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 /*
  89  * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
  90  * provide own defines and kvm_is_error_hva
  91  */
  92 #ifndef KVM_HVA_ERR_BAD
  93
  94 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  95 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  96
  97 static inline bool kvm_is_error_hva(unsigned long addr)
  98 {
  99         return addr >= PAGE_OFFSET;
 100 }
 101
 102 #endif
 103
 104 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
 105
 106 static inline bool is_error_page(struct page *page)
 107 {
 108         return IS_ERR(page);
 109 }
 110
 111 /*
 112  * vcpu->requests bit members
 113  */
 114 #define KVM_REQ_TLB_FLUSH          0
 115 #define KVM_REQ_MIGRATE_TIMER      1
 116 #define KVM_REQ_REPORT_TPR_ACCESS  2
 117 #define KVM_REQ_MMU_RELOAD         3
 118 #define KVM_REQ_TRIPLE_FAULT       4
 119 #define KVM_REQ_PENDING_TIMER      5
 120 #define KVM_REQ_UNHALT             6
 121 #define KVM_REQ_MMU_SYNC           7
 122 #define KVM_REQ_CLOCK_UPDATE       8
 123 #define KVM_REQ_KICK               9
 124 #define KVM_REQ_DEACTIVATE_FPU    10
 125 #define KVM_REQ_EVENT             11
 126 #define KVM_REQ_APF_HALT          12
 127 #define KVM_REQ_STEAL_UPDATE      13
 128 #define KVM_REQ_NMI               14
 129 #define KVM_REQ_PMU               15
 130 #define KVM_REQ_PMI               16
 131 #define KVM_REQ_WATCHDOG          17
 132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
 133 #define KVM_REQ_MCLOCK_INPROGRESS 19
 134 #define KVM_REQ_EPR_EXIT          20
 135 #define KVM_REQ_SCAN_IOAPIC       21
 136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
 137
 138 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 139 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 140
 141 struct kvm;
 142 struct kvm_vcpu;
 143 extern struct kmem_cache *kvm_vcpu_cache;
 144
 145 extern spinlock_t kvm_lock;
 146 extern struct list_head vm_list;
 147
 148 struct kvm_io_range {
 149         gpa_t addr;
 150         int len;
 151         struct kvm_io_device *dev;
 152 };
 153
 154 #define NR_IOBUS_DEVS 1000
 155
 156 struct kvm_io_bus {
 157         int dev_count;
 158         int ioeventfd_count;
 159         struct kvm_io_range range[];
 160 };
 161
 162 enum kvm_bus {
 163         KVM_MMIO_BUS,
 164         KVM_PIO_BUS,
 165         KVM_VIRTIO_CCW_NOTIFY_BUS,
 166         KVM_NR_BUSES
 167 };
 168
 169 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 170                      int len, const void *val);
 171 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 172                             int len, const void *val, long cookie);
 173 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 174                     void *val);
 175 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 176                             int len, struct kvm_io_device *dev);
 177 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 178                               struct kvm_io_device *dev);
 179
 180 #ifdef CONFIG_KVM_ASYNC_PF
 181 struct kvm_async_pf {
 182         struct work_struct work;
 183         struct list_head link;
 184         struct list_head queue;
 185         struct kvm_vcpu *vcpu;
 186         struct mm_struct *mm;
 187         gva_t gva;
 188         unsigned long addr;
 189         struct kvm_arch_async_pf arch;
 190         bool   wakeup_all;
 191 };
 192
 193 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 194 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 195 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
 196                        struct kvm_arch_async_pf *arch);
 197 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 198 #endif
 199
 200 enum {
 201         OUTSIDE_GUEST_MODE,
 202         IN_GUEST_MODE,
 203         EXITING_GUEST_MODE,
 204         READING_SHADOW_PAGE_TABLES,
 205 };
 206
 207 /*
 208  * Sometimes a large or cross-page mmio needs to be broken up into separate
 209  * exits for userspace servicing.
