| 1 | /* |
| 2 | * Copyright (c) 2006-2012 Apple Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | |
| 29 | #include <pexpert/pexpert.h> |
| 30 | #include <i386/cpuid.h> |
| 31 | #include <i386/cpu_data.h> |
| 32 | #include <i386/mp.h> |
| 33 | #include <i386/proc_reg.h> |
| 34 | #include <i386/vmx.h> |
| 35 | #include <i386/vmx/vmx_asm.h> |
| 36 | #include <i386/vmx/vmx_shims.h> |
| 37 | #include <i386/vmx/vmx_cpu.h> |
| 38 | #include <mach/mach_host.h> /* for host_info() */ |
| 39 | |
| 40 | #define VMX_KPRINTF(x...) /* kprintf("vmx: " x) */ |
| 41 | |
| 42 | int vmx_use_count = 0; |
| 43 | boolean_t vmx_exclusive = FALSE; |
| 44 | |
| 45 | lck_grp_t *vmx_lck_grp = NULL; |
| 46 | lck_mtx_t *vmx_lck_mtx = NULL; |
| 47 | |
| 48 | /* ----------------------------------------------------------------------------- |
| 49 | vmx_is_available() |
| 50 | Is the VMX facility available on this CPU? |
| 51 | -------------------------------------------------------------------------- */ |
| 52 | static inline boolean_t |
| 53 | vmx_is_available(void) |
| 54 | { |
| 55 | return (0 != (cpuid_features() & CPUID_FEATURE_VMX)); |
| 56 | } |
| 57 | |
| 58 | /* ----------------------------------------------------------------------------- |
| 59 | vmxon_is_enabled() |
| 60 | Is the VMXON instruction enabled on this CPU? |
| 61 | -------------------------------------------------------------------------- */ |
| 62 | static inline boolean_t |
| 63 | vmxon_is_enabled(void) |
| 64 | { |
| 65 | return (vmx_is_available() && |
| 66 | (rdmsr64(MSR_IA32_FEATURE_CONTROL) & MSR_IA32_FEATCTL_VMXON)); |
| 67 | } |
| 68 | |
| 69 | #if MACH_ASSERT |
| 70 | /* ----------------------------------------------------------------------------- |
| 71 | vmx_is_cr0_valid() |
| 72 | Is CR0 valid for executing VMXON on this CPU? |
| 73 | -------------------------------------------------------------------------- */ |
| 74 | static inline boolean_t |
| 75 | vmx_is_cr0_valid(vmx_specs_t *specs) |
| 76 | { |
| 77 | uintptr_t cr0 = get_cr0(); |
| 78 | return (0 == ((~cr0 & specs->cr0_fixed_0)|(cr0 & ~specs->cr0_fixed_1))); |
| 79 | } |
| 80 | |
| 81 | /* ----------------------------------------------------------------------------- |
| 82 | vmx_is_cr4_valid() |
| 83 | Is CR4 valid for executing VMXON on this CPU? |
| 84 | -------------------------------------------------------------------------- */ |
| 85 | static inline boolean_t |
| 86 | vmx_is_cr4_valid(vmx_specs_t *specs) |
| 87 | { |
| 88 | uintptr_t cr4 = get_cr4(); |
| 89 | return (0 == ((~cr4 & specs->cr4_fixed_0)|(cr4 & ~specs->cr4_fixed_1))); |
| 90 | } |
| 91 | |
| 92 | #endif |
| 93 | |
| 94 | static void |
| 95 | vmx_enable(void) |
| 96 | { |
| 97 | uint64_t msr_image; |
| 98 | |
| 99 | if (!vmx_is_available()) |
| 100 | return; |
| 101 | |
| 102 | /* |
| 103 | * We don't count on EFI initializing MSR_IA32_FEATURE_CONTROL |
