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 <asm/signal.h>
  26
  27 #include <linux/kvm.h>
  28 #include <linux/kvm_para.h>
  29
  30 #include <linux/kvm_types.h>
  31
  32 #include <asm/kvm_host.h>
  33
  34 #ifndef KVM_MMIO_SIZE
  35 #define KVM_MMIO_SIZE 8
  36 #endif
  37
  38 /*
  39  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  40  * in kvm, other bits are visible for userspace which are defined in
  41  * include/linux/kvm_h.
  42  */
  43 #define KVM_MEMSLOT_INVALID     (1UL << 16)
  44
  45 /*
  46  * If we support unaligned MMIO, at most one fragment will be split into two:
  47  */
  48 #ifdef KVM_UNALIGNED_MMIO
  49 #  define KVM_EXTRA_MMIO_FRAGMENTS 1
  50 #else
  51 #  define KVM_EXTRA_MMIO_FRAGMENTS 0
  52 #endif
  53
  54 #define KVM_USER_MMIO_SIZE 8
  55
  56 #define KVM_MAX_MMIO_FRAGMENTS \
  57         (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
  58
  59 /*
  60  * For the normal pfn, the highest 12 bits should be zero,
  61  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  62  * mask bit 63 to indicate the noslot pfn.
  63  */
  64 #define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  65 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  66 #define KVM_PFN_NOSLOT          (0x1ULL << 63)
  67
  68 #define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  69 #define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  70 #define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  71
  72 /*
  73  * error pfns indicate that the gfn is in slot but faild to
  74  * translate it to pfn on host.
  75  */
  76 static inline bool is_error_pfn(pfn_t pfn)
  77 {
  78         return !!(pfn & KVM_PFN_ERR_MASK);
  79 }
  80
  81 /*
  82  * error_noslot pfns indicate that the gfn can not be
  83  * translated to pfn - it is not in slot or failed to
  84  * translate it to pfn.
  85  */
  86 static inline bool is_error_noslot_pfn(pfn_t pfn)
  87 {
  88         return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
  89 }
  90
  91 /* noslot pfn indicates that the gfn is not in slot. */
  92 static inline bool is_noslot_pfn(pfn_t pfn)
  93 {
  94         return pfn == KVM_PFN_NOSLOT;
  95 }
  96
  97 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  98 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  99
 100 static inline bool kvm_is_error_hva(unsigned long addr)
 101 {
 102         return addr >= PAGE_OFFSET;
 103 }
 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_IMMEDIATE_EXIT    15
 131 #define KVM_REQ_PMU               16
 132 #define KVM_REQ_PMI               17
 133 #define KVM_REQ_WATCHDOG          18
 134 #define KVM_REQ_MASTERCLOCK_UPDATE 19
 135 #define KVM_REQ_MCLOCK_INPROGRESS 20
 136
 137 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 138 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 139
 140 struct kvm;
 141 struct kvm_vcpu;
 142 extern struct kmem_cache *kvm_vcpu_cache;
 143
 144 struct kvm_io_range {
 145         gpa_t addr;
 146         int len;
 147         struct kvm_io_device *dev;
 148 };
 149
 150 #define NR_IOBUS_DEVS 1000
 151
 152 struct kvm_io_bus {
 153         int                   dev_count;
 154         struct kvm_io_range range[];
 155 };
 156
 157 enum kvm_bus {
 158         KVM_MMIO_BUS,
 159         KVM_PIO_BUS,
 160         KVM_NR_BUSES
 161 };
 162
 163 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 164                      int len, const void *val);
 165 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 166                     void *val);
 167 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 168                             int len, struct kvm_io_device *dev);
 169 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 170                               struct kvm_io_device *dev);
 171
 172 #ifdef CONFIG_KVM_ASYNC_PF
 173 struct kvm_async_pf {
 174         struct work_struct work;
 175         struct list_head link;
 176         struct list_head queue;
 177         struct kvm_vcpu *vcpu;
 178         struct mm_struct *mm;
 179         gva_t gva;
 180         unsigned long addr;
 181         struct kvm_arch_async_pf arch;
 182         struct page *page;
 183         bool done;
 184 };
 185
 186 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 187 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 188 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 189                        struct kvm_arch_async_pf *arch);
 190 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 191 #endif
 192
 193 enum {
 194         OUTSIDE_GUEST_MODE,
 195         IN_GUEST_MODE,
 196         EXITING_GUEST_MODE,
 197         READING_SHADOW_PAGE_TABLES,
 198 };
 199
 200 /*
 201  * Sometimes a large or cross-page mmio needs to be broken up into separate
 202  * exits for userspace servicing.