 210  */
 211 struct kvm_mmio_fragment {
 212         gpa_t gpa;
 213         void *data;
 214         unsigned len;
 215 };
 216
 217 struct kvm_vcpu {
 218         struct kvm *kvm;
 219 #ifdef CONFIG_PREEMPT_NOTIFIERS
 220         struct preempt_notifier preempt_notifier;
 221 #endif
 222         int cpu;
 223         int vcpu_id;
 224         int srcu_idx;
 225         int mode;
 226         unsigned long requests;
 227         unsigned long guest_debug;
 228
 229         struct mutex mutex;
 230         struct kvm_run *run;
 231
 232         int fpu_active;
 233         int guest_fpu_loaded, guest_xcr0_loaded;
 234         wait_queue_head_t wq;
 235         struct pid *pid;
 236         int sigset_active;
 237         sigset_t sigset;
 238         struct kvm_vcpu_stat stat;
 239
 240 #ifdef CONFIG_HAS_IOMEM
 241         int mmio_needed;
 242         int mmio_read_completed;
 243         int mmio_is_write;
 244         int mmio_cur_fragment;
 245         int mmio_nr_fragments;
 246         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 247 #endif
 248
 249 #ifdef CONFIG_KVM_ASYNC_PF
 250         struct {
 251                 u32 queued;
 252                 struct list_head queue;
 253                 struct list_head done;
 254                 spinlock_t lock;
 255         } async_pf;
 256 #endif
 257
 258 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 259         /*
 260          * Cpu relax intercept or pause loop exit optimization
 261          * in_spin_loop: set when a vcpu does a pause loop exit
 262          *  or cpu relax intercepted.
 263          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 264          */
 265         struct {
 266                 bool in_spin_loop;
 267                 bool dy_eligible;
 268         } spin_loop;
 269 #endif
 270         bool preempted;
 271         struct kvm_vcpu_arch arch;
 272 };
 273
 274 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 275 {
 276         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 277 }
 278
 279 /*
 280  * Some of the bitops functions do not support too long bitmaps.
 281  * This number must be determined not to exceed such limits.
 282  */
 283 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 284
 285 struct kvm_memory_slot {
 286         gfn_t base_gfn;
 287         unsigned long npages;
 288         unsigned long *dirty_bitmap;
 289         struct kvm_arch_memory_slot arch;
 290         unsigned long userspace_addr;
 291         u32 flags;
 292         short id;
 293 };
 294
 295 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 296 {
 297         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 298 }
 299
 300 struct kvm_kernel_irq_routing_entry {
 301         u32 gsi;
 302         u32 type;
 303         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 304                    struct kvm *kvm, int irq_source_id, int level,
 305                    bool line_status);
 306         union {
 307                 struct {
 308                         unsigned irqchip;
 309                         unsigned pin;
 310                 } irqchip;
 311                 struct msi_msg msi;
 312         };
 313         struct hlist_node link;
 314 };
 315
 316 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 317
 318 struct kvm_irq_routing_table {
 319         int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
 320         struct kvm_kernel_irq_routing_entry *rt_entries;
 321         u32 nr_rt_entries;
 322         /*
 323          * Array indexed by gsi. Each entry contains list of irq chips
 324          * the gsi is connected to.
 325          */
 326         struct hlist_head map[0];
 327 };
 328
 329 #else
 330
 331 struct kvm_irq_routing_table {};
 332
 333 #endif
 334
 335 #ifndef KVM_PRIVATE_MEM_SLOTS
 336 #define KVM_PRIVATE_MEM_SLOTS 0
 337 #endif
 338
 339 #ifndef KVM_MEM_SLOTS_NUM
 340 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 341 #endif
 342
 343 /*
 344  * Note:
 345  * memslots are not sorted by id anymore, please use id_to_memslot()
 346  * to get the memslot by its id.