| 104 | * and turning VMXON on and locking the bit, so we do that now. |
| 105 | */ |
| 106 | msr_image = rdmsr64(MSR_IA32_FEATURE_CONTROL); |
| 107 | if (0 == ((msr_image & MSR_IA32_FEATCTL_LOCK))) |
| 108 | wrmsr64(MSR_IA32_FEATURE_CONTROL, |
| 109 | (msr_image | |
| 110 | MSR_IA32_FEATCTL_VMXON | |
| 111 | MSR_IA32_FEATCTL_LOCK)); |
| 112 | |
| 113 | set_cr4(get_cr4() | CR4_VMXE); |
| 114 | } |
| 115 | |
| 116 | void |
| 117 | vmx_init() |
| 118 | { |
| 119 | vmx_lck_grp = lck_grp_alloc_init("vmx" , LCK_GRP_ATTR_NULL); |
| 120 | assert(vmx_lck_grp); |
| 121 | |
| 122 | vmx_lck_mtx = lck_mtx_alloc_init(vmx_lck_grp, LCK_ATTR_NULL); |
| 123 | assert(vmx_lck_mtx); |
| 124 | } |
| 125 | |
| 126 | /* ----------------------------------------------------------------------------- |
| 127 | vmx_get_specs() |
| 128 | Obtain VMX facility specifications for this CPU and |
| 129 | enter them into the vmx_specs_t structure. If VMX is not available or |
| 130 | disabled on this CPU, set vmx_present to false and return leaving |
| 131 | the remainder of the vmx_specs_t uninitialized. |
| 132 | -------------------------------------------------------------------------- */ |
| 133 | void |
| 134 | vmx_cpu_init() |
| 135 | { |
| 136 | vmx_specs_t *specs = ¤t_cpu_datap()->cpu_vmx.specs; |
| 137 | |
| 138 | vmx_enable(); |
| 139 | |
| 140 | VMX_KPRINTF("[%d]vmx_cpu_init() initialized: %d\n" , |
| 141 | cpu_number(), specs->initialized); |
| 142 | |
| 143 | /* if we have read the data on boot, we won't read it again on wakeup */ |
| 144 | if (specs->initialized) |
| 145 | return; |
| 146 | else |
| 147 | specs->initialized = TRUE; |
| 148 | |
| 149 | /* See if VMX is present, return if it is not */ |
| 150 | specs->vmx_present = vmx_is_available() && vmxon_is_enabled(); |
| 151 | VMX_KPRINTF("[%d]vmx_cpu_init() vmx_present: %d\n" , |
| 152 | cpu_number(), specs->vmx_present); |
| 153 | if (!specs->vmx_present) |
| 154 | return; |
| 155 | |
| 156 | #define rdmsr_mask(msr, mask) (uint32_t)(rdmsr64(msr) & (mask)) |
| 157 | specs->vmcs_id = rdmsr_mask(MSR_IA32_VMX_BASIC, VMX_VCR_VMCS_REV_ID); |
| 158 | |
| 159 | /* Obtain VMX-fixed bits in CR0 */ |
| 160 | specs->cr0_fixed_0 = rdmsr_mask(MSR_IA32_VMX_CR0_FIXED0, 0xFFFFFFFF); |
| 161 | specs->cr0_fixed_1 = rdmsr_mask(MSR_IA32_VMX_CR0_FIXED1, 0xFFFFFFFF); |
| 162 | |
| 163 | /* Obtain VMX-fixed bits in CR4 */ |
| 164 | specs->cr4_fixed_0 = rdmsr_mask(MSR_IA32_VMX_CR4_FIXED0, 0xFFFFFFFF); |
| 165 | specs->cr4_fixed_1 = rdmsr_mask(MSR_IA32_VMX_CR4_FIXED1, 0xFFFFFFFF); |
| 166 | } |
| 167 | |
| 168 | /* ----------------------------------------------------------------------------- |
| 169 | vmx_on() |
| 170 | Enter VMX root operation on this CPU. |
| 171 | -------------------------------------------------------------------------- */ |
| 172 | static void |
| 173 | vmx_on(void *arg __unused) |
| 174 | { |
