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 #ifndef KVM_MMIO_SIZE
  37 #define KVM_MMIO_SIZE 8
  38 #endif
  39
  40 /*
  41  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  42  * in kvm, other bits are visible for userspace which are defined in
  43  * include/linux/kvm_h.
  44  */
  45 #define KVM_MEMSLOT_INVALID     (1UL << 16)
  46
  47 /* Two fragments for cross MMIO pages. */
  48 #define KVM_MAX_MMIO_FRAGMENTS  2
  49
  50 /*
  51  * For the normal pfn, the highest 12 bits should be zero,
  52  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  53  * mask bit 63 to indicate the noslot pfn.
  54  */
  55 #define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  57 #define KVM_PFN_NOSLOT          (0x1ULL << 63)
  58
  59 #define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  60 #define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  61 #define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  62
  63 /*
  64  * error pfns indicate that the gfn is in slot but faild to
  65  * translate it to pfn on host.
  66  */
  67 static inline bool is_error_pfn(pfn_t pfn)
  68 {
  69         return !!(pfn & KVM_PFN_ERR_MASK);
  70 }
  71
  72 /*
  73  * error_noslot pfns indicate that the gfn can not be
  74  * translated to pfn - it is not in slot or failed to
  75  * translate it to pfn.
  76  */
  77 static inline bool is_error_noslot_pfn(pfn_t pfn)
  78 {
  79         return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
  80 }
  81
  82 /* noslot pfn indicates that the gfn is not in slot. */
  83 static inline bool is_noslot_pfn(pfn_t pfn)
  84 {
  85         return pfn == KVM_PFN_NOSLOT;
  86 }
  87
  88 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  89 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  90
  91 static inline bool kvm_is_error_hva(unsigned long addr)
  92 {
  93         return addr >= PAGE_OFFSET;
  94 }
  95
  96 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
  97
  98 static inline bool is_error_page(struct page *page)
  99 {
 100         return IS_ERR(page);
 101 }
 102
 103 /*
 104  * vcpu->requests bit members
 105  */
 106 #define KVM_REQ_TLB_FLUSH          0
 107 #define KVM_REQ_MIGRATE_TIMER      1
 108 #define KVM_REQ_REPORT_TPR_ACCESS  2
 109 #define KVM_REQ_MMU_RELOAD         3
 110 #define KVM_REQ_TRIPLE_FAULT       4
 111 #define KVM_REQ_PENDING_TIMER      5
 112 #define KVM_REQ_UNHALT             6
 113 #define KVM_REQ_MMU_SYNC           7
 114 #define KVM_REQ_CLOCK_UPDATE       8
 115 #define KVM_REQ_KICK               9
 116 #define KVM_REQ_DEACTIVATE_FPU    10
 117 #define KVM_REQ_EVENT             11
 118 #define KVM_REQ_APF_HALT          12
 119 #define KVM_REQ_STEAL_UPDATE      13
 120 #define KVM_REQ_NMI               14
 121 #define KVM_REQ_PMU               15
 122 #define KVM_REQ_PMI               16
 123 #define KVM_REQ_WATCHDOG          17
 124 #define KVM_REQ_MASTERCLOCK_UPDATE 18
 125 #define KVM_REQ_MCLOCK_INPROGRESS 19
 126 #define KVM_REQ_EPR_EXIT          20
 127 #define KVM_REQ_SCAN_IOAPIC       21
 128
 129 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 131
 132 struct kvm;
 133 struct kvm_vcpu;
 134 extern struct kmem_cache *kvm_vcpu_cache;
 135
 136 extern raw_spinlock_t kvm_lock;
 137 extern struct list_head vm_list;
 138
 139 struct kvm_io_range {
 140         gpa_t addr;
 141         int len;
 142         struct kvm_io_device *dev;
 143 };
 144
 145 #define NR_IOBUS_DEVS 1000
 146
 147 struct kvm_io_bus {
 148         int                   dev_count;
 149         struct kvm_io_range range[];
 150 };
 151
 152 enum kvm_bus {
 153         KVM_MMIO_BUS,
 154         KVM_PIO_BUS,
 155         KVM_VIRTIO_CCW_NOTIFY_BUS,
 156         KVM_NR_BUSES
 157 };
 158
 159 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 160                      int len, const void *val);
 161 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 162                     void *val);
 163 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 164                             int len, struct kvm_io_device *dev);
 165 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 166                               struct kvm_io_device *dev);
 167
 168 #ifdef CONFIG_KVM_ASYNC_PF
 169 struct kvm_async_pf {
 170         struct work_struct work;
 171         struct list_head link;
 172         struct list_head queue;
 173         struct kvm_vcpu *vcpu;
