1 #ifndef __KVM_HOST_H 2 #define __KVM_HOST_H 3 4 /* 5 * This work is licensed under the terms of the GNU GPL, version 2. See 6 * the COPYING file in the top-level directory. 7 */ 8 9 #include <linux/types.h> 10 #include <linux/hardirq.h> 11 #include <linux/list.h> 12 #include <linux/mutex.h> 13 #include <linux/spinlock.h> 14 #include <linux/signal.h> 15 #include <linux/sched.h> 16 #include <linux/bug.h> 17 #include <linux/mm.h> 18 #include <linux/mmu_notifier.h> 19 #include <linux/preempt.h> 20 #include <linux/msi.h> 21 #include <linux/slab.h> 22 #include <linux/rcupdate.h> 23 #include <linux/ratelimit.h> 24 #include <linux/err.h> 25 #include <linux/irqflags.h> 26 #include <linux/context_tracking.h> 27 #include <asm/signal.h> 28 29 #include <linux/kvm.h> 30 #include <linux/kvm_para.h> 31 32 #include <linux/kvm_types.h> 33 34 #include <asm/kvm_host.h> 35 36 /* 37 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used 38 * in kvm, other bits are visible for userspace which are defined in 39 * include/linux/kvm_h. 40 */ 41 #define KVM_MEMSLOT_INVALID (1UL << 16) 42 #define KVM_MEMSLOT_INCOHERENT (1UL << 17) 43 44 /* Two fragments for cross MMIO pages. */ 45 #define KVM_MAX_MMIO_FRAGMENTS 2 46 47 #ifndef KVM_ADDRESS_SPACE_NUM 48 #define KVM_ADDRESS_SPACE_NUM 1 49 #endif 50 51 /* 52 * For the normal pfn, the highest 12 bits should be zero, 53 * so we can mask bit 62 ~ bit 52 to indicate the error pfn, 54 * mask bit 63 to indicate the noslot pfn. 55 */ 56 #define KVM_PFN_ERR_MASK (0x7ffULL << 52) 57 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52) 58 #define KVM_PFN_NOSLOT (0x1ULL << 63) 59 60 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK) 61 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1) 62 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2) 63 64 /* 65 * error pfns indicate that the gfn is in slot but faild to 66 * translate it to pfn on host. 67 */ 68 static inline bool is_error_pfn(pfn_t pfn) 69 { 70 return !!(pfn & KVM_PFN_ERR_MASK); 71 } 72 73 /* 74 * error_noslot pfns indicate that the gfn can not be 75 * translated to pfn - it is not in slot or failed to 76 * translate it to pfn. 77 */ 78 static inline bool is_error_noslot_pfn(pfn_t pfn) 79 { 80 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK); 81 } 82 83 /* noslot pfn indicates that the gfn is not in slot. */ 84 static inline bool is_noslot_pfn(pfn_t pfn) 85 { 86 return pfn == KVM_PFN_NOSLOT; 87 } 88 89 /* 90 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390) 91 * provide own defines and kvm_is_error_hva 92 */ 93 #ifndef KVM_HVA_ERR_BAD 94 95 #define KVM_HVA_ERR_BAD (PAGE_OFFSET) 96 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE) 97 98 static inline bool kvm_is_error_hva(unsigned long addr) 99 { 100 return addr >= PAGE_OFFSET; 101 } 102 103 #endif 104 105 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT)) 106 107 static inline bool is_error_page(struct page *page) 108 { 109 return IS_ERR(page); 110 } 111 112 /* 113 * vcpu->requests bit members 114 */ 115 #define KVM_REQ_TLB_FLUSH 0 116 #define KVM_REQ_MIGRATE_TIMER 1 117 #define KVM_REQ_REPORT_TPR_ACCESS 2 118 #define KVM_REQ_MMU_RELOAD 3 119 #define KVM_REQ_TRIPLE_FAULT 4 120 #define KVM_REQ_PENDING_TIMER 5 121 #define KVM_REQ_UNHALT 6 122 #define KVM_REQ_MMU_SYNC 7 123 #define KVM_REQ_CLOCK_UPDATE 8 124 #define KVM_REQ_KICK 9 125 #define KVM_REQ_DEACTIVATE_FPU 10 126 #define KVM_REQ_EVENT 11 127 #define KVM_REQ_APF_HALT 12 128 #define KVM_REQ_STEAL_UPDATE 13 129 #define KVM_REQ_NMI 14 130 #define KVM_REQ_PMU 15 131 #define KVM_REQ_PMI 16 132 #define KVM_REQ_WATCHDOG 17 133 #define KVM_REQ_MASTERCLOCK_UPDATE 18 134 #define KVM_REQ_MCLOCK_INPROGRESS 19 135 #define KVM_REQ_EPR_EXIT 20 136 #define KVM_REQ_SCAN_IOAPIC 21 137 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22 138 #define KVM_REQ_ENABLE_IBS 23 139 #define KVM_REQ_DISABLE_IBS 24 140 #define KVM_REQ_APIC_PAGE_RELOAD 25 141 #define KVM_REQ_SMI 26 142 #define KVM_REQ_HV_CRASH 27 143 144 #define KVM_USERSPACE_IRQ_SOURCE_ID 0 145 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1 146 147 extern struct kmem_cache *kvm_vcpu_cache; 148 149 extern spinlock_t kvm_lock; 150 extern struct list_head vm_list; 151 152 struct kvm_io_range { 153 gpa_t addr; 154 int len; 155 struct kvm_io_device *dev; 156 }; 157 158 #define NR_IOBUS_DEVS 1000 159 160 struct