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 2021 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 
  52 SDT_PROVIDER_DECLARE(vmm);
  53 
  54 struct vm;
  55 struct vm_exception;
  56 struct seg_desc;
  57 struct vm_exit;
  58 struct vie;
  59 struct vm_run;
  60 struct vhpet;
  61 struct vioapic;
  62 struct vlapic;
  63 struct vmspace;
  64 struct vm_object;
  65 struct vm_guest_paging;
  66 struct pmap;
  67 
  68 typedef int     (*vmm_init_func_t)(int ipinum);
  69 typedef int     (*vmm_cleanup_func_t)(void);
  70 typedef void    (*vmm_resume_func_t)(void);
  71 typedef void *  (*vmi_init_func_t)(struct vm *vm, struct pmap *pmap);
  72 typedef int     (*vmi_run_func_t)(void *vmi, int vcpu, uint64_t rip,
  73     struct pmap *pmap);
  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 vmspace *(*vmi_vmspace_alloc)(vm_offset_t min, vm_offset_t max);
  86 typedef void    (*vmi_vmspace_free)(struct vmspace *vmspace);
  87 typedef struct vlapic *(*vmi_vlapic_init)(void *vmi, int vcpu);
  88 typedef void    (*vmi_vlapic_cleanup)(void *vmi, struct vlapic *vlapic);
  89 typedef void    (*vmi_savectx)(void *vmi, int vcpu);
  90 typedef void    (*vmi_restorectx)(void *vmi, int vcpu);
  91 
  92 struct vmm_ops {
  93         vmm_init_func_t         init;           /* module wide initialization */
  94         vmm_cleanup_func_t      cleanup;
  95         vmm_resume_func_t       resume;
  96 
  97         vmi_init_func_t         vminit;         /* vm-specific initialization */
  98         vmi_run_func_t          vmrun;
  99         vmi_cleanup_func_t      vmcleanup;
 100         vmi_get_register_t      vmgetreg;
 101         vmi_set_register_t      vmsetreg;
 102         vmi_get_desc_t          vmgetdesc;
 103         vmi_set_desc_t          vmsetdesc;
 104         vmi_get_cap_t           vmgetcap;
 105         vmi_set_cap_t           vmsetcap;
 106         vmi_vmspace_alloc       vmspace_alloc;
 107         vmi_vmspace_free        vmspace_free;
 108         vmi_vlapic_init         vlapic_init;
 109         vmi_vlapic_cleanup      vlapic_cleanup;
 110 
 111         vmi_savectx             vmsavectx;
 112         vmi_restorectx          vmrestorectx;
 113 };
 114 
 115 extern struct vmm_ops vmm_ops_intel;
 116 extern struct vmm_ops vmm_ops_amd;
 117 
 118 int vm_create(const char *name, struct vm **retvm);
 119 void vm_destroy(struct vm *vm);
 120 int vm_reinit(struct vm *vm);
 121 const char *vm_name(struct vm *vm);
 122 uint16_t vm_get_maxcpus(struct vm *vm);
 123 void vm_get_topology(struct vm *vm, uint16_t *sockets, uint16_t *cores,
 124     uint16_t *threads, uint16_t *maxcpus);
 125 int vm_set_topology(struct vm *vm, uint16_t sockets, uint16_t cores,
 126     uint16_t threads, uint16_t maxcpus);
 127 
 128 /*
 129  * APIs that modify the guest memory map require all vcpus to be frozen.
 130  */
 131 int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off,
 132     size_t len, int prot, int flags);
 133 int vm_munmap_memseg(struct vm *vm, vm_paddr_t gpa, size_t len);
 134 int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem);
 135 void vm_free_memseg(struct vm *vm, int ident);
 136 int vm_map_mmio(struct vm *vm, vm_paddr_t gpa, size_t len, vm_paddr_t hpa);
 137 int vm_unmap_mmio(struct vm *vm, vm_paddr_t gpa, size_t len);
 138 int vm_assign_pptdev(struct vm *vm, int pptfd);
 139 int vm_unassign_pptdev(struct vm *vm, int pptfd);
 140 
 141 /*
 142  * APIs that inspect the guest memory map require only a *single* vcpu to
 143  * be frozen. This acts like a read lock on the guest memory map since any
 144  * modification requires *all* vcpus to be frozen.