 174         struct mm_struct *mm;
 175         gva_t gva;
 176         unsigned long addr;
 177         struct kvm_arch_async_pf arch;
 178         struct page *page;
 179         bool done;
 180 };
 181
 182 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 183 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 184 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 185                        struct kvm_arch_async_pf *arch);
 186 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 187 #endif
 188
 189 enum {
 190         OUTSIDE_GUEST_MODE,
 191         IN_GUEST_MODE,
 192         EXITING_GUEST_MODE,
 193         READING_SHADOW_PAGE_TABLES,
 194 };
 195
 196 /*
 197  * Sometimes a large or cross-page mmio needs to be broken up into separate
 198  * exits for userspace servicing.
 199  */
 200 struct kvm_mmio_fragment {
 201         gpa_t gpa;
 202         void *data;
 203         unsigned len;
 204 };
 205
 206 struct kvm_vcpu {
 207         struct kvm *kvm;
 208 #ifdef CONFIG_PREEMPT_NOTIFIERS
 209         struct preempt_notifier preempt_notifier;
 210 #endif
 211         int cpu;
 212         int vcpu_id;
 213         int srcu_idx;
 214         int mode;
 215         unsigned long requests;
 216         unsigned long guest_debug;
 217
 218         struct mutex mutex;
 219         struct kvm_run *run;
 220
 221         int fpu_active;
 222         int guest_fpu_loaded, guest_xcr0_loaded;
 223         wait_queue_head_t wq;
 224         struct pid *pid;
 225         int sigset_active;
 226         sigset_t sigset;
 227         struct kvm_vcpu_stat stat;
 228
 229 #ifdef CONFIG_HAS_IOMEM
 230         int mmio_needed;
 231         int mmio_read_completed;
 232         int mmio_is_write;
 233         int mmio_cur_fragment;
 234         int mmio_nr_fragments;
 235         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 236 #endif
 237
 238 #ifdef CONFIG_KVM_ASYNC_PF
 239         struct {
 240                 u32 queued;
 241                 struct list_head queue;
 242                 struct list_head done;
 243                 spinlock_t lock;
 244         } async_pf;
 245 #endif
 246
 247 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 248         /*
 249          * Cpu relax intercept or pause loop exit optimization
 250          * in_spin_loop: set when a vcpu does a pause loop exit
 251          *  or cpu relax intercepted.
 252          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 253          */
 254         struct {
 255                 bool in_spin_loop;
 256                 bool dy_eligible;
 257         } spin_loop;
 258 #endif
 259         bool preempted;
 260         struct kvm_vcpu_arch arch;
 261 };
 262
 263 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 264 {
 265         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 266 }
 267
 268 /*
 269  * Some of the bitops functions do not support too long bitmaps.
 270  * This number must be determined not to exceed such limits.
 271  */
 272 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 273
 274 struct kvm_memory_slot {
 275         gfn_t base_gfn;
 276         unsigned long npages;
 277         unsigned long *dirty_bitmap;
 278         struct kvm_arch_memory_slot arch;
 279         unsigned long userspace_addr;
 280         u32 flags;
 281         short id;
 282 };
 283
 284 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 285 {
 286         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 287 }
 288
 289 struct kvm_kernel_irq_routing_entry {
 290         u32 gsi;
 291         u32 type;
 292         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 293                    struct kvm *kvm, int irq_source_id, int level,
 294                    bool line_status);
 295         union {
 296                 struct {
 297                         unsigned irqchip;
 298                         unsigned pin;
 299                 } irqchip;
 300                 struct msi_msg msi;
 301         };
 302         struct hlist_node link;
 303 };
 304
 305 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 306
 307 struct kvm_irq_routing_table {
 308         int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
 309         struct kvm_kernel_irq_routing_entry *rt_entries;
 310         u32 nr_rt_entries;
 311         /*
 312          * Array indexed by gsi. Each entry contains list of irq chips
 313          * the gsi is connected to.