kvm_io_bus { 161 int dev_count; 162 int ioeventfd_count; 163 struct kvm_io_range range[]; 164 }; 165 166 enum kvm_bus { 167 KVM_MMIO_BUS, 168 KVM_PIO_BUS, 169 KVM_VIRTIO_CCW_NOTIFY_BUS, 170 KVM_FAST_MMIO_BUS, 171 KVM_NR_BUSES 172 }; 173 174 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr, 175 int len, const void *val); 176 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, 177 gpa_t addr, int len, const void *val, long cookie); 178 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr, 179 int len, void *val); 180 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, 181 int len, struct kvm_io_device *dev); 182 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, 183 struct kvm_io_device *dev); 184 185 #ifdef CONFIG_KVM_ASYNC_PF 186 struct kvm_async_pf { 187 struct work_struct work; 188 struct list_head link; 189 struct list_head queue; 190 struct kvm_vcpu *vcpu; 191 struct mm_struct *mm; 192 gva_t gva; 193 unsigned long addr; 194 struct kvm_arch_async_pf arch; 195 bool wakeup_all; 196 }; 197 198 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu); 199 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu); 200 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva, 201 struct kvm_arch_async_pf *arch); 202 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu); 203 #endif 204 205 enum { 206 OUTSIDE_GUEST_MODE, 207 IN_GUEST_MODE, 208 EXITING_GUEST_MODE, 209 READING_SHADOW_PAGE_TABLES, 210 }; 211 212 /* 213 * Sometimes a large or cross-page mmio needs to be broken up into separate 214 * exits for userspace servicing. 215 */ 216 struct kvm_mmio_fragment { 217 gpa_t gpa; 218 void *data; 219 unsigned len; 220 }; 221 222 struct kvm_vcpu { 223 struct kvm *kvm; 224 #ifdef CONFIG_PREEMPT_NOTIFIERS 225 struct preempt_notifier preempt_notifier; 226 #endif 227 int cpu; 228 int vcpu_id; 229 int srcu_idx; 230 int mode; 231 unsigned long requests; 232 unsigned long guest_debug; 233 234 struct mutex mutex; 235 struct kvm_run *run; 236 237 int fpu_active; 238 int guest_fpu_loaded, guest_xcr0_loaded; 239 unsigned char fpu_counter; 240 wait_queue_head_t wq; 241 struct pid *pid; 242 int sigset_active; 243 sigset_t sigset; 244 struct kvm_vcpu_stat stat; 245 246 #ifdef CONFIG_HAS_IOMEM 247 int mmio_needed; 248 int mmio_read_completed; 249 int mmio_is_write; 250 int mmio_cur_fragment; 251 int mmio_nr_fragments; 252 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS]; 253 #endif 254 255 #ifdef CONFIG_KVM_ASYNC_PF 256 struct { 257 u32 queued; 258 struct list_head queue; 259 struct list_head done; 260 spinlock_t lock; 261 } async_pf; 262 #endif 263 264 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT 265 /* 266 * Cpu relax intercept or pause loop exit optimization 267 * in_spin_loop: set when a vcpu does a pause loop exit 268 * or cpu relax intercepted. 269 * dy_eligible: indicates whether vcpu is eligible for directed yield. 270 */ 271 struct { 272 bool in_spin_loop; 273 bool dy_eligible; 274 } spin_loop; 275 #endif 276 bool preempted; 277 struct kvm_vcpu_arch arch; 278 }; 279 280 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu) 281 { 282 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE); 283 } 284 285 /* 286 * Some of the bitops functions do not support too long bitmaps. 287 * This number must be determined not to exceed such limits. 288 */ 289 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1) 290 291 struct kvm_memory_slot { 292 gfn_t base_gfn; 293 unsigned long npages; 294 unsigned long *dirty_bitmap; 295 struct kvm_arch_memory_slot arch; 296 unsigned long userspace_addr; 297 u32 flags; 298 short id; 299 }; 300 301 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot) 302 { 303 return ALIGN(memslot->npages, BITS_PER_LONG) / 8; 304 } 305 306 struct kvm_s390_adapter_int { 307 u64 ind_addr; 308 u64 summary_addr; 309 u64 ind_offset; 310 u32 summary_offset; 311 u32 adapter_id; 312 }; 313 314 struct kvm_kernel_irq_routing_entry { 315 u32 gsi; 316 u32 type; 317 int (*set)(struct kvm_kernel_irq_routing_entry *e, 318 struct kvm *kvm, int irq_source_id, int level, 319 bool line_status); 320 union { 321 struct { 322 unsigned irqchip; 323 unsigned pin; 324 } irqchip; 325 struct msi_msg msi; 326 struct kvm_s390_adapter_int adapter; 327 }; 328 struct hlist_node link; 329 }; 330 331 #ifndef KVM_PRIVATE_MEM_SLOTS 332 #define KVM_PRIVATE_MEM_SLOTS 0 333 #endif 334 335 #ifndef KVM_MEM_SLOTS_NUM 336 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS) 337 #endif 338 339 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE 340 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu) 341 { 342 return 0; 343 } 344 #endif 345 346 /* 347 * Note: 348 * memslots are not sorted by id anymore, please use id_to_memslot() 349 * to get the memslot by its id. 350 */ 351 struct kvm_memslots { 352 u64 generation; 353 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM]; 354 /* The mapping table from slot id to the index in memslots[]. */ 355 short id_to_index[KVM_MEM_SLOTS_NUM]; 356 atomic_t lru_slot; 357 int used_slots; 358 }; 359 360 struct kvm { 361 spinlock_t mmu_lock; 362 struct mutex slots_lock; 363 struct mm_struct *mm; /* userspace tied to this vm */ 364 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM]; 365 struct srcu_struct srcu; 366 struct srcu_struct irq_srcu; 367 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS]; 368 atomic_t online_vcpus; 369 int last_boosted_vcpu; 370 struct list_head vm_list; 371 struct mutex lock; 372 struct kvm_io_bus *buses[KVM_NR_BUSES]; 373 #ifdef CONFIG_HAVE_KVM_EVENTFD 374 struct { 375 spinlock_t lock; 376 struct list_head items; 377 struct list_head resampler_list; 378 struct mutex resampler_lock; 379 } irqfds; 380 struct list_head ioeventfds; 381 #endif 382 struct kvm_vm_stat stat; 383 struct kvm_arch arch; 384 atomic_t users_count; 385 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET 386 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring; 387 spinlock_t ring_lock; 388 struct list_head coalesced_zones; 389 #endif 390 391 struct mutex irq_lock; 392 #ifdef CONFIG_HAVE_KVM_IRQCHIP 393 /* 394 * Update side is protected by irq_lock. 395 */ 396 struct kvm_irq_routing_table __rcu *irq_routing; 397 #endif 398 #ifdef CONFIG_HAVE_KVM_IRQFD 399 struct hlist_head irq_ack_notifier_list; 400 #endif 401 402 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) 403 struct mmu_notifier mmu_notifier; 404 unsigned long mmu_notifier_seq; 405 long mmu_notifier_count; 406 #endif 407 long tlbs_dirty; 408 struct list_head devices; 409 }; 410 411 #define kvm_err(fmt, ...) \ 412 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) 413 #define kvm_info(fmt, ...) \ 414 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) 415 #define kvm_debug(fmt, ...) \ 416 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) 417 #define kvm_pr_unimpl(fmt, ...) \ 418 pr_err_ratelimited("kvm [%i]: " fmt, \ 419 task_tgid_nr(current), ## __VA_ARGS__) 420 421 /* The guest did something we don't support. */ 422 #define vcpu_unimpl(vcpu, fmt, ...) \ 423 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__) 424 425 #define vcpu_debug(vcpu, fmt, ...) \ 426 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__) 427 428 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i) 429 { 430 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case 431 * the caller has read kvm->online_vcpus before (as is the case 432 * for kvm_for_each_vcpu, for example). 433 */ 434 smp_rmb(); 435 return kvm->vcpus[i]; 436 } 437 438 #define kvm_for_each_vcpu(idx, vcpup, kvm) \ 439 for (idx = 0; \ 440 idx < atomic_read(&kvm->online_vcpus) && \ 441 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \ 442 idx++) 443 444 #define kvm_for_each_memslot(memslot, slots) \ 445 for (memslot = &slots->memslots[0]; \ 446 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\ 447 memslot++) 448 449 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id); 450 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu); 451 452 int __must_check vcpu_load(struct kvm_vcpu *vcpu); 453 void vcpu_put(struct kvm_vcpu *vcpu); 454 455 #ifdef __KVM_HAVE_IOAPIC 456 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm); 457 #else 458 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm) 459 { 460 } 461 #endif 462 463 #ifdef CONFIG_HAVE_KVM_IRQFD 464 int kvm_irqfd_init(void); 465 void kvm_irqfd_exit(void); 466 #else 467 static inline int kvm_irqfd_init(void) 468 { 469 return 0; 470 } 471 472 static inline void kvm_irqfd_exit(void) 473 { 474 } 475 #endif 476 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, 477 struct module *module); 478 void kvm_exit(void); 479 480 void kvm_get_kvm(struct kvm *kvm); 481 void kvm_put_kvm(struct kvm *kvm); 482 