 203  */
 204 struct kvm_mmio_fragment {
 205         gpa_t gpa;
 206         void *data;
 207         unsigned len;
 208 };
 209
 210 struct kvm_vcpu {
 211         struct kvm *kvm;
 212 #ifdef CONFIG_PREEMPT_NOTIFIERS
 213         struct preempt_notifier preempt_notifier;
 214 #endif
 215         int cpu;
 216         int vcpu_id;
 217         int srcu_idx;
 218         int mode;
 219         unsigned long requests;
 220         unsigned long guest_debug;
 221
 222         struct mutex mutex;
 223         struct kvm_run *run;
 224
 225         int fpu_active;
 226         int guest_fpu_loaded, guest_xcr0_loaded;
 227         wait_queue_head_t wq;
 228         struct pid *pid;
 229         int sigset_active;
 230         sigset_t sigset;
 231         struct kvm_vcpu_stat stat;
 232
 233 #ifdef CONFIG_HAS_IOMEM
 234         int mmio_needed;
 235         int mmio_read_completed;
 236         int mmio_is_write;
 237         int mmio_cur_fragment;
 238         int mmio_nr_fragments;
 239         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 240 #endif
 241
 242 #ifdef CONFIG_KVM_ASYNC_PF
 243         struct {
 244                 u32 queued;
 245                 struct list_head queue;
 246                 struct list_head done;
 247                 spinlock_t lock;
 248         } async_pf;
 249 #endif
 250
 251 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 252         /*
 253          * Cpu relax intercept or pause loop exit optimization
 254          * in_spin_loop: set when a vcpu does a pause loop exit
 255          *  or cpu relax intercepted.
 256          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 257          */
 258         struct {
 259                 bool in_spin_loop;
 260                 bool dy_eligible;
 261         } spin_loop;
 262 #endif
 263         struct kvm_vcpu_arch arch;
 264 };
 265
 266 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 267 {
 268         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 269 }
 270
 271 /*
 272  * Some of the bitops functions do not support too long bitmaps.
 273  * This number must be determined not to exceed such limits.
 274  */
 275 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 276
 277 struct kvm_memory_slot {
 278         gfn_t base_gfn;
 279         unsigned long npages;
 280         unsigned long flags;
 281         unsigned long *dirty_bitmap;
 282         struct kvm_arch_memory_slot arch;
 283         unsigned long userspace_addr;
 284         int user_alloc;
 285         int id;
 286 };
 287
 288 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 289 {
 290         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 291 }
 292
 293 struct kvm_kernel_irq_routing_entry {
 294         u32 gsi;
 295         u32 type;
 296         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 297                    struct kvm *kvm, int irq_source_id, int level);
 298         union {
 299                 struct {
 300                         unsigned irqchip;
 301                         unsigned pin;
 302                 } irqchip;
 303                 struct msi_msg msi;
 304         };
 305         struct hlist_node link;
 306 };
 307
 308 #ifdef __KVM_HAVE_IOAPIC
 309
 310 struct kvm_irq_routing_table {
 311         int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
 312         struct kvm_kernel_irq_routing_entry *rt_entries;
 313         u32 nr_rt_entries;
 314         /*
 315          * Array indexed by gsi. Each entry contains list of irq chips
 316          * the gsi is connected to.
 317          */
 318         struct hlist_head map[0];
 319 };
 320
 321 #else
 322
 323 struct kvm_irq_routing_table {};
 324
 325 #endif
 326
 327 #ifndef KVM_MEM_SLOTS_NUM
 328 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 329 #endif
 330
 331 /*
 332  * Note:
 333  * memslots are not sorted by id anymore, please use id_to_memslot()
 334  * to get the memslot by its id.