| 175 | vmx_cpu_t *cpu = ¤t_cpu_datap()->cpu_vmx; |
| 176 | addr64_t vmxon_region_paddr; |
| 177 | int result; |
| 178 | |
| 179 | VMX_KPRINTF("[%d]vmx_on() entry state: %d\n" , |
| 180 | cpu_number(), cpu->specs.vmx_on); |
| 181 | |
| 182 | assert(cpu->specs.vmx_present); |
| 183 | |
| 184 | if (NULL == cpu->vmxon_region) |
| 185 | panic("vmx_on: VMXON region not allocated" ); |
| 186 | vmxon_region_paddr = vmx_paddr(cpu->vmxon_region); |
| 187 | |
| 188 | /* |
| 189 | * Enable VMX operation. |
| 190 | */ |
| 191 | if (FALSE == cpu->specs.vmx_on) { |
| 192 | assert(vmx_is_cr0_valid(&cpu->specs)); |
| 193 | assert(vmx_is_cr4_valid(&cpu->specs)); |
| 194 | |
| 195 | result = __vmxon(vmxon_region_paddr); |
| 196 | |
| 197 | if (result != VMX_SUCCEED) { |
| 198 | panic("vmx_on: unexpected return %d from __vmxon()" , result); |
| 199 | } |
| 200 | |
| 201 | cpu->specs.vmx_on = TRUE; |
| 202 | } |
| 203 | VMX_KPRINTF("[%d]vmx_on() return state: %d\n" , |
| 204 | cpu_number(), cpu->specs.vmx_on); |
| 205 | } |
| 206 | |
| 207 | /* ----------------------------------------------------------------------------- |
| 208 | vmx_off() |
| 209 | Leave VMX root operation on this CPU. |
| 210 | -------------------------------------------------------------------------- */ |
| 211 | static void |
| 212 | vmx_off(void *arg __unused) |
| 213 | { |
| 214 | vmx_cpu_t *cpu = ¤t_cpu_datap()->cpu_vmx; |
| 215 | int result; |
| 216 | |
| 217 | VMX_KPRINTF("[%d]vmx_off() entry state: %d\n" , |
| 218 | cpu_number(), cpu->specs.vmx_on); |
| 219 | |
| 220 | if (TRUE == cpu->specs.vmx_on) { |
| 221 | /* Tell the CPU to release the VMXON region */ |
| 222 | result = __vmxoff(); |
| 223 | |
| 224 | if (result != VMX_SUCCEED) { |
| 225 | panic("vmx_off: unexpected return %d from __vmxoff()" , result); |
| 226 | } |
| 227 | |
| 228 | cpu->specs.vmx_on = FALSE; |
| 229 | } |
| 230 | |
| 231 | VMX_KPRINTF("[%d]vmx_off() return state: %d\n" , |
| 232 | cpu_number(), cpu->specs.vmx_on); |
| 233 | } |
| 234 | |
| 235 | /* ----------------------------------------------------------------------------- |
| 236 | vmx_allocate_vmxon_regions() |
| 237 | Allocate, clear and init VMXON regions for all CPUs. |
| 238 | -------------------------------------------------------------------------- */ |
| 239 | static void |
| 240 | vmx_allocate_vmxon_regions(void) |
| 241 | { |
| 242 | unsigned int i; |
| 243 | |
| 244 | for (i=0; i<real_ncpus; i++) { |
| 245 | vmx_cpu_t *cpu = &cpu_datap(i)->cpu_vmx; |
| 246 | |
| 247 | /* The size is defined to be always <= 4K, so we just allocate a page */ |
| 248 | cpu->vmxon_region = vmx_pcalloc(); |
| 249 | if (NULL == cpu->vmxon_region) |
| 250 | panic("vmx_allocate_vmxon_regions: unable to allocate VMXON region" ); |
| 251 | *(uint32_t*)(cpu->vmxon_region) = cpu->specs.vmcs_id; |
| 252 | } |
| 253 | } |
| 254 | |