483 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id) 484 { 485 return rcu_dereference_check(kvm->memslots[as_id], 486 srcu_read_lock_held(&kvm->srcu) 487 || lockdep_is_held(&kvm->slots_lock)); 488 } 489 490 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm) 491 { 492 return __kvm_memslots(kvm, 0); 493 } 494 495 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu) 496 { 497 int as_id = kvm_arch_vcpu_memslots_id(vcpu); 498 499 return __kvm_memslots(vcpu->kvm, as_id); 500 } 501 502 static inline struct kvm_memory_slot * 503 id_to_memslot(struct kvm_memslots *slots, int id) 504 { 505 int index = slots->id_to_index[id]; 506 struct kvm_memory_slot *slot; 507 508 slot = &slots->memslots[index]; 509 510 WARN_ON(slot->id != id); 511 return slot; 512 } 513 514 /* 515 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations: 516 * - create a new memory slot 517 * - delete an existing memory slot 518 * - modify an existing memory slot 519 * -- move it in the guest physical memory space 520 * -- just change its flags 521 * 522 * Since flags can be changed by some of these operations, the following 523 * differentiation is the best we can do for __kvm_set_memory_region(): 524 */ 525 enum kvm_mr_change { 526 KVM_MR_CREATE, 527 KVM_MR_DELETE, 528 KVM_MR_MOVE, 529 KVM_MR_FLAGS_ONLY, 530 }; 531 532 int kvm_set_memory_region(struct kvm *kvm, 533 const struct kvm_userspace_memory_region *mem); 534 int __kvm_set_memory_region(struct kvm *kvm, 535 const struct kvm_userspace_memory_region *mem); 536 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free, 537 struct kvm_memory_slot *dont); 538 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot, 539 unsigned long npages); 540 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots); 541 int kvm_arch_prepare_memory_region(struct kvm *kvm, 542 struct kvm_memory_slot *memslot, 543 const struct kvm_userspace_memory_region *mem, 544 enum kvm_mr_change change); 545 void kvm_arch_commit_memory_region(struct kvm *kvm, 546 const struct kvm_userspace_memory_region *mem, 547 const struct kvm_memory_slot *old, 548 const struct kvm_memory_slot *new, 549 enum kvm_mr_change change); 550 bool kvm_largepages_enabled(void); 551 void kvm_disable_largepages(void); 552 /* flush all memory translations */ 553 void kvm_arch_flush_shadow_all(struct kvm *kvm); 554 /* flush memory translations pointing to 'slot' */ 555 void kvm_arch_flush_shadow_memslot(struct kvm *kvm, 556 struct kvm_memory_slot *slot); 557 558 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn, 559 struct page **pages, int nr_pages); 560 561 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn); 562 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn); 563 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable); 564 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn); 565 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn, 566 bool *writable); 567 void kvm_release_page_clean(struct page *page); 568 void kvm_release_page_dirty(struct page *page); 569 void kvm_set_page_accessed(struct page *page); 570 571 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn); 572 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn); 573 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, 574 bool *writable); 575 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn); 576 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn); 577 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic, 578 bool *async, bool write_fault, bool *writable); 579 580 void kvm_release_pfn_clean(pfn_t pfn); 581 void kvm_set_pfn_dirty(pfn_t pfn); 582 void kvm_set_pfn_accessed(pfn_t pfn); 583 void kvm_get_pfn(pfn_t pfn); 584 585 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, 586 int len); 587 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, 588 unsigned long len); 589 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len); 590 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, 591 void *data, unsigned long len); 592 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, 593 int offset, int len); 594 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, 595 unsigned long len); 596 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, 597 void *data, unsigned long len); 598 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, 599 gpa_t gpa, unsigned long len); 600 