 335  */
 336 struct kvm_memslots {
 337         u64 generation;
 338         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 339         /* The mapping table from slot id to the index in memslots[]. */
 340         int id_to_index[KVM_MEM_SLOTS_NUM];
 341 };
 342
 343 struct kvm {
 344         spinlock_t mmu_lock;
 345         struct mutex slots_lock;
 346         struct mm_struct *mm; /* userspace tied to this vm */
 347         struct kvm_memslots *memslots;
 348         struct srcu_struct srcu;
 349 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 350         u32 bsp_vcpu_id;
 351 #endif
 352         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 353         atomic_t online_vcpus;
 354         int last_boosted_vcpu;
 355         struct list_head vm_list;
 356         struct mutex lock;
 357         struct kvm_io_bus *buses[KVM_NR_BUSES];
 358 #ifdef CONFIG_HAVE_KVM_EVENTFD
 359         struct {
 360                 spinlock_t        lock;
 361                 struct list_head  items;
 362                 struct list_head  resampler_list;
 363                 struct mutex      resampler_lock;
 364         } irqfds;
 365         struct list_head ioeventfds;
 366 #endif
 367         struct kvm_vm_stat stat;
 368         struct kvm_arch arch;
 369         atomic_t users_count;
 370 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 371         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 372         spinlock_t ring_lock;
 373         struct list_head coalesced_zones;
 374 #endif
 375
 376         struct mutex irq_lock;
 377 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 378         /*
 379          * Update side is protected by irq_lock and,
 380          * if configured, irqfds.lock.
 381          */
 382         struct kvm_irq_routing_table __rcu *irq_routing;
 383         struct hlist_head mask_notifier_list;
 384         struct hlist_head irq_ack_notifier_list;
 385 #endif
 386
 387 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 388         struct mmu_notifier mmu_notifier;
 389         unsigned long mmu_notifier_seq;
 390         long mmu_notifier_count;
 391 #endif
 392         long tlbs_dirty;
 393 };
 394
 395 #define kvm_err(fmt, ...) \
 396         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 397 #define kvm_info(fmt, ...) \
 398         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 399 #define kvm_debug(fmt, ...) \
 400         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 401 #define kvm_pr_unimpl(fmt, ...) \
 402         pr_err_ratelimited("kvm [%i]: " fmt, \
 403                            task_tgid_nr(current), ## __VA_ARGS__)
 404
 405 /* The guest did something we don't support. */
 406 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 407         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 408
 409 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 410 {
 411         smp_rmb();
 412         return kvm->vcpus[i];
 413 }
 414
 415 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 416         for (idx = 0; \
 417              idx < atomic_read(&kvm->online_vcpus) && \
 418              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 419              idx++)
 420
 421 #define kvm_for_each_memslot(memslot, slots)    \
 422         for (memslot = &slots->memslots[0];     \
 423               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 424                 memslot++)
 425
 426 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 427 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 428
 429 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 430 void vcpu_put(struct kvm_vcpu *vcpu);
 431
 432 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 433                   struct module *module);
 434 void kvm_exit(void);
 435
 436 void kvm_get_kvm(struct kvm *kvm);
 437 void kvm_put_kvm(struct kvm *kvm);
 438 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
 439
 440 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 441 {
 442         return rcu_dereference_check(kvm->memslots,
 443                         srcu_read_lock_held(&kvm->srcu)
 444                         || lockdep_is_held(&kvm->slots_lock));
 445 }
 446
 447 static inline struct kvm_memory_slot *
 448 id_to_memslot(struct kvm_memslots *slots, int id)
 449 {
 450         int index = slots->id_to_index[id];
 451         struct kvm_memory_slot *slot;
 452
 453         slot = &slots->memslots[index];
 454
 455         WARN_ON(slot->id != id);
 456         return slot;
 457 }
 458
 459 int kvm_set_memory_region(struct kvm *kvm,
 460                           struct kvm_userspace_memory_region *mem,
 461                           int user_alloc);
 462 int __kvm_set_memory_region(struct kvm *kvm,
 463                             struct kvm_userspace_memory_region *mem,
 464                             int user_alloc);
 465 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 466                            struct kvm_memory_slot *dont);
 467 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 468 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 469                                 struct kvm_memory_slot *memslot,
 470                                 struct kvm_memory_slot old,
 471                                 struct kvm_userspace_memory_region *mem,
 472                                 int user_alloc);
 473 void kvm_arch_commit_memory_region(struct kvm *kvm,
 474                                 struct kvm_userspace_memory_region *mem,
 475                                 struct kvm_memory_slot old,
 476                                 int user_alloc);
 477 bool kvm_largepages_enabled(void);
 478 void kvm_disable_largepages(void);
 479 /* flush all memory translations */
 480 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 481 /* flush memory translations pointing to 'slot' */
 482 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 483                                    struct kvm_memory_slot *slot);
 484
