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 2020 Oxide Computer Company
43 */
44
45 #ifndef _VMM_KERNEL_H_
46 #define _VMM_KERNEL_H_
47
48 #include <sys/sdt.h>
49 #include <x86/segments.h>
50
51 SDT_PROVIDER_DECLARE(vmm);
52
53 struct vm;
54 struct vm_exception;
55 struct seg_desc;
56 struct vm_exit;
57 struct vie;
58 struct vm_run;
59 struct vhpet;
60 struct vioapic;
61 struct vlapic;
62 struct vmspace;
63 struct vm_object;
64 struct vm_guest_paging;
65 struct pmap;
66
67 struct vm_eventinfo {
68 uint_t *rptr; /* runblock cookie */
69 int *sptr; /* suspend cookie */
70 int *iptr; /* reqidle cookie */
71 };
72
73 typedef int (*vmm_init_func_t)(int ipinum);
74 typedef int (*vmm_cleanup_func_t)(void);
75 typedef void (*vmm_resume_func_t)(void);
76 typedef void * (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
77 typedef int (*vmi_run_func_t)(void *vmi, int vcpu, uint64_t rip,
78 struct pmap *pmap, struct vm_eventinfo *info);
79 typedef void (*vmi_cleanup_func_t)(void *vmi);
80 typedef int (*vmi_get_register_t)(void *vmi, int vcpu, int num,
81 uint64_t *retval);
82 typedef int (*vmi_set_register_t)(void *vmi, int vcpu, int num,
83 uint64_t val);
84 typedef int (*vmi_get_desc_t)(void *vmi, int vcpu, int num,
85 struct seg_desc *desc);
86 typedef int (*vmi_set_desc_t)(void *vmi, int vcpu, int num,
87 struct seg_desc *desc);
88 typedef int (*vmi_get_cap_t)(void *vmi, int vcpu, int num, int *retval);
89 typedef int (*vmi_set_cap_t)(void *vmi, int vcpu, int num, int val);
90 typedef struct vmspace *(*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
91 typedef void (*vmi_vmspace_free)(struct vmspace *vmspace);
92 typedef struct vlapic *(*vmi_vlapic_init)(void *vmi, int vcpu);
93 typedef void (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
94 #ifndef __FreeBSD__
95 typedef void (*vmi_savectx)(void *vmi, int vcpu);
96 typedef void (*vmi_restorectx)(void *vmi, int vcpu);
97 #endif
98
99 struct vmm_ops {
100 vmm_init_func_t init; /* module wide initialization */
101 vmm_cleanup_func_t cleanup;
102 vmm_resume_func_t resume;
103
104 vmi_init_func_t vminit; /* vm-specific initialization */
105 vmi_run_func_t vmrun;
106 vmi_cleanup_func_t vmcleanup;
107 vmi_get_register_t vmgetreg;
108 vmi_set_register_t vmsetreg;
109 vmi_get_desc_t vmgetdesc;
110 vmi_set_desc_t vmsetdesc;
111 vmi_get_cap_t vmgetcap;
112 vmi_set_cap_t vmsetcap;
113 vmi_vmspace_alloc vmspace_alloc;
114 vmi_vmspace_free vmspace_free;
115 vmi_vlapic_init vlapic_init;
116 vmi_vlapic_cleanup vlapic_cleanup;
117
118 #ifndef __FreeBSD__
119 vmi_savectx vmsavectx;
120 vmi_restorectx vmrestorectx;
121 #endif
122 };
123
124 extern struct vmm_ops vmm_ops_intel;
125 extern struct vmm_ops vmm_ops_amd;
126
127 int vm_create(const char *name, struct vm **retvm);
128 void vm_destroy(struct vm *vm);
129 int vm_reinit(struct vm *vm);
130 const char *vm_name(struct vm *vm);
131 uint16_t vm_get_maxcpus(struct vm *vm);
132 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
133 uint16_t *threads, uint16_t *maxcpus);
134 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
135 uint16_t threads, uint16_t maxcpus);
136
137 /*
138 * APIs that modify the guest memory map require all vcpus to be frozen.