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len); 601 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len); 602 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn); 603 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn); 604 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn); 605 void mark_page_dirty(struct kvm *kvm, gfn_t gfn); 606 607 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu); 608 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn); 609 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn); 610 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn); 611 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn); 612 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn); 613 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable); 614 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset, 615 int len); 616 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, 617 unsigned long len); 618 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, 619 unsigned long len); 620 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data, 621 int offset, int len); 622 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data, 623 unsigned long len); 624 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn); 625 626 void kvm_vcpu_block(struct kvm_vcpu *vcpu); 627 void kvm_vcpu_kick(struct kvm_vcpu *vcpu); 628 int kvm_vcpu_yield_to(struct kvm_vcpu *target); 629 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu); 630 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu); 631 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu); 632 633 void kvm_flush_remote_tlbs(struct kvm *kvm); 634 void kvm_reload_remote_mmus(struct kvm *kvm); 635 void kvm_make_mclock_inprogress_request(struct kvm *kvm); 636 void kvm_make_scan_ioapic_request(struct kvm *kvm); 637 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req); 638 639 long kvm_arch_dev_ioctl(struct file *filp, 640 unsigned int ioctl, unsigned long arg); 641 long kvm_arch_vcpu_ioctl(struct file *filp, 642 unsigned int ioctl, unsigned long arg); 643 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf); 644 645 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext); 646 647 int kvm_get_dirty_log(struct kvm *kvm, 648 struct kvm_dirty_log *log, int *is_dirty); 649 650 int kvm_get_dirty_log_protect(struct kvm *kvm, 651 struct kvm_dirty_log *log, bool *is_dirty); 652 653 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm, 654 struct kvm_memory_slot *slot, 655 gfn_t gfn_offset, 656 unsigned long mask); 657 658 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, 659 struct kvm_dirty_log *log); 660 661 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level, 662 bool line_status); 663 long kvm_arch_vm_ioctl(struct file *filp, 664 unsigned int ioctl, unsigned long arg); 665 666 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu); 667 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu); 668 669 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, 670 struct kvm_translation *tr); 671 672 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs); 673 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs); 674 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, 675 struct kvm_sregs *sregs); 676 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, 677 struct kvm_sregs *sregs); 678 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, 679 struct kvm_mp_state *mp_state); 680 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, 681 struct kvm_mp_state *mp_state); 682 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, 683 struct kvm_guest_debug *dbg); 684 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run); 685 686 int kvm_arch_init(void *opaque); 687 void kvm_arch_exit(void); 688 689 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu); 690 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu); 691 692 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu); 693 694 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu); 695 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu); 696 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu); 697 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id); 698 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu); 699 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu); 700 