 485 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 486                             int nr_pages);
 487
 488 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 489 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 490 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 491 void kvm_release_page_clean(struct page *page);
 492 void kvm_release_page_dirty(struct page *page);
 493 void kvm_set_page_dirty(struct page *page);
 494 void kvm_set_page_accessed(struct page *page);
 495
 496 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 497 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 498                        bool write_fault, bool *writable);
 499 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 500 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 501                       bool *writable);
 502 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 503 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 504
 505 void kvm_release_pfn_dirty(pfn_t pfn);
 506 void kvm_release_pfn_clean(pfn_t pfn);
 507 void kvm_set_pfn_dirty(pfn_t pfn);
 508 void kvm_set_pfn_accessed(pfn_t pfn);
 509 void kvm_get_pfn(pfn_t pfn);
 510
 511 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 512                         int len);
 513 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 514                           unsigned long len);
 515 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 516 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 517                            void *data, unsigned long len);
 518 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 519                          int offset, int len);
 520 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 521                     unsigned long len);
 522 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 523                            void *data, unsigned long len);
 524 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 525                               gpa_t gpa);
 526 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 527 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 528 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 529 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 530 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 531 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 532 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 533                              gfn_t gfn);
 534
 535 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 536 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 537 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 538 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 539 void kvm_resched(struct kvm_vcpu *vcpu);
 540 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 541 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 542
 543 void kvm_flush_remote_tlbs(struct kvm *kvm);
 544 void kvm_reload_remote_mmus(struct kvm *kvm);
 545 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 546
 547 long kvm_arch_dev_ioctl(struct file *filp,
 548                         unsigned int ioctl, unsigned long arg);
 549 long kvm_arch_vcpu_ioctl(struct file *filp,
 550                          unsigned int ioctl, unsigned long arg);
 551 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 552
 553 int kvm_dev_ioctl_check_extension(long ext);
 554
 555 int kvm_get_dirty_log(struct kvm *kvm,
 556                         struct kvm_dirty_log *log, int *is_dirty);
 557 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 558                                 struct kvm_dirty_log *log);
 559
 560 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 561                                    struct
 562                                    kvm_userspace_memory_region *mem,
 563                                    int user_alloc);
 564 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
 565 long kvm_arch_vm_ioctl(struct file *filp,
 566                        unsigned int ioctl, unsigned long arg);
 567
 568 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 569 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 570
 571 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 572                                     struct kvm_translation *tr);
 573
 574 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 575 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 576 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 577                                   struct kvm_sregs *sregs);
 578 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 579                                   struct kvm_sregs *sregs);
 580 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 581                                     struct kvm_mp_state *mp_state);
 582 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 583                                     struct kvm_mp_state *mp_state);
 584 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 585                                         struct kvm_guest_debug *dbg);
 586 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 587
 588 int kvm_arch_init(void *opaque);
 589 void kvm_arch_exit(void);
 590
 591 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 592 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 593
 594 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 595 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 596 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 597 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 598 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 599 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 600 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 601
 602 int kvm_arch_hardware_enable(void *garbage);
 603 void kvm_arch_hardware_disable(void *garbage);
 604 int kvm_arch_hardware_setup(void);
 605 void kvm_arch_hardware_unsetup(void);
 606 void kvm_arch_check_processor_compat(void *rtn);
 607 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 608 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 609
 610 void kvm_free_physmem(struct kvm *kvm);
 611
