include/linux/kvm_host.h

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