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu); 701 702 int kvm_arch_hardware_enable(void); 703 void kvm_arch_hardware_disable(void); 704 int kvm_arch_hardware_setup(void); 705 void kvm_arch_hardware_unsetup(void); 706 void kvm_arch_check_processor_compat(void *rtn); 707 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu); 708 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu); 709 710 void *kvm_kvzalloc(unsigned long size); 711 712 #ifndef __KVM_HAVE_ARCH_VM_ALLOC 713 static inline struct kvm *kvm_arch_alloc_vm(void) 714 { 715 return kzalloc(sizeof(struct kvm), GFP_KERNEL); 716 } 717 718 static inline void kvm_arch_free_vm(struct kvm *kvm) 719 { 720 kfree(kvm); 721 } 722 #endif 723 724 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA 725 void kvm_arch_register_noncoherent_dma(struct kvm *kvm); 726 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm); 727 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm); 728 #else 729 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm) 730 { 731 } 732 733 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm) 734 { 735 } 736 737 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm) 738 { 739 return false; 740 } 741 #endif 742 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE 743 void kvm_arch_start_assignment(struct kvm *kvm); 744 void kvm_arch_end_assignment(struct kvm *kvm); 745 bool kvm_arch_has_assigned_device(struct kvm *kvm); 746 #else 747 static inline void kvm_arch_start_assignment(struct kvm *kvm) 748 { 749 } 750 751 static inline void kvm_arch_end_assignment(struct kvm *kvm) 752 { 753 } 754 755 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm) 756 { 757 return false; 758 } 759 #endif 760 761 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu) 762 { 763 #ifdef __KVM_HAVE_ARCH_WQP 764 return vcpu->arch.wqp; 765 #else 766 return &vcpu->wq; 767 #endif 768 } 769 770 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED 771 /* 772 * returns true if the virtual interrupt controller is initialized and 773 * ready to accept virtual IRQ. On some architectures the virtual interrupt 774 * controller is dynamically instantiated and this is not always true. 775 */ 776 bool kvm_arch_intc_initialized(struct kvm *kvm); 777 #else 778 static inline bool kvm_arch_intc_initialized(struct kvm *kvm) 779 { 780 return true; 781 } 782 #endif 783 784 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type); 785 void kvm_arch_destroy_vm(struct kvm *kvm); 786 void kvm_arch_sync_events(struct kvm *kvm); 787 788 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu); 789 void kvm_vcpu_kick(struct kvm_vcpu *vcpu); 790 791 bool kvm_is_reserved_pfn(pfn_t pfn); 792 793 struct kvm_irq_ack_notifier { 794 struct hlist_node link; 795 unsigned gsi; 796 void (*irq_acked)(struct kvm_irq_ack_notifier *kian); 797 }; 798 799 int kvm_irq_map_gsi(struct kvm *kvm, 800 struct kvm_kernel_irq_routing_entry *entries, int gsi); 801 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin); 802 803 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level, 804 bool line_status); 805 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level); 806 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm, 807 int irq_source_id, int level, bool line_status); 808 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin); 809 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin); 810 void kvm_register_irq_ack_notifier(struct kvm *kvm, 811 struct kvm_irq_ack_notifier *kian); 812 void kvm_unregister_irq_ack_notifier(struct kvm *kvm, 813 struct kvm_irq_ack_notifier *kian); 814 int kvm_request_irq_source_id(struct kvm *kvm); 815 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id); 816 817 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT 818 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot); 819 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot); 820 #else 821 static inline int kvm_iommu_map_pages(struct kvm *kvm, 822 struct kvm_memory_slot *slot) 823 { 824 return 0; 825 } 826 827 static inline void kvm_iommu_unmap_pages(struct kvm *kvm, 828 struct kvm_memory_slot *slot) 829 { 830 } 831 #endif 832 833 /* must be called with irqs disabled */ 834 static inline void __kvm_guest_enter(void) 835 { 836 guest_enter(); 837 /* KVM does not hold any references to rcu protected data when it 838 * switches CPU into a guest mode. In fact switching to a guest mode 839 * is very similar to exiting to userspace from rcu point of view. In 840 * addition CPU may stay in a guest mode for quite a long time (up to 841 * one time slice). Lets treat guest mode as quiescent state, just like 842 * we do with user-mode execution. 