 612 void *kvm_kvzalloc(unsigned long size);
 613 void kvm_kvfree(const void *addr);
 614
 615 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 616 static inline struct kvm *kvm_arch_alloc_vm(void)
 617 {
 618         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 619 }
 620
 621 static inline void kvm_arch_free_vm(struct kvm *kvm)
 622 {
 623         kfree(kvm);
 624 }
 625 #endif
 626
 627 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 628 {
 629 #ifdef __KVM_HAVE_ARCH_WQP
 630         return vcpu->arch.wqp;
 631 #else
 632         return &vcpu->wq;
 633 #endif
 634 }
 635
 636 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 637 void kvm_arch_destroy_vm(struct kvm *kvm);
 638 void kvm_free_all_assigned_devices(struct kvm *kvm);
 639 void kvm_arch_sync_events(struct kvm *kvm);
 640
 641 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 642 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 643
 644 bool kvm_is_mmio_pfn(pfn_t pfn);
 645
 646 struct kvm_irq_ack_notifier {
 647         struct hlist_node link;
 648         unsigned gsi;
 649         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 650 };
 651
 652 struct kvm_assigned_dev_kernel {
 653         struct kvm_irq_ack_notifier ack_notifier;
 654         struct list_head list;
 655         int assigned_dev_id;
 656         int host_segnr;
 657         int host_busnr;
 658         int host_devfn;
 659         unsigned int entries_nr;
 660         int host_irq;
 661         bool host_irq_disabled;
 662         bool pci_2_3;
 663         struct msix_entry *host_msix_entries;
 664         int guest_irq;
 665         struct msix_entry *guest_msix_entries;
 666         unsigned long irq_requested_type;
 667         int irq_source_id;
 668         int flags;
 669         struct pci_dev *dev;
 670         struct kvm *kvm;
 671         spinlock_t intx_lock;
 672         spinlock_t intx_mask_lock;
 673         char irq_name[32];
 674         struct pci_saved_state *pci_saved_state;
 675 };
 676
 677 struct kvm_irq_mask_notifier {
 678         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 679         int irq;
 680         struct hlist_node link;
 681 };
 682
 683 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 684                                     struct kvm_irq_mask_notifier *kimn);
 685 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 686                                       struct kvm_irq_mask_notifier *kimn);
 687 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 688                              bool mask);
 689
 690 #ifdef __KVM_HAVE_IOAPIC
 691 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
 692                                    union kvm_ioapic_redirect_entry *entry,
 693                                    unsigned long *deliver_bitmask);
 694 #endif
 695 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 696 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 697                 int irq_source_id, int level);
 698 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 699 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 700                                    struct kvm_irq_ack_notifier *kian);
 701 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 702                                    struct kvm_irq_ack_notifier *kian);
 703 int kvm_request_irq_source_id(struct kvm *kvm);
 704 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 705
 706 /* For vcpu->arch.iommu_flags */
 707 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 708
 709 #ifdef CONFIG_IOMMU_API
 710 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 711 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 712 int kvm_iommu_map_guest(struct kvm *kvm);
 713 int kvm_iommu_unmap_guest(struct kvm *kvm);
 714 int kvm_assign_device(struct kvm *kvm,
 715                       struct kvm_assigned_dev_kernel *assigned_dev);
 716 int kvm_deassign_device(struct kvm *kvm,
 717                         struct kvm_assigned_dev_kernel *assigned_dev);
 718 #else /* CONFIG_IOMMU_API */
 719 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 720                                       struct kvm_memory_slot *slot)
 721 {
 722         return 0;
 723 }
 724
 725 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 726                                          struct kvm_memory_slot *slot)
 727 {
 728 }
 729
 730 static inline int kvm_iommu_map_guest(struct kvm *kvm)
 731 {
 732         return -ENODEV;
 733 }
 734
 735 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 736 {
 737         return 0;
 738 }
 739
 740 static inline int kvm_assign_device(struct kvm *kvm,
 741                 struct kvm_assigned_dev_kernel *assigned_dev)
 742 {
 743         return 0;
 744 }
 745
 746 static inline int kvm_deassign_device(struct kvm *kvm,
 747                 struct kvm_assigned_dev_kernel *assigned_dev)
 748 {
 749         return 0;
 750 }
 751 #endif /* CONFIG_IOMMU_API */
 752
 753 static inline void kvm_guest_enter(void)
 754 {
 755         BUG_ON(preemptible());
 756         vtime_account(current);
 757         current->flags |= PF_VCPU;
 758         /* KVM does not hold any references to rcu protected data when it
 759          * switches CPU into a guest mode. In fact switching to a guest mode
 760          * is very similar to exiting to userspase from rcu point of view. In
 761          * addition CPU may stay in a guest mode for quite a long time (up to
 762          * one time slice). Lets treat guest mode as quiescent state, just like
 763          * we do with user-mode execution.
 764          */
 765         rcu_virt_note_context_switch(smp_processor_id());
 766 }
 767
 768 static inline void kvm_guest_exit(void)
 769 {
 770         vtime_account(current);
 771         current->flags &= ~PF_VCPU;
 772 }
 773
 774 /*
 775  * search_memslots() and __gfn_to_memslot() are here because they are
 776  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 777  * gfn_to_memslot() itself isn't here as an inline because that would
 778  * bloat other code too much.