139 */
140 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
141 size_t len, int prot, int flags);
142 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
143 void vm_free_memseg(struct vm *vm, int ident);
144 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
145 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
146 #ifdef __FreeBSD__
147 int vm_assign_pptdev(struct vm *vm, int bus, int slot, int func);
148 int vm_unassign_pptdev(struct vm *vm, int bus, int slot, int func);
149 #else
150 int vm_assign_pptdev(struct vm *vm, int pptfd);
151 int vm_unassign_pptdev(struct vm *vm, int pptfd);
152 #endif /* __FreeBSD__ */
153
154 /*
155 * APIs that inspect the guest memory map require only a *single* vcpu to
156 * be frozen. This acts like a read lock on the guest memory map since any
157 * modification requires *all* vcpus to be frozen.
158 */
159 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
160 vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
161 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
162 struct vm_object **objptr);
163 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
164 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
165 int prot, void **cookie);
166 void vm_gpa_release(void *cookie);
167 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
168
169 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
170 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
171 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
172 struct seg_desc *ret_desc);
173 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
174 struct seg_desc *desc);
175 int vm_run(struct vm *vm, int vcpuid, const struct vm_entry *);
176 int vm_suspend(struct vm *vm, enum vm_suspend_how how);
177 int vm_inject_nmi(struct vm *vm, int vcpu);
178 int vm_nmi_pending(struct vm *vm, int vcpuid);
179 void vm_nmi_clear(struct vm *vm, int vcpuid);
180 int vm_inject_extint(struct vm *vm, int vcpu);
181 int vm_extint_pending(struct vm *vm, int vcpuid);
182 void vm_extint_clear(struct vm *vm, int vcpuid);
183 struct vlapic *vm_lapic(struct vm *vm, int cpu);
184 struct vioapic *vm_ioapic(struct vm *vm);
185 struct vhpet *vm_hpet(struct vm *vm);
186 int vm_get_capability(struct vm *vm, int vcpu, int type, int *val);
187 int vm_set_capability(struct vm *vm, int vcpu, int type, int val);
188 int vm_get_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state *state);
189 int vm_set_x2apic_state(struct vm *vm, int vcpu, enum x2apic_state state);
190 int vm_apicid2vcpuid(struct vm *vm, int apicid);
191 int vm_activate_cpu(struct vm *vm, int vcpu);
192 int vm_suspend_cpu(struct vm *vm, int vcpu);
193 int vm_resume_cpu(struct vm *vm, int vcpu);
194 struct vm_exit *vm_exitinfo(struct vm *vm, int vcpuid);
195 struct vie *vm_vie_ctx(struct vm *vm, int vcpuid);
196 void vm_exit_suspended(struct vm *vm, int vcpuid, uint64_t rip);
197 void vm_exit_debug(struct vm *vm, int vcpuid, uint64_t rip);
198 void vm_exit_runblock(struct vm *vm, int vcpuid, uint64_t rip);
199 void vm_exit_astpending(struct vm *vm, int vcpuid, uint64_t rip);
200 void vm_exit_reqidle(struct vm *vm, int vcpuid, uint64_t rip);
201 int vm_service_mmio_read(struct vm *vm, int cpuid, uint64_t gpa, uint64_t *rval,
202 int rsize);
203 int vm_service_mmio_write(struct vm *vm, int cpuid, uint64_t gpa, uint64_t wval,
204 int wsize);
205 void vm_req_spinup_ap(struct vm *vm, int req_vcpuid, uint64_t req_rip);
206
207 #ifdef _SYS__CPUSET_H_
208 cpuset_t vm_active_cpus(struct vm *vm);
209 cpuset_t vm_debug_cpus(struct vm *vm);
210 cpuset_t vm_suspended_cpus(struct vm *vm);
211 #endif /* _SYS__CPUSET_H_ */
212
213 static __inline int
214 vcpu_runblocked(struct vm_eventinfo *info)
215 {
216
217 return (*info->rptr != 0);
218 }
219
220 static __inline int
221 vcpu_suspended(struct vm_eventinfo *info)
222 {
223
224 return (*info->sptr);
225 }
226
227 static __inline int
228 vcpu_reqidle(struct vm_eventinfo *info)
229 {
230
231 return (*info->iptr);
232 }
233
234 int vcpu_debugged(struct vm *vm, int vcpuid);
235
236 /*
237 * Return true if device indicated by bus/slot/func is supposed to be a
238 * pci passthrough device.