 314          */
 315         struct hlist_head map[0];
 316 };
 317
 318 #else
 319
 320 struct kvm_irq_routing_table {};
 321
 322 #endif
 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 };
 343
 344 struct kvm {
 345         spinlock_t mmu_lock;
 346         struct mutex slots_lock;
 347         struct mm_struct *mm; /* userspace tied to this vm */
 348         struct kvm_memslots *memslots;
 349         struct srcu_struct srcu;
 350 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 351         u32 bsp_vcpu_id;
 352 #endif
 353         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 354         atomic_t online_vcpus;
 355         int last_boosted_vcpu;
 356         struct list_head vm_list;
 357         struct mutex lock;
 358         struct kvm_io_bus *buses[KVM_NR_BUSES];
 359 #ifdef CONFIG_HAVE_KVM_EVENTFD
 360         struct {
 361                 spinlock_t        lock;
 362                 struct list_head  items;
 363                 struct list_head  resampler_list;
 364                 struct mutex      resampler_lock;
 365         } irqfds;
 366         struct list_head ioeventfds;
 367 #endif
 368         struct kvm_vm_stat stat;
 369         struct kvm_arch arch;
 370         atomic_t users_count;
 371 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 372         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 373         spinlock_t ring_lock;
 374         struct list_head coalesced_zones;
 375 #endif
 376
 377         struct mutex irq_lock;
 378 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 379         /*
 380          * Update side is protected by irq_lock and,
 381          * if configured, irqfds.lock.
 382          */
 383         struct kvm_irq_routing_table __rcu *irq_routing;
 384         struct hlist_head mask_notifier_list;
 385         struct hlist_head irq_ack_notifier_list;
 386 #endif
 387
 388 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 389         struct mmu_notifier mmu_notifier;
 390         unsigned long mmu_notifier_seq;
 391         long mmu_notifier_count;
 392 #endif
 393         long tlbs_dirty;
 394         struct list_head devices;
 395 };
 396
 397 #define kvm_err(fmt, ...) \
 398         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 399 #define kvm_info(fmt, ...) \
 400         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 401 #define kvm_debug(fmt, ...) \
 402         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 403 #define kvm_pr_unimpl(fmt, ...) \
 404         pr_err_ratelimited("kvm [%i]: " fmt, \
 405                            task_tgid_nr(current), ## __VA_ARGS__)
 406
 407 /* The guest did something we don't support. */
 408 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 409         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 410
 411 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 412 {
 413         smp_rmb();
 414         return kvm->vcpus[i];
 415 }
 416
 417 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 418         for (idx = 0; \
 419              idx < atomic_read(&kvm->online_vcpus) && \
 420              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 421              idx++)
 422
 423 #define kvm_for_each_memslot(memslot, slots)    \
 424         for (memslot = &slots->memslots[0];     \
 425               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 426                 memslot++)
 427
 428 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 429 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 430
 431 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 432 void vcpu_put(struct kvm_vcpu *vcpu);
 433
 434 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 435 int kvm_irqfd_init(void);
 436 void kvm_irqfd_exit(void);
 437 #else
 438 static inline int kvm_irqfd_init(void)
 439 {
 440         return 0;
 441 }
 442
 443 static inline void kvm_irqfd_exit(void)
 444 {
 445 }
 446 #endif
 447 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 448                   struct module *module);
 449 void kvm_exit(void);
 450
 451 void kvm_get_kvm(struct kvm *kvm);
 452 void kvm_put_kvm(struct kvm *kvm);
 453 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
 454                      u64 last_generation);
 455
 456 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 457 {
 458         return rcu_dereference_check(kvm->memslots,
 459                         srcu_read_lock_held(&kvm->srcu)
 460                         || lockdep_is_held(&kvm->slots_lock));
 461 }
 462
 463 static inline struct kvm_memory_slot *
 464 id_to_memslot(struct kvm_memslots *slots, int id)
 465 {
 466         int index = slots->id_to_index[id];
 467         struct kvm_memory_slot *slot;
 468
 469         slot = &slots->memslots[index];
 470
 471         WARN_ON(slot->id != id);
 472         return slot;
 473 }
 474
 475 /*
 476  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
 477  * - create a new memory slot
 478  * - delete an existing memory slot
 479  * - modify an existing memory slot
 480  *   -- move it in the guest physical memory space
 481  *   -- just change its flags
 482  *
 483  * Since flags can be changed by some of these operations, the following
 484  * differentiation is the best we can do for __kvm_set_memory_region():
 485  */
 486 enum kvm_mr_change {
 487         KVM_MR_CREATE,
