_MODE_COMPATIBILITY, /* IA-32E mode (CS.L = 0) */
444 CPU_MODE_64BIT, /* IA-32E mode (CS.L = 1) */
447 enum vm_paging_mode {
454 struct vm_guest_paging {
457 enum vm_cpu_mode cpu_mode;
458 enum vm_paging_mode paging_mode;
462 * The data structures 'vie' and 'vie_op' are meant to be opaque to the
463 * consumers of instruction decoding. The only reason why their contents
464 * need to be exposed is because they are part of the 'vm_exit' structure.
467 uint8_t op_byte; /* actual opcode byte */
468 uint8_t op_type; /* type of operation (e.g. MOV) */
472 #define VIE_INST_SIZE 15
474 uint8_t inst[VIE_INST_SIZE]; /* instruction bytes */
475 uint8_t num_valid; /* size of the instruction */
476 uint8_t num_processed;
478 uint8_t addrsize:4, opsize:4; /* address and operand sizes */
479 uint8_t rex_w:1, /* REX prefix */
484 repz_present:1, /* REP/REPE/REPZ prefix */
485 repnz_present:1, /* REPNE/REPNZ prefix */
486 opsize_override:1, /* Operand size override */
487 addrsize_override:1, /* Address size override */
488 segment_override:1; /* Segment override */
490 uint8_t mod:2, /* ModRM byte */
494 uint8_t ss:2, /* SIB byte */
502 int base_register; /* VM_REG_GUEST_xyz */
503 int index_register; /* VM_REG_GUEST_xyz */
504 int segment_register; /* VM_REG_GUEST_xyz */
506 int64_t displacement; /* optional addr displacement */
507 int64_t immediate; /* optional immediate operand */
509 uint8_t decoded; /* set to 1 if successfully decoded */
511 struct vie_op op; /* opcode description */
524 VM_EXITCODE_INST_EMUL,
525 VM_EXITCODE_SPINUP_AP,
526 VM_EXITCODE_DEPRECATED1, /* used to be SPINDOWN_CPU */
527 VM_EXITCODE_RENDEZVOUS,
528 VM_EXITCODE_IOAPIC_EOI,
529 VM_EXITCODE_SUSPENDED,
530 VM_EXITCODE_INOUT_STR,
531 VM_EXITCODE_TASK_SWITCH,
540 uint16_t bytes:3; /* 1 or 2 or 4 */
545 uint32_t eax; /* valid for out */
548 struct vm_inout_str {
549 struct vm_inout inout; /* must be the first element */
550 struct vm_guest_paging paging;
554 uint64_t count; /* rep=1 (%rcx), rep=0 (1) */
556 enum vm_reg_name seg_name;
557 struct seg_desc seg_desc;
560 enum task_switch_reason {
564 TSR_IDT_GATE, /* task gate in IDT */
567 struct vm_task_switch {
568 uint16_t tsssel; /* new TSS selector */
569 int ext; /* task switch due to external event */
571 int errcode_valid; /* push 'errcode' on the new stack */
572 enum task_switch_reason reason;
573 struct vm_guest_paging paging;
577 enum vm_exitcode exitcode;
578 int inst_length; /* 0 means unknown */
581 struct vm_inout inout;
582 struct vm_inout_str inout_str;
592 struct vm_guest_paging paging;
596 * VMX specific payload. Used when there is no "better"
597 * exitcode to represent the VM-exit.
600 int status; /* vmx inst status */
602 * 'exit_reason' and 'exit_qualification' are valid
603 * only if 'status' is zero.
605 uint32_t exit_reason;
606 uint64_t exit_qualification;
608 * 'inst_error' and 'inst_type' are valid
609 * only if 'status' is non-zero.
615 * SVM specific payload.
623 uint32_t code; /* ecx value */
637 enum vm_suspend_how how;
639 struct vm_task_switch task_switch;
643 /* APIs to inject faults into the guest */
644 void vm_inject_fault(void *vm, int vcpuid, int vector, int errcode_valid,
648 vm_inject_ud(void *vm, int vcpuid)
650 vm_inject_fault(vm, vcpuid, IDT_UD, 0, 0);
654 vm_inject_gp(void *vm, int vcpuid)
656 vm_inject_fault(vm, vcpuid, IDT_GP, 1, 0);
660 vm_inject_ac(void *vm, int vcpuid, int errcode)
662 vm_inject_fault(vm, vcpuid, IDT_AC, 1, errcode);
666 vm_inject_ss(void *vm, int vcpuid, int errcode)
668 vm_inject_fault(vm, vcpuid, IDT_SS, 1, errcode);
671 void vm_inject_pf(void *vm, int vcpuid, int error_code, uint64_t cr2);
673 int vm_restart_instruction(void *vm, int vcpuid);