 779  */
 780 static inline struct kvm_memory_slot *
 781 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 782 {
 783         struct kvm_memory_slot *memslot;
 784
 785         kvm_for_each_memslot(memslot, slots)
 786                 if (gfn >= memslot->base_gfn &&
 787                       gfn < memslot->base_gfn + memslot->npages)
 788                         return memslot;
 789
 790         return NULL;
 791 }
 792
 793 static inline struct kvm_memory_slot *
 794 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 795 {
 796         return search_memslots(slots, gfn);
 797 }
 798
 799 static inline unsigned long
 800 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 801 {
 802         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 803 }
 804
 805 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 806 {
 807         return gfn_to_memslot(kvm, gfn)->id;
 808 }
 809
 810 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 811 {
 812         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 813         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 814                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 815 }
 816
 817 static inline gfn_t
 818 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 819 {
 820         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 821
 822         return slot->base_gfn + gfn_offset;
 823 }
 824
 825 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 826 {
 827         return (gpa_t)gfn << PAGE_SHIFT;
 828 }
 829
 830 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 831 {
 832         return (gfn_t)(gpa >> PAGE_SHIFT);
 833 }
 834
 835 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 836 {
 837         return (hpa_t)pfn << PAGE_SHIFT;
 838 }
 839
 840 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 841 {
 842         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 843 }
 844
 845 enum kvm_stat_kind {
 846         KVM_STAT_VM,
 847         KVM_STAT_VCPU,
 848 };
 849
 850 struct kvm_stats_debugfs_item {
 851         const char *name;
 852         int offset;
 853         enum kvm_stat_kind kind;
 854         struct dentry *dentry;
 855 };
 856 extern struct kvm_stats_debugfs_item debugfs_entries[];
 857 extern struct dentry *kvm_debugfs_dir;
 858
 859 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 860 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 861 {
 862         if (unlikely(kvm->mmu_notifier_count))
 863                 return 1;
 864         /*
 865          * Ensure the read of mmu_notifier_count happens before the read
 866          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 867          * mmu_notifier_invalidate_range_end to make sure that the caller
 868          * either sees the old (non-zero) value of mmu_notifier_count or
 869          * the new (incremented) value of mmu_notifier_seq.
 870          * PowerPC Book3s HV KVM calls this under a per-page lock
 871          * rather than under kvm->mmu_lock, for scalability, so
 872          * can't rely on kvm->mmu_lock to keep things ordered.
 873          */
 874         smp_rmb();
 875         if (kvm->mmu_notifier_seq != mmu_seq)
 876                 return 1;
 877         return 0;
 878 }
 879 #endif
 880
 881 #ifdef KVM_CAP_IRQ_ROUTING
 882
 883 #define KVM_MAX_IRQ_ROUTES 1024
 884
 885 int kvm_setup_default_irq_routing(struct kvm *kvm);
 886 int kvm_set_irq_routing(struct kvm *kvm,
 887                         const struct kvm_irq_routing_entry *entries,
 888                         unsigned nr,
 889                         unsigned flags);
 890 void kvm_free_irq_routing(struct kvm *kvm);
 891
 892 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 893
 894 #else
 895
 896 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 897
 898 #endif
 899
 900 #ifdef CONFIG_HAVE_KVM_EVENTFD
 901
 902 void kvm_eventfd_init(struct kvm *kvm);
 903 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 904 void kvm_irqfd_release(struct kvm *kvm);
 905 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 906 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 907
 908 #else
 909
 910 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 911
 912 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 913 {
 914         return -EINVAL;
 915 }
 916
 917 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 918
 919 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 920 static inline void kvm_irq_routing_update(struct kvm *kvm,
 921                                           struct kvm_irq_routing_table *irq_rt)
 922 {
 923         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 924 }
 925 #endif
 926
 927 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 928 {
 929         return -ENOSYS;
 930 }
 931
 932 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 933
 934 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 935 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 936 {
 937         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 938 }
 939
 940 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 941
 942 #else
 943
 944 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 945
 946 #endif
 947
 948 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
 949
 950 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 951                                   unsigned long arg);
 952
 953 #else
 954
 955 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 956                                                 unsigned long arg)
 957 {
 958         return -ENOTTY;
 959 }
 960
 961 #endif
 962
 963 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
 964 {
 965         set_bit(req, &vcpu->requests);
 966 }
 967
 968 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
 969 {
 970         if (test_bit(req, &vcpu->requests)) {
 971                 clear_bit(req, &vcpu->requests);
 972                 return true;
 973         } else {
 974                 return false;
 975         }
 976 }
 977
 978 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 979
 980 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 981 {
 982         vcpu->spin_loop.in_spin_loop = val;
 983 }
 984 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 985 {
 986         vcpu->spin_loop.dy_eligible = val;
 987 }
 988
 989 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 990
 991 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 992 {
 993 }
 994
 995 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 996 {
 997 }
 998
 999 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1000 {
1001         return true;
1002 }
1003
1004 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1005 #endif
1006