 347  */
 348 struct kvm_memslots {
 349         u64 generation;
 350         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 351         /* The mapping table from slot id to the index in memslots[]. */
 352         short id_to_index[KVM_MEM_SLOTS_NUM];
 353 };
 354
 355 struct kvm {
 356         spinlock_t mmu_lock;
 357         struct mutex slots_lock;
 358         struct mm_struct *mm; /* userspace tied to this vm */
 359         struct kvm_memslots *memslots;
 360         struct srcu_struct srcu;
 361 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 362         u32 bsp_vcpu_id;
 363 #endif
 364         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 365         atomic_t online_vcpus;
 366         int last_boosted_vcpu;
 367         struct list_head vm_list;
 368         struct mutex lock;
 369         struct kvm_io_bus *buses[KVM_NR_BUSES];
 370 #ifdef CONFIG_HAVE_KVM_EVENTFD
 371         struct {
 372                 spinlock_t        lock;
 373                 struct list_head  items;
 374                 struct list_head  resampler_list;
 375                 struct mutex      resampler_lock;
 376         } irqfds;
 377         struct list_head ioeventfds;
 378 #endif
 379         struct kvm_vm_stat stat;
 380         struct kvm_arch arch;
 381         atomic_t users_count;
 382 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 383         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 384         spinlock_t ring_lock;
 385         struct list_head coalesced_zones;
 386 #endif
 387
 388         struct mutex irq_lock;
 389 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 390         /*
 391          * Update side is protected by irq_lock and,
 392          * if configured, irqfds.lock.
 393          */
 394         struct kvm_irq_routing_table __rcu *irq_routing;
 395         struct hlist_head mask_notifier_list;
 396         struct hlist_head irq_ack_notifier_list;
 397 #endif
 398
 399 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 400         struct mmu_notifier mmu_notifier;
 401         unsigned long mmu_notifier_seq;
 402         long mmu_notifier_count;
 403 #endif
 404         /* Protected by mmu_lock */
 405         bool tlbs_dirty;
 406
 407         struct list_head devices;
 408 };
 409
 410 #define kvm_err(fmt, ...) \
 411         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 412 #define kvm_info(fmt, ...) \
 413         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 414 #define kvm_debug(fmt, ...) \
 415         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 416 #define kvm_pr_unimpl(fmt, ...) \
 417         pr_err_ratelimited("kvm [%i]: " fmt, \
 418                            task_tgid_nr(current), ## __VA_ARGS__)
 419
 420 /* The guest did something we don't support. */
 421 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 422         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 423
 424 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 425 {
 426         smp_rmb();
 427         return kvm->vcpus[i];
 428 }
 429
 430 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 431         for (idx = 0; \
 432              idx < atomic_read(&kvm->online_vcpus) && \
 433              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 434              idx++)
 435
 436 #define kvm_for_each_memslot(memslot, slots)    \
 437         for (memslot = &slots->memslots[0];     \
 438               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 439                 memslot++)
 440
 441 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 442 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 443
 444 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 445 void vcpu_put(struct kvm_vcpu *vcpu);
 446
 447 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 448 int kvm_irqfd_init(void);
 449 void kvm_irqfd_exit(void);
 450 #else
 451 static inline int kvm_irqfd_init(void)
 452 {
 453         return 0;
 454 }
 455
 456 static inline void kvm_irqfd_exit(void)
 457 {
 458 }
 459 #endif
 460 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 461                   struct module *module);
 462 void kvm_exit(void);
 463
 464 void kvm_get_kvm(struct kvm *kvm);
 465 void kvm_put_kvm(struct kvm *kvm);
 466
 467 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 468 {
 469         return rcu_dereference_check(kvm->memslots,
 470                         srcu_read_lock_held(&kvm->srcu)
 471                         || lockdep_is_held(&kvm->slots_lock));
 472 }
 473
 474 static inline struct kvm_memory_slot *
 475 id_to_memslot(struct kvm_memslots *slots, int id)
 476 {
 477         int index = slots->id_to_index[id];
 478         struct kvm_memory_slot *slot;
 479
 480         slot = &slots->memslots[index];
 481
 482         WARN_ON(slot->id != id);
 483         return slot;
 484 }
 485
 486 /*
 487  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
 488  * - create a new memory slot
 489  * - delete an existing memory slot
 490  * - modify an existing memory slot
 491  *   -- move it in the guest physical memory space
 492  *   -- just change its flags
 493  *
 494  * Since flags can be changed by some of these operations, the following
 495  * differentiation is the best we can do for __kvm_set_memory_region():
 496  */
 497 enum kvm_mr_change {
 498         KVM_MR_CREATE,
 499         KVM_MR_DELETE,
 500         KVM_MR_MOVE,
 501         KVM_MR_FLAGS_ONLY,
 502 };
 503
 504 int kvm_set_memory_region(struct kvm *kvm,
 505                           struct kvm_userspace_memory_region *mem);
 506 int __kvm_set_memory_region(struct kvm *kvm,
 507                             struct kvm_userspace_memory_region *mem);
 508 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
 509                            struct kvm_memory_slot *dont);