| 255 | /* ----------------------------------------------------------------------------- |
| 256 | vmx_free_vmxon_regions() |
| 257 | Free VMXON regions for all CPUs. |
| 258 | -------------------------------------------------------------------------- */ |
| 259 | static void |
| 260 | vmx_free_vmxon_regions(void) |
| 261 | { |
| 262 | unsigned int i; |
| 263 | |
| 264 | for (i=0; i<real_ncpus; i++) { |
| 265 | vmx_cpu_t *cpu = &cpu_datap(i)->cpu_vmx; |
| 266 | |
| 267 | vmx_pfree(cpu->vmxon_region); |
| 268 | cpu->vmxon_region = NULL; |
| 269 | } |
| 270 | } |
| 271 | |
| 272 | /* ----------------------------------------------------------------------------- |
| 273 | vmx_globally_available() |
| 274 | Checks whether VT can be turned on for all CPUs. |
| 275 | -------------------------------------------------------------------------- */ |
| 276 | static boolean_t |
| 277 | vmx_globally_available(void) |
| 278 | { |
| 279 | unsigned int i; |
| 280 | unsigned int ncpus = ml_get_max_cpus(); |
| 281 | boolean_t available = TRUE; |
| 282 | |
| 283 | for (i=0; i<ncpus; i++) { |
| 284 | vmx_cpu_t *cpu = &cpu_datap(i)->cpu_vmx; |
| 285 | |
| 286 | if (!cpu->specs.vmx_present) |
| 287 | available = FALSE; |
| 288 | } |
| 289 | VMX_KPRINTF("VMX available: %d\n" , available); |
| 290 | return available; |
| 291 | } |
| 292 | |
| 293 | |
| 294 | /* ----------------------------------------------------------------------------- |
| 295 | vmx_turn_on() |
| 296 | Turn on VT operation on all CPUs. |
| 297 | -------------------------------------------------------------------------- */ |
| 298 | int |
| 299 | host_vmxon(boolean_t exclusive) |
| 300 | { |
| 301 | int error; |
| 302 | |
| 303 | assert(0 == get_preemption_level()); |
| 304 | |
| 305 | if (!vmx_globally_available()) |
| 306 | return VMX_UNSUPPORTED; |
| 307 | |
| 308 | lck_mtx_lock(vmx_lck_mtx); |
| 309 | |
| 310 | if (vmx_exclusive || (exclusive && vmx_use_count)) { |
| 311 | error = VMX_INUSE; |
| 312 | } else { |
| 313 | if (0 == vmx_use_count) { |
| 314 | vmx_allocate_vmxon_regions(); |
| 315 | vmx_exclusive = exclusive; |
| 316 | vmx_use_count = 1; |
| 317 | mp_cpus_call(CPUMASK_ALL, ASYNC, vmx_on, NULL); |
| 318 | |
| 319 | } else { |
| 320 | vmx_use_count++; |
| 321 | } |
| 322 | |
| 323 | VMX_KPRINTF("VMX use count: %d\n" , vmx_use_count); |
| 324 | error = VMX_OK; |
| 325 | } |
| 326 | |
| 327 | lck_mtx_unlock(vmx_lck_mtx); |
| 328 | |
| 329 | return error; |
| 330 | } |
| 331 | |
| 332 | /* ----------------------------------------------------------------------------- |
| 333 | vmx_turn_off() |
| 334 | Turn off VT operation on all CPUs. |
| 335 | -------------------------------------------------------------------------- */ |
| 336 | void |
| 337 | host_vmxoff() |
| 338 | { |
| 339 | assert(0 == get_preemption_level()); |
| 340 | |
| 341 | lck_mtx_lock(vmx_lck_mtx); |
| 342 | |
| 343 | if (1 == vmx_use_count) { |