843 */ 844 if (!context_tracking_cpu_is_enabled()) 845 rcu_virt_note_context_switch(smp_processor_id()); 846 } 847 848 /* must be called with irqs disabled */ 849 static inline void __kvm_guest_exit(void) 850 { 851 guest_exit(); 852 } 853 854 static inline void kvm_guest_enter(void) 855 { 856 unsigned long flags; 857 858 local_irq_save(flags); 859 __kvm_guest_enter(); 860 local_irq_restore(flags); 861 } 862 863 static inline void kvm_guest_exit(void) 864 { 865 unsigned long flags; 866 867 local_irq_save(flags); 868 __kvm_guest_exit(); 869 local_irq_restore(flags); 870 } 871 872 /* 873 * search_memslots() and __gfn_to_memslot() are here because they are 874 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c. 875 * gfn_to_memslot() itself isn't here as an inline because that would 876 * bloat other code too much. 877 */ 878 static inline struct kvm_memory_slot * 879 search_memslots(struct kvm_memslots *slots, gfn_t gfn) 880 { 881 int start = 0, end = slots->used_slots; 882 int slot = atomic_read(&slots->lru_slot); 883 struct kvm_memory_slot *memslots = slots->memslots; 884 885 if (gfn >= memslots[slot].base_gfn && 886 gfn < memslots[slot].base_gfn + memslots[slot].npages) 887 return &memslots[slot]; 888 889 while (start < end) { 890 slot = start + (end - start) / 2; 891 892 if (gfn >= memslots[slot].base_gfn) 893 end = slot; 894 else 895 start = slot + 1; 896 } 897 898 if (gfn >= memslots[start].base_gfn && 899 gfn < memslots[start].base_gfn + memslots[start].npages) { 900 atomic_set(&slots->lru_slot, start); 901 return &memslots[start]; 902 } 903 904 return NULL; 905 } 906 907 static inline struct kvm_memory_slot * 908 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn) 909 { 910 return search_memslots(slots, gfn); 911 } 912 913 static inline unsigned long 914 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn) 915 { 916 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE; 917 } 918 919 static inline int memslot_id(struct kvm *kvm, gfn_t gfn) 920 { 921 return gfn_to_memslot(kvm, gfn)->id; 922 } 923 924 static inline gfn_t 925 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot) 926 { 927 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT; 928 929 return slot->base_gfn + gfn_offset; 930 } 931 932 static inline gpa_t gfn_to_gpa(gfn_t gfn) 933 { 934 return (gpa_t)gfn << PAGE_SHIFT; 935 } 936 937 static inline gfn_t gpa_to_gfn(gpa_t gpa) 938 { 939 return (gfn_t)(gpa >> PAGE_SHIFT); 940 } 941 942 static inline hpa_t pfn_to_hpa(pfn_t pfn) 943 { 944 return (hpa_t)pfn << PAGE_SHIFT; 945 } 946 947 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa) 948 { 949 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa)); 950 951 return kvm_is_error_hva(hva); 952 } 953 954 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu) 955 { 956 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests); 957 } 958 959 enum kvm_stat_kind { 960 KVM_STAT_VM, 961 KVM_STAT_VCPU, 962 }; 963 964 struct kvm_stats_debugfs_item { 965 const char *name; 966 int offset; 967 enum kvm_stat_kind kind; 968 struct dentry *dentry; 969 }; 970 extern struct kvm_stats_debugfs_item debugfs_entries[]; 971 extern struct dentry *kvm_debugfs_dir; 972 973 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) 974 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq) 975 { 976 if (unlikely(kvm->mmu_notifier_count)) 977 return 1; 978 /* 979 * Ensure the read of mmu_notifier_count happens before the read 980 * of mmu_notifier_seq. This interacts with the smp_wmb() in 981 * mmu_notifier_invalidate_range_end to make sure that the caller 982 * either sees the old (non-zero) value of mmu_notifier_count or 983 * the new (incremented) value of mmu_notifier_seq. 984 * PowerPC Book3s HV KVM calls this under a per-page lock 985 * rather than under kvm->mmu_lock, for scalability, so 986 * can't rely on kvm->mmu_lock to keep things ordered. 