 488         KVM_MR_DELETE,
 489         KVM_MR_MOVE,
 490         KVM_MR_FLAGS_ONLY,
 491 };
 492
 493 int kvm_set_memory_region(struct kvm *kvm,
 494                           struct kvm_userspace_memory_region *mem);
 495 int __kvm_set_memory_region(struct kvm *kvm,
 496                             struct kvm_userspace_memory_region *mem);
 497 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 498                            struct kvm_memory_slot *dont);
 499 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 500 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 501                                 struct kvm_memory_slot *memslot,
 502                                 struct kvm_userspace_memory_region *mem,
 503                                 enum kvm_mr_change change);
 504 void kvm_arch_commit_memory_region(struct kvm *kvm,
 505                                 struct kvm_userspace_memory_region *mem,
 506                                 const struct kvm_memory_slot *old,
 507                                 enum kvm_mr_change change);
 508 bool kvm_largepages_enabled(void);
 509 void kvm_disable_largepages(void);
 510 /* flush all memory translations */
 511 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 512 /* flush memory translations pointing to 'slot' */
 513 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 514                                    struct kvm_memory_slot *slot);
 515
 516 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 517                             int nr_pages);
 518
 519 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 520 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 521 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 522 void kvm_release_page_clean(struct page *page);
 523 void kvm_release_page_dirty(struct page *page);
 524 void kvm_set_page_dirty(struct page *page);
 525 void kvm_set_page_accessed(struct page *page);
 526
 527 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 528 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 529                        bool write_fault, bool *writable);
 530 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 531 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 532                       bool *writable);
 533 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 534 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 535
 536 void kvm_release_pfn_dirty(pfn_t pfn);
 537 void kvm_release_pfn_clean(pfn_t pfn);
 538 void kvm_set_pfn_dirty(pfn_t pfn);
 539 void kvm_set_pfn_accessed(pfn_t pfn);
 540 void kvm_get_pfn(pfn_t pfn);
 541
 542 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 543                         int len);
 544 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 545                           unsigned long len);
 546 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 547 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 548                            void *data, unsigned long len);
 549 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 550                          int offset, int len);
 551 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 552                     unsigned long len);
 553 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 554                            void *data, unsigned long len);
 555 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 556                               gpa_t gpa, unsigned long len);
 557 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 558 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 559 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 560 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 561 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 562 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 563 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 564                              gfn_t gfn);
 565
 566 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 567 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 568 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 569 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 570 void kvm_resched(struct kvm_vcpu *vcpu);
 571 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 572 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 573
 574 void kvm_flush_remote_tlbs(struct kvm *kvm);
 575 void kvm_reload_remote_mmus(struct kvm *kvm);
 576 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 577 void kvm_make_scan_ioapic_request(struct kvm *kvm);
 578
 579 long kvm_arch_dev_ioctl(struct file *filp,
 580                         unsigned int ioctl, unsigned long arg);
 581 long kvm_arch_vcpu_ioctl(struct file *filp,
 582                          unsigned int ioctl, unsigned long arg);
 583 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 584
 585 int kvm_dev_ioctl_check_extension(long ext);
 586
 587 int kvm_get_dirty_log(struct kvm *kvm,
 588                         struct kvm_dirty_log *log, int *is_dirty);
 589 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 590                                 struct kvm_dirty_log *log);
 591
 592 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 593                                    struct kvm_userspace_memory_region *mem);
 594 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 595                         bool line_status);