 510 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 511                             unsigned long npages);
 512 void kvm_arch_memslots_updated(struct kvm *kvm);
 513 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 514                                 struct kvm_memory_slot *memslot,
 515                                 struct kvm_userspace_memory_region *mem,
 516                                 enum kvm_mr_change change);
 517 void kvm_arch_commit_memory_region(struct kvm *kvm,
 518                                 struct kvm_userspace_memory_region *mem,
 519                                 const struct kvm_memory_slot *old,
 520                                 enum kvm_mr_change change);
 521 bool kvm_largepages_enabled(void);
 522 void kvm_disable_largepages(void);
 523 /* flush all memory translations */
 524 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 525 /* flush memory translations pointing to 'slot' */
 526 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 527                                    struct kvm_memory_slot *slot);
 528
 529 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 530                             int nr_pages);
 531
 532 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 533 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 534 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
 535 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 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 bool 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
 588 long kvm_arch_dev_ioctl(struct file *filp,
 589                         unsigned int ioctl, unsigned long arg);
 590 long kvm_arch_vcpu_ioctl(struct file *filp,
 591                          unsigned int ioctl, unsigned long arg);
 592 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 593
 594 int kvm_dev_ioctl_check_extension(long ext);
 595
 596 int kvm_get_dirty_log(struct kvm *kvm,
 597                         struct kvm_dirty_log *log, int *is_dirty);
 598 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 599                                 struct kvm_dirty_log *log);
 600
 601 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 602                         bool line_status);
 603 long kvm_arch_vm_ioctl(struct file *filp,
 604                        unsigned int ioctl, unsigned long arg);
 605
 606 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 607 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 608
 609 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 610                                     struct kvm_translation *tr);
 611
 612 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 613 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 614 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 615                                   struct kvm_sregs *sregs);
 616 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 617                                   struct kvm_sregs *sregs);
 618 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 619                                     struct kvm_mp_state *mp_state);
 620 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 621                                     struct kvm_mp_state *mp_state);
 622 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 623                                         struct kvm_guest_debug *dbg);
 624 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 625
 626 int kvm_arch_init(void *opaque);
 627 void kvm_arch_exit(void);
 628
 629 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 630 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 631
 632 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 633 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 634 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 635 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 636 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 637 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 638 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 639
 640 int kvm_arch_hardware_enable(void *garbage);
 641 void kvm_arch_hardware_disable(void *garbage);
 642 int kvm_arch_hardware_setup(void);
 643 void kvm_arch_hardware_unsetup(void);
 644 void kvm_arch_check_processor_compat(void *rtn);
 645 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 646 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 647
 648 void *kvm_kvzalloc(unsigned long size);
 649 void kvm_kvfree(const void *addr);
 650
 651 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 652 static inline struct kvm *kvm_arch_alloc_vm(void)
 653 {
 654         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 655 }
 656
 657 static inline void kvm_arch_free_vm(struct kvm *kvm)
 658 {
 659         kfree(kvm);
 660 }
 661 #endif
 662
 663 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
 664 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
 665 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
 666 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
 667 #else
 668 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
 669 {
 670 }
 671
 672 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
 673 {
 674 }
 675
 676 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
 677 {
 678         return false;
 679 }
 680 #endif
 681
 682 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 683 {
 684 #ifdef __KVM_HAVE_ARCH_WQP
 685         return vcpu->arch.wqp;
 686 #else
 687         return &vcpu->wq;
 688 #endif
 689 }
 690
 691 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 692 void kvm_arch_destroy_vm(struct kvm *kvm);
 693 void kvm_arch_sync_events(struct kvm *kvm);
 694
 695 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 696 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 697
 698 bool kvm_is_mmio_pfn(pfn_t pfn);