239 *
240 * Return false otherwise.
241 */
242 bool vmm_is_pptdev(int bus, int slot, int func);
243
244 void *vm_iommu_domain(struct vm *vm);
245
246 enum vcpu_state {
247 VCPU_IDLE,
248 VCPU_FROZEN,
249 VCPU_RUNNING,
250 VCPU_SLEEPING,
251 };
252
253 int vcpu_set_state(struct vm *vm, int vcpu, enum vcpu_state state,
254 bool from_idle);
255 enum vcpu_state vcpu_get_state(struct vm *vm, int vcpu, int *hostcpu);
256 void vcpu_block_run(struct vm *, int);
257 void vcpu_unblock_run(struct vm *, int);
258
259 #ifndef __FreeBSD__
260 uint64_t vcpu_tsc_offset(struct vm *vm, int vcpuid);
261 #endif
262
263 static __inline int
264 vcpu_is_running(struct vm *vm, int vcpu, int *hostcpu)
265 {
266 return (vcpu_get_state(vm, vcpu, hostcpu) == VCPU_RUNNING);
267 }
268
269 #ifdef _SYS_THREAD_H
270 static __inline int
271 vcpu_should_yield(struct vm *vm, int vcpu)
272 {
273
274 if (curthread->t_astflag)
275 return (1);
276 else if (CPU->cpu_runrun)
277 return (1);
278 else
279 return (0);
280 }
281 #endif /* _SYS_THREAD_H */
282
283 typedef enum vcpu_notify {
284 VCPU_NOTIFY_NONE,
285 VCPU_NOTIFY_APIC, /* Posted intr notification (if possible) */
286 VCPU_NOTIFY_EXIT, /* IPI to cause VM exit */
287 } vcpu_notify_t;
288
289 void *vcpu_stats(struct vm *vm, int vcpu);
290 void vcpu_notify_event(struct vm *vm, int vcpuid);
291 void vcpu_notify_event_type(struct vm *vm, int vcpuid, vcpu_notify_t);
292 struct vmspace *vm_get_vmspace(struct vm *vm);
293 struct vatpic *vm_atpic(struct vm *vm);
294 struct vatpit *vm_atpit(struct vm *vm);
295 struct vpmtmr *vm_pmtmr(struct vm *vm);
296 struct vrtc *vm_rtc(struct vm *vm);
297
298 /*
299 * Inject exception 'vector' into the guest vcpu. This function returns 0 on
300 * success and non-zero on failure.
301 *
302 * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
303 * this function directly because they enforce the trap-like or fault-like
304 * behavior of an exception.
305 *
306 * This function should only be called in the context of the thread that is
307 * executing this vcpu.
308 */
309 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
310 uint32_t errcode, int restart_instruction);
311
312 /*
313 * This function is called after a VM-exit that occurred during exception or
314 * interrupt delivery through the IDT. The format of 'intinfo' is described
315 * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
316 *
317 * If a VM-exit handler completes the event delivery successfully then it
318 * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
319 * if the task switch emulation is triggered via a task gate then it should
320 * call this function with 'intinfo=0' to indicate that the external event
321 * is not pending anymore.
322 *
323 * Return value is 0 on success and non-zero on failure.