 145  */
 146 int vm_mmap_getnext(struct vm *vm, vm_paddr_t *gpa, int *segid,
 147     vm_ooffset_t *segoff, size_t *len, int *prot, int *flags);
 148 int vm_get_memseg(struct vm *vm, int ident, size_t *len, bool *sysmem,
 149     struct vm_object **objptr);
 150 vm_paddr_t vmm_sysmem_maxaddr(struct vm *vm);
 151 void *vm_gpa_hold(struct vm *, int vcpuid, vm_paddr_t gpa, size_t len,
 152     int prot, void **cookie);
 153 void vm_gpa_release(void *cookie);
 154 bool vm_mem_allocated(struct vm *vm, int vcpuid, vm_paddr_t gpa);
 155 
 156 int vm_get_register(struct vm *vm, int vcpu, int reg, uint64_t *retval);
 157 int vm_set_register(struct vm *vm, int vcpu, int reg, uint64_t val);
 158 int vm_get_seg_desc(struct vm *vm, int vcpu, int reg,
 159     struct seg_desc *ret_desc);
 160 int vm_set_seg_desc(struct vm *vm, int vcpu, int reg,
 161     const struct seg_desc *desc);
 162 int vm_get_run_state(struct vm *vm, int vcpuid, uint32_t *state,
 163     uint8_t *sipi_vec);
 164 int vm_set_run_state(struct vm *vm, int vcpuid, uint32_t state,
 165     uint8_t sipi_vec);
 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 vatpic *vm_atpic(struct vm *vm);
 265 struct vatpit *vm_atpit(struct vm *vm);
 266 struct vpmtmr *vm_pmtmr(struct vm *vm);
 267 struct vrtc *vm_rtc(struct vm *vm);
 268 
 269 /*
 270  * Inject exception 'vector' into the guest vcpu. This function returns 0 on
 271  * success and non-zero on failure.
 272  *
 273  * Wrapper functions like 'vm_inject_gp()' should be preferred to calling
 274  * this function directly because they enforce the trap-like or fault-like
 275  * behavior of an exception.
 276  *
 277  * This function should only be called in the context of the thread that is
 278  * executing this vcpu.
 279  */
 280 int vm_inject_exception(struct vm *vm, int vcpuid, int vector, int err_valid,
 281     uint32_t errcode, int restart_instruction);
 282 
 283 /*
 284  * This function is called after a VM-exit that occurred during exception or
 285  * interrupt delivery through the IDT. The format of 'intinfo' is described
 286  * in Figure 15-1, "EXITINTINFO for All Intercepts", APM, Vol 2.
 287  *
 288  * If a VM-exit handler completes the event delivery successfully then it
 289  * should call vm_exit_intinfo() to extinguish the pending event. For e.g.,
 290  * if the task switch emulation is triggered via a task gate then it should
 291  * call this function with 'intinfo=0' to indicate that the external event
 292  * is not pending anymore.
 293  *
 294  * Return value is 0 on success and non-zero on failure.
 295  */
 296 int vm_exit_intinfo(struct vm *vm, int vcpuid, uint64_t intinfo);
 297 
 298 /*
 299  * This function is called before every VM-entry to retrieve a pending
 300  * event that should be injected into the guest. This function combines
 301  * nested events into a double or triple fault.
 302  *
 303  * Returns 0 if there are no events that need to be injected into the guest
 304  * and non-zero otherwise.
 305  */
 306 int vm_entry_intinfo(struct vm *vm, int vcpuid, uint64_t *info);
 307 
 308 int vm_get_intinfo(struct vm *vm, int vcpuid, uint64_t *info1, uint64_t *info2);
 309 
 310 enum vm_reg_name vm_segment_name(int seg_encoding);
 311 
 312 struct vm_copyinfo {
 313         uint64_t        gpa;
 314         size_t          len;
 315         void            *hva;
 316         void            *cookie;
 317 };
 318 
 319 /*
 320  * Set up 'copyinfo[]' to copy to/from guest linear address space starting
 321  * at 'gla' and 'len' bytes long. The 'prot' should be set to PROT_READ for
 322  * a copyin or PROT_WRITE for a copyout.
 323  *
 324  * retval       is_fault        Interpretation
 325  *   0             0            Success
 326  *   0             1            An exception was injected into the guest
 327  * EFAULT         N/A           Unrecoverable error
 328  *
 329  * The 'copyinfo[]' can be passed to 'vm_copyin()' or 'vm_copyout()' only if
 330  * the return value is 0. The 'copyinfo[]' resources should be freed by calling
 331  * 'vm_copy_teardown()' after the copy is done.