 699
 700 struct kvm_irq_ack_notifier {
 701         struct hlist_node link;
 702         unsigned gsi;
 703         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 704 };
 705
 706 struct kvm_assigned_dev_kernel {
 707         struct kvm_irq_ack_notifier ack_notifier;
 708         struct list_head list;
 709         int assigned_dev_id;
 710         int host_segnr;
 711         int host_busnr;
 712         int host_devfn;
 713         unsigned int entries_nr;
 714         int host_irq;
 715         bool host_irq_disabled;
 716         bool pci_2_3;
 717         struct msix_entry *host_msix_entries;
 718         int guest_irq;
 719         struct msix_entry *guest_msix_entries;
 720         unsigned long irq_requested_type;
 721         int irq_source_id;
 722         int flags;
 723         struct pci_dev *dev;
 724         struct kvm *kvm;
 725         spinlock_t intx_lock;
 726         spinlock_t intx_mask_lock;
 727         char irq_name[32];
 728         struct pci_saved_state *pci_saved_state;
 729 };
 730
 731 struct kvm_irq_mask_notifier {
 732         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 733         int irq;
 734         struct hlist_node link;
 735 };
 736
 737 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 738                                     struct kvm_irq_mask_notifier *kimn);
 739 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 740                                       struct kvm_irq_mask_notifier *kimn);
 741 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 742                              bool mask);
 743
 744 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 745                 bool line_status);
 746 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 747 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 748                 int irq_source_id, int level, bool line_status);
 749 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 750 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 751 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 752                                    struct kvm_irq_ack_notifier *kian);
 753 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 754                                    struct kvm_irq_ack_notifier *kian);
 755 int kvm_request_irq_source_id(struct kvm *kvm);
 756 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 757
 758 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 759 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 760 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 761 int kvm_iommu_map_guest(struct kvm *kvm);
 762 int kvm_iommu_unmap_guest(struct kvm *kvm);
 763 int kvm_assign_device(struct kvm *kvm,
 764                       struct kvm_assigned_dev_kernel *assigned_dev);
 765 int kvm_deassign_device(struct kvm *kvm,
 766                         struct kvm_assigned_dev_kernel *assigned_dev);
 767 #else
 768 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 769                                       struct kvm_memory_slot *slot)
 770 {
 771         return 0;
 772 }
 773
 774 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 775                                          struct kvm_memory_slot *slot)
 776 {
 777 }
 778
 779 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 780 {
 781         return 0;
 782 }
 783 #endif
 784
 785 static inline void kvm_guest_enter(void)
 786 {
 787         unsigned long flags;
 788
 789         BUG_ON(preemptible());
 790
 791         local_irq_save(flags);
 792         guest_enter();
 793         local_irq_restore(flags);
 794
 795         /* KVM does not hold any references to rcu protected data when it
 796          * switches CPU into a guest mode. In fact switching to a guest mode
 797          * is very similar to exiting to userspace from rcu point of view. In
 798          * addition CPU may stay in a guest mode for quite a long time (up to
 799          * one time slice). Lets treat guest mode as quiescent state, just like
 800          * we do with user-mode execution.
 801          */
 802         rcu_virt_note_context_switch(smp_processor_id());
 803 }
 804
 805 static inline void kvm_guest_exit(void)
 806 {
 807         unsigned long flags;
 808
 809         local_irq_save(flags);
 810         guest_exit();
 811         local_irq_restore(flags);
 812 }
 813
 814 /*
 815  * search_memslots() and __gfn_to_memslot() are here because they are
 816  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 817  * gfn_to_memslot() itself isn't here as an inline because that would
 818  * bloat other code too much.
 819  */
 820 static inline struct kvm_memory_slot *
 821 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 822 {
 823         struct kvm_memory_slot *memslot;
 824
 825         kvm_for_each_memslot(memslot, slots)
 826                 if (gfn >= memslot->base_gfn &&
 827                       gfn < memslot->base_gfn + memslot->npages)
 828                         return memslot;
 829
 830         return NULL;
 831 }
 832
 833 static inline struct kvm_memory_slot *
 834 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 835 {
 836         return search_memslots(slots, gfn);
 837 }
 838
 839 static inline unsigned long
 840 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 841 {
 842         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 843 }
 844
 845 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 846 {
 847         return gfn_to_memslot(kvm, gfn)->id;
 848 }
 849
 850 static inline gfn_t
 851 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 852 {
 853         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 854
 855         return slot->base_gfn + gfn_offset;
 856 }
 857
 858 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 859 {
 860         return (gpa_t)gfn << PAGE_SHIFT;
 861 }
 862
 