| 344 | vmx_exclusive = FALSE; |
| 345 | vmx_use_count = 0; |
| 346 | mp_cpus_call(CPUMASK_ALL, ASYNC, vmx_off, NULL); |
| 347 | vmx_free_vmxon_regions(); |
| 348 | } else { |
| 349 | vmx_use_count--; |
| 350 | } |
| 351 | |
| 352 | lck_mtx_unlock(vmx_lck_mtx); |
| 353 | |
| 354 | VMX_KPRINTF("VMX use count: %d\n" , vmx_use_count); |
| 355 | } |
| 356 | |
| 357 | /* ----------------------------------------------------------------------------- |
| 358 | vmx_suspend() |
| 359 | Turn off VT operation on this CPU if it was on. |
| 360 | Called when a CPU goes offline. |
| 361 | -------------------------------------------------------------------------- */ |
| 362 | void |
| 363 | vmx_suspend() |
| 364 | { |
| 365 | VMX_KPRINTF("vmx_suspend\n" ); |
| 366 | |
| 367 | if (vmx_use_count) |
| 368 | vmx_off(NULL); |
| 369 | } |
| 370 | |
| 371 | /* ----------------------------------------------------------------------------- |
| 372 | vmx_suspend() |
| 373 | Restore the previous VT state. Called when CPU comes back online. |
| 374 | -------------------------------------------------------------------------- */ |
| 375 | void |
| 376 | vmx_resume(boolean_t is_wake_from_hibernate) |
| 377 | { |
| 378 | VMX_KPRINTF("vmx_resume\n" ); |
| 379 | |
| 380 | vmx_enable(); |
| 381 | |
| 382 | if (vmx_use_count == 0) |
| 383 | return; |
| 384 | |
| 385 | /* |
| 386 | * When resuming from hiberate on the boot cpu, |
| 387 | * we must mark VMX as off since that's the state at wake-up |
| 388 | * because the restored state in memory records otherwise. |
| 389 | * This results in vmx_on() doing the right thing. |
| 390 | */ |
| 391 | if (is_wake_from_hibernate) { |
| 392 | vmx_cpu_t *cpu = ¤t_cpu_datap()->cpu_vmx; |
| 393 | cpu->specs.vmx_on = FALSE; |
| 394 | } |
| 395 | |
| 396 | vmx_on(NULL); |
| 397 | } |
| 398 | |
| 399 | /* ----------------------------------------------------------------------------- |
| 400 | vmx_hv_support() |
| 401 | Determine if the VMX feature set is sufficent for kernel HV support. |
| 402 | -------------------------------------------------------------------------- */ |
| 403 | boolean_t |
| 404 | vmx_hv_support() |
| 405 | { |
| 406 | if (!vmx_is_available()) |
| 407 | return FALSE; |
| 408 | |
| 409 | #define CHK(msr, shift, mask) if (!VMX_CAP(msr, shift, mask)) return FALSE; |
| 410 | |
| 411 | /* 'EPT' and 'Unrestricted Mode' are part of the secondary processor-based |
| 412 | * VM-execution controls */ |
| 413 | CHK(MSR_IA32_VMX_BASIC, 0, VMX_BASIC_TRUE_CTLS) |
| 414 | CHK(MSR_IA32_VMX_TRUE_PROCBASED_CTLS, 32, VMX_TRUE_PROCBASED_SECONDARY_CTLS) |
| 415 | |
| 416 | /* if we have these, check for 'EPT' and 'Unrestricted Mode' */ |
| 417 | CHK(MSR_IA32_VMX_PROCBASED_CTLS2, 32, VMX_PROCBASED_CTLS2_EPT) |
| 418 | CHK(MSR_IA32_VMX_PROCBASED_CTLS2, 32, VMX_PROCBASED_CTLS2_UNRESTRICTED) |
| 419 | |
| 420 | return TRUE; |
| 421 | } |
| 422 | |