 332  */
 333 int vm_copy_setup(struct vm *vm, int vcpuid, struct vm_guest_paging *paging,
 334     uint64_t gla, size_t len, int prot, struct vm_copyinfo *copyinfo,
 335     int num_copyinfo, int *is_fault);
 336 void vm_copy_teardown(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
 337     int num_copyinfo);
 338 void vm_copyin(struct vm *vm, int vcpuid, struct vm_copyinfo *copyinfo,
 339     void *kaddr, size_t len);
 340 void vm_copyout(struct vm *vm, int vcpuid, const void *kaddr,
 341     struct vm_copyinfo *copyinfo, size_t len);
 342 
 343 int vcpu_trace_exceptions(struct vm *vm, int vcpuid);
 344 
 345 /* APIs to inject faults into the guest */
 346 void vm_inject_fault(struct vm *vm, int vcpuid, int vector, int errcode_valid,
 347     int errcode);
 348 
 349 void vm_inject_ud(struct vm *vm, int vcpuid);
 350 void vm_inject_gp(struct vm *vm, int vcpuid);
 351 void vm_inject_ac(struct vm *vm, int vcpuid, int errcode);
 352 void vm_inject_ss(struct vm *vm, int vcpuid, int errcode);
 353 void vm_inject_pf(struct vm *vm, int vcpuid, int errcode, uint64_t cr2);
 354 
 355 /*
 356  * Both SVM and VMX have complex logic for injecting events such as exceptions
 357  * or interrupts into the guest.  Within those two backends, the progress of
 358  * event injection is tracked by event_inject_state, hopefully making it easier
 359  * to reason about.
 360  */
 361 enum event_inject_state {
 362         EIS_CAN_INJECT  = 0, /* exception/interrupt can be injected */
 363         EIS_EV_EXISTING = 1, /* blocked by existing event */
 364         EIS_EV_INJECTED = 2, /* blocked by injected event */
 365         EIS_GI_BLOCK    = 3, /* blocked by guest interruptability */
 366 
 367         /*
 368          * Flag to request an immediate exit from VM context after event
 369          * injection in order to perform more processing
 370          */
 371         EIS_REQ_EXIT    = (1 << 15),
 372 };
 373 
 374 void vmm_sol_glue_init(void);
 375 void vmm_sol_glue_cleanup(void);
 376 
 377 int vmm_mod_load(void);
 378 int vmm_mod_unload(void);
 379 
 380 void vmm_call_trap(uint64_t);
 381 
 382 /*
 383  * Because of tangled headers, this is not exposed directly via the vmm_drv
 384  * interface, but rather mirrored as vmm_drv_iop_cb_t in vmm_drv.h.
 385  */
 386 typedef int (*ioport_handler_t)(void *, bool, uint16_t, uint8_t, uint32_t *);
 387 
 388 int vm_ioport_access(struct vm *vm, int vcpuid, bool in, uint16_t port,
 389     uint8_t bytes, uint32_t *val);
 390 
 391 int vm_ioport_attach(struct vm *vm, uint16_t port, ioport_handler_t func,
 392     void *arg, void **cookie);
 393 int vm_ioport_detach(struct vm *vm, void **cookie, ioport_handler_t *old_func,
 394     void **old_arg);
 395 
 396 int vm_ioport_hook(struct vm *, uint16_t, ioport_handler_t, void *, void **);
 397 void vm_ioport_unhook(struct vm *, void **);
 398 
 399 enum vcpu_ustate {
 400         VU_INIT = 0,    /* initialized but has not yet attempted to run */
 401         VU_RUN,         /* running in guest context */
 402         VU_IDLE,        /* idle (HLTed, wait-for-SIPI, etc) */
 403         VU_EMU_KERN,    /* emulation performed in-kernel */
 404         VU_EMU_USER,    /* emulation performed in userspace */
 405         VU_SCHED,       /* off-cpu for interrupt, preempt, lock contention */
 406         VU_MAX
 407 };
 408 
 409 void vcpu_ustate_change(struct vm *, int, enum vcpu_ustate);
 410 
 411 typedef struct vmm_kstats {
 412         kstat_named_t   vk_name;
 413 } vmm_kstats_t;
 414 
 415 typedef struct vmm_vcpu_kstats {
 416         kstat_named_t   vvk_vcpu;
 417         kstat_named_t   vvk_time_init;
 418         kstat_named_t   vvk_time_run;
 419         kstat_named_t   vvk_time_idle;
 420         kstat_named_t   vvk_time_emu_kern;
 421         kstat_named_t   vvk_time_emu_user;
 422         kstat_named_t   vvk_time_sched;
 423 } vmm_vcpu_kstats_t;
 424 
 425 #define VMM_KSTAT_CLASS "misc"
 426 
 427 int vmm_kstat_update_vcpu(struct kstat *, int);
 428 
 429 #endif /* _VMM_KERNEL_H_ */