324 */
325 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
326
327 /*
328 * This function is called before every VM-entry to retrieve a pending
329 * event that should be injected into the guest. This function combines
330 * nested events into a double or triple fault.
331 *
332 * Returns 0 if there are no events that need to be injected into the guest
333 * and non-zero otherwise.
334 */
335 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
336
337 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
338
339 enum vm_reg_name vm_segment_name(int seg_encoding);
340
341 struct vm_copyinfo {
342 uint64_t gpa;
343 size_t len;
344 void *hva;
345 void *cookie;
346 };
347
348 /*
349 * Set up 'copyinfo[]' to copy to/from guest linear address space starting
350 * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
351 * a copyin or PROT_WRITE for a copyout.
352 *
353 * retval is_fault Interpretation
354 * 0 0 Success
355 * 0 1 An exception was injected into the guest
356 * EFAULT N/A Unrecoverable error
357 *
358 * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
359 * the return value is 0. The 'copyinfo[]' resources should be freed by calling
360 * 'vm_copy_teardown()' after the copy is done.
361 */
362 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
363 uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
364 int num_copyinfo, int *is_fault);
365 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
366 int num_copyinfo);
367 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
368 void *kaddr, size_t len);
369 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
370 struct vm_copyinfo *copyinfo, size_t len);
371
372 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
373
374 /* APIs to inject faults into the guest */
375 void vm_inject_fault(struct vm *vm, int vcpuid, int vector, int errcode_valid,
376 int errcode);
377
378 void vm_inject_ud(struct vm *vm, int vcpuid);
379 void vm_inject_gp(struct vm *vm, int vcpuid);
380 void vm_inject_ac(struct vm *vm, int vcpuid, int errcode);
381 void vm_inject_ss(struct vm *vm, int vcpuid, int errcode);
382 void vm_inject_pf(struct vm *vm, int vcpuid, int errcode, uint64_t cr2);
383
384 /*
385 * Both SVM and VMX have complex logic for injecting events such as exceptions
386 * or interrupts into the guest. Within those two backends, the progress of
387 * event injection is tracked by event_inject_state, hopefully making it easier
388 * to reason about.
389 */
390 enum event_inject_state {
391 EIS_CAN_INJECT = 0, /* exception/interrupt can be injected */
392 EIS_EV_EXISTING = 1, /* blocked by existing event */
393 EIS_EV_INJECTED = 2, /* blocked by injected event */
394 EIS_GI_BLOCK = 3, /* blocked by guest interruptability */
395
396 /*
397 * Flag to request an immediate exit from VM context after event
398 * injection in order to perform more processing
399 */
400 EIS_REQ_EXIT = (1 << 15),
401 };
402
403 #ifndef __FreeBSD__
404
405 void vmm_sol_glue_init(void);
406 void vmm_sol_glue_cleanup(void);
407
408 int vmm_mod_load(void);
409 int vmm_mod_unload(void);
410
411 void vmm_call_trap(uint64_t);
412
413 /*
414 * Because of tangled headers, this is not exposed directly via the vmm_drv
415 * interface, but rather mirrored as vmm_drv_iop_cb_t in vmm_drv.h.
416 */
417 typedef int (*ioport_handler_t)(void *, bool, uint16_t, uint8_t, uint32_t *);
418
419 int vm_ioport_access(struct vm *vm, int vcpuid, bool in, uint16_t port,
420 uint8_t bytes, uint32_t *val);
421
422 int vm_ioport_attach(struct vm *vm, uint16_t port, ioport_handler_t func,
423 void *arg, void **cookie);
424 int vm_ioport_detach(struct vm *vm, void **cookie, ioport_handler_t *old_func,
425 void **old_arg);
426
427 int vm_ioport_hook(struct vm *, uint16_t, ioport_handler_t, void *, void **);
428 void vm_ioport_unhook(struct vm *, void **);
429
430 #endif /* __FreeBSD */
431
432 #endif /* _VMM_KERNEL_H_ */