 596 long kvm_arch_vm_ioctl(struct file *filp,
 597                        unsigned int ioctl, unsigned long arg);
 598
 599 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 600 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 601
 602 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 603                                     struct kvm_translation *tr);
 604
 605 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 606 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 607 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 608                                   struct kvm_sregs *sregs);
 609 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 610                                   struct kvm_sregs *sregs);
 611 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 612                                     struct kvm_mp_state *mp_state);
 613 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 614                                     struct kvm_mp_state *mp_state);
 615 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 616                                         struct kvm_guest_debug *dbg);
 617 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 618
 619 int kvm_arch_init(void *opaque);
 620 void kvm_arch_exit(void);
 621
 622 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 623 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 624
 625 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 626 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 627 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 628 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 629 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 630 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 631 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 632
 633 int kvm_arch_hardware_enable(void *garbage);
 634 void kvm_arch_hardware_disable(void *garbage);
 635 int kvm_arch_hardware_setup(void);
 636 void kvm_arch_hardware_unsetup(void);
 637 void kvm_arch_check_processor_compat(void *rtn);
 638 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 639 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 640
 641 void kvm_free_physmem(struct kvm *kvm);
 642
 643 void *kvm_kvzalloc(unsigned long size);
 644 void kvm_kvfree(const void *addr);
 645
 646 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 647 static inline struct kvm *kvm_arch_alloc_vm(void)
 648 {
 649         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 650 }
 651
 652 static inline void kvm_arch_free_vm(struct kvm *kvm)
 653 {
 654         kfree(kvm);
 655 }
 656 #endif
 657
 658 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 659 {
 660 #ifdef __KVM_HAVE_ARCH_WQP
 661         return vcpu->arch.wqp;
 662 #else
 663         return &vcpu->wq;
 664 #endif
 665 }
 666
 667 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 668 void kvm_arch_destroy_vm(struct kvm *kvm);
 669 void kvm_arch_sync_events(struct kvm *kvm);
 670
 671 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 672 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 673
 674 bool kvm_is_mmio_pfn(pfn_t pfn);
 675
 676 struct kvm_irq_ack_notifier {
 677         struct hlist_node link;
 678         unsigned gsi;
 679         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 680 };
 681
 682 struct kvm_assigned_dev_kernel {
 683         struct kvm_irq_ack_notifier ack_notifier;
 684         struct list_head list;
 685         int assigned_dev_id;
 686         int host_segnr;
 687         int host_busnr;
 688         int host_devfn;
 689         unsigned int entries_nr;
 690         int host_irq;
 691         bool host_irq_disabled;
 692         bool pci_2_3;
 693         struct msix_entry *host_msix_entries;
 694         int guest_irq;
 695         struct msix_entry *guest_msix_entries;
 696         unsigned long irq_requested_type;
 697         int irq_source_id;
 698         int flags;
 699         struct pci_dev *dev;
 700         struct kvm *kvm;
 701         spinlock_t intx_lock;
 702         spinlock_t intx_mask_lock;
 703         char irq_name[32];
 704         struct pci_saved_state *pci_saved_state;
 705 };
 706
 707 struct kvm_irq_mask_notifier {
 708         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 709         int irq;
 710         struct hlist_node link;
 711 };
 712
 713 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 714                                     struct kvm_irq_mask_notifier *kimn);
 715 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 716                                       struct kvm_irq_mask_notifier *kimn);
 717 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 718                              bool mask);
 719
 720 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 721                 bool line_status);
 722 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 723 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 724                 int irq_source_id, int level, bool line_status);
 725 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 726 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 727 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 728                                    struct kvm_irq_ack_notifier *kian);
 729 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 730                                    struct kvm_irq_ack_notifier *kian);