987 */ 988 smp_rmb(); 989 if (kvm->mmu_notifier_seq != mmu_seq) 990 return 1; 991 return 0; 992 } 993 #endif 994 995 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING 996 997 #ifdef CONFIG_S390 998 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that... 999 #else 1000 #define KVM_MAX_IRQ_ROUTES 1024 1001 #endif 1002 1003 int kvm_setup_default_irq_routing(struct kvm *kvm); 1004 int kvm_set_irq_routing(struct kvm *kvm, 1005 const struct kvm_irq_routing_entry *entries, 1006 unsigned nr, 1007 unsigned flags); 1008 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e, 1009 const struct kvm_irq_routing_entry *ue); 1010 void kvm_free_irq_routing(struct kvm *kvm); 1011 1012 #else 1013 1014 static inline void kvm_free_irq_routing(struct kvm *kvm) {} 1015 1016 #endif 1017 1018 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi); 1019 1020 #ifdef CONFIG_HAVE_KVM_EVENTFD 1021 1022 void kvm_eventfd_init(struct kvm *kvm); 1023 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args); 1024 1025 #ifdef CONFIG_HAVE_KVM_IRQFD 1026 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args); 1027 void kvm_irqfd_release(struct kvm *kvm); 1028 void kvm_irq_routing_update(struct kvm *); 1029 #else 1030 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args) 1031 { 1032 return -EINVAL; 1033 } 1034 1035 static inline void kvm_irqfd_release(struct kvm *kvm) {} 1036 #endif 1037 1038 #else 1039 1040 static inline void kvm_eventfd_init(struct kvm *kvm) {} 1041 1042 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args) 1043 { 1044 return -EINVAL; 1045 } 1046 1047 static inline void kvm_irqfd_release(struct kvm *kvm) {} 1048 1049 #ifdef CONFIG_HAVE_KVM_IRQCHIP 1050 static inline void kvm_irq_routing_update(struct kvm *kvm) 1051 { 1052 } 1053 #endif 1054 1055 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args) 1056 { 1057 return -ENOSYS; 1058 } 1059 1060 #endif /* CONFIG_HAVE_KVM_EVENTFD */ 1061 1062 #ifdef CONFIG_KVM_APIC_ARCHITECTURE 1063 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu); 1064 #else 1065 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; } 1066 #endif 1067 1068 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu) 1069 { 1070 set_bit(req, &vcpu->requests); 1071 } 1072 1073 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu) 1074 { 1075 if (test_bit(req, &vcpu->requests)) { 1076 clear_bit(req, &vcpu->requests); 1077 return true; 1078 } else { 1079 return false; 1080 } 1081 } 1082 1083 extern bool kvm_rebooting; 1084 1085 struct kvm_device { 1086 struct kvm_device_ops *ops; 1087 struct kvm *kvm; 1088 void *private; 1089 struct list_head vm_node; 1090 }; 1091 1092 /* create, destroy, and name are mandatory */ 1093 struct kvm_device_ops { 1094 const char *name; 1095 int (*create)(struct kvm_device *dev, u32 type); 1096 1097 /* 1098 * Destroy is responsible for freeing dev. 1099 * 1100 * Destroy may be called before or after destructors are called 1101 * on emulated I/O regions, depending on whether a reference is 1102 * held by a vcpu or other kvm component that gets destroyed 1103 * after the emulated I/O. 1104 */ 1105 void (*destroy)(struct kvm_device *dev); 1106 1107 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); 1108 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); 1109 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); 1110 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl, 1111 unsigned long arg); 1112 }; 1113 1114 void kvm_device_get(struct kvm_device *dev); 1115 void kvm_device_put(struct kvm_device *dev); 1116 struct kvm_device *kvm_device_from_filp(struct file *filp); 1117 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type); 1118 void kvm_unregister_device_ops(u32 type); 1119 1120 extern struct kvm_device_ops kvm_mpic_ops; 1121 extern struct kvm_device_ops kvm_xics_ops; 1122 extern struct kvm_device_ops kvm_arm_vgic_v2_ops; 1123 extern struct kvm_device_ops kvm_arm_vgic_v3_ops; 1124 1125 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT 1126 1127 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val) 1128 { 1129 vcpu->spin_loop.in_spin_loop = val; 1130 } 1131 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val) 1132 { 1133 vcpu->spin_loop.dy_eligible = val; 1134 } 1135 1136 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */ 1137 1138 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val) 1139 { 1140 } 1141 1142 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val) 1143 { 1144 } 1145 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */ 1146 #endif 1147 1148