 731 int kvm_request_irq_source_id(struct kvm *kvm);
 732 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 733
 734 /* For vcpu->arch.iommu_flags */
 735 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 736
 737 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 738 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 739 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 740 int kvm_iommu_map_guest(struct kvm *kvm);
 741 int kvm_iommu_unmap_guest(struct kvm *kvm);
 742 int kvm_assign_device(struct kvm *kvm,
 743                       struct kvm_assigned_dev_kernel *assigned_dev);
 744 int kvm_deassign_device(struct kvm *kvm,
 745                         struct kvm_assigned_dev_kernel *assigned_dev);
 746 #else
 747 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 748                                       struct kvm_memory_slot *slot)
 749 {
 750         return 0;
 751 }
 752
 753 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 754                                          struct kvm_memory_slot *slot)
 755 {
 756 }
 757
 758 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 759 {
 760         return 0;
 761 }
 762 #endif
 763
 764 static inline void kvm_guest_enter(void)
 765 {
 766         unsigned long flags;
 767
 768         BUG_ON(preemptible());
 769
 770         local_irq_save(flags);
 771         guest_enter();
 772         local_irq_restore(flags);
 773
 774         /* KVM does not hold any references to rcu protected data when it
 775          * switches CPU into a guest mode. In fact switching to a guest mode
 776          * is very similar to exiting to userspase from rcu point of view. In
 777          * addition CPU may stay in a guest mode for quite a long time (up to
 778          * one time slice). Lets treat guest mode as quiescent state, just like
 779          * we do with user-mode execution.
 780          */
 781         rcu_virt_note_context_switch(smp_processor_id());
 782 }
 783
 784 static inline void kvm_guest_exit(void)
 785 {
 786         unsigned long flags;
 787
 788         local_irq_save(flags);
 789         guest_exit();
 790         local_irq_restore(flags);
 791 }
 792
 793 /*
 794  * search_memslots() and __gfn_to_memslot() are here because they are
 795  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 796  * gfn_to_memslot() itself isn't here as an inline because that would
 797  * bloat other code too much.
 798  */
 799 static inline struct kvm_memory_slot *
 800 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 801 {
 802         struct kvm_memory_slot *memslot;
 803
 804         kvm_for_each_memslot(memslot, slots)
 805                 if (gfn >= memslot->base_gfn &&
 806                       gfn < memslot->base_gfn + memslot->npages)
 807                         return memslot;
 808
 809         return NULL;
 810 }
 811
 812 static inline struct kvm_memory_slot *
 813 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 814 {
 815         return search_memslots(slots, gfn);
 816 }
 817
 818 static inline unsigned long
 819 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 820 {
 821         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 822 }
 823
 824 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 825 {
 826         return gfn_to_memslot(kvm, gfn)->id;
 827 }
 828
 829 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 830 {
 831         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 832         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 833                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 834 }
 835
 836 static inline gfn_t
 837 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 838 {
 839         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 840
 841         return slot->base_gfn + gfn_offset;
 842 }
 843
 844 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 845 {
 846         return (gpa_t)gfn << PAGE_SHIFT;
 847 }
 848
 849 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 850 {
 851         return (gfn_t)(gpa >> PAGE_SHIFT);
 852 }
 853
 854 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 855 {
 856         return (hpa_t)pfn << PAGE_SHIFT;
 857 }
 858
 859 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 860 {
 861         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 862 }
 863
 864 enum kvm_stat_kind {
 865         KVM_STAT_VM,
 866         KVM_STAT_VCPU,
 867 };
 868
 869 struct kvm_stats_debugfs_item {
 870         const char *name;
 871         int offset;
 872         enum kvm_stat_kind kind;
 873         struct dentry *dentry;
 874 };
 875 extern struct kvm_stats_debugfs_item debugfs_entries[];
 876 extern struct dentry *kvm_debugfs_dir;
 877
 878 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 879 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 880 {
 881         if (unlikely(kvm->mmu_notifier_count))
 882                 return 1;
 883         /*
 884          * Ensure the read of mmu_notifier_count happens before the read
 885          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 886          * mmu_notifier_invalidate_range_end to make sure that the caller
 887          * either sees the old (non-zero) value of mmu_notifier_count or
 888          * the new (incremented) value of mmu_notifier_seq.
 889          * PowerPC Book3s HV KVM calls this under a per-page lock
 890          * rather than under kvm->mmu_lock, for scalability, so
 891          * can't rely on kvm->mmu_lock to keep things ordered.
 892          */
 893         smp_rmb();
 894         if (kvm->mmu_notifier_seq != mmu_seq)
 895                 return 1;
 896         return 0;
 897 }
 898 #endif
 899
 900 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 901
 902 #define KVM_MAX_IRQ_ROUTES 1024
 903
 904 int kvm_setup_default_irq_routing(struct kvm *kvm);
 905 int kvm_set_irq_routing(struct kvm *kvm,
 906                         const struct kvm_irq_routing_entry *entries,
 907                         unsigned nr,
 908                         unsigned flags);
 909 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
 910                           struct kvm_kernel_irq_routing_entry *e,
 911                           const struct kvm_irq_routing_entry *ue);
 912 void kvm_free_irq_routing(struct kvm *kvm);
 913
 914 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 915
 916 #else
 917
 918 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 919
 920 #endif
 921
 922 #ifdef CONFIG_HAVE_KVM_EVENTFD
 923
 924 void kvm_eventfd_init(struct kvm *kvm);
 925 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 926
 927 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 928 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 929 void kvm_irqfd_release(struct kvm *kvm);
 930 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 931 #else
 932 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 933 {
 934         return -EINVAL;
 935 }
 936
 937 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 938 #endif
 939
 940 #else
 941
 942 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 943
 944 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 945 {
 946         return -EINVAL;
 947 }
 948
 949 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 950
 951 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 952 static inline void kvm_irq_routing_update(struct kvm *kvm,
 953                                           struct kvm_irq_routing_table *irq_rt)
 954 {
 955         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 956 }
 957 #endif
 958
 959 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 960 {
 961         return -ENOSYS;
 962 }
 963
 964 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 965
 966 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 967 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 968 {
 969         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 970 }
 971
 972 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 973
 974 #else
 975
 976 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 977
 978 #endif
 979
 980 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 981
 982 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 983                                   unsigned long arg);
 984
 985 void kvm_free_all_assigned_devices(struct kvm *kvm);
 986
 987 #else
 988
 989 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 990                                                 unsigned long arg)
 991 {
 992         return -ENOTTY;
 993 }
 994
 995 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
 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_ops;
1017
1018 struct kvm_device {
1019         struct kvm_device_ops *ops;
1020         struct kvm *kvm;
1021         void *private;
1022         struct list_head vm_node;
1023 };
1024
1025 /* create, destroy, and name are mandatory */
1026 struct kvm_device_ops {
1027         const char *name;
1028         int (*create)(struct kvm_device *dev, u32 type);
1029
1030         /*
1031          * Destroy is responsible for freeing dev.
1032          *
1033          * Destroy may be called before or after destructors are called
1034          * on emulated I/O regions, depending on whether a reference is
1035          * held by a vcpu or other kvm component that gets destroyed
1036          * after the emulated I/O.
1037          */
1038         void (*destroy)(struct kvm_device *dev);
1039
1040         int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1041         int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1042         int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1043         long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1044                       unsigned long arg);
1045 };
1046
1047 void kvm_device_get(struct kvm_device *dev);
1048 void kvm_device_put(struct kvm_device *dev);
1049 struct kvm_device *kvm_device_from_filp(struct file *filp);
1050
1051 extern struct kvm_device_ops kvm_mpic_ops;
1052 extern struct kvm_device_ops kvm_xics_ops;
1053
1054 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1055
1056 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1057 {
1058         vcpu->spin_loop.in_spin_loop = val;
1059 }
1060 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1061 {
1062         vcpu->spin_loop.dy_eligible = val;
1063 }
1064
1065 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1066
1067 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1068 {
1069 }
1070
1071 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1072 {
1073 }
1074
1075 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1076 {
1077         return true;
1078 }
1079
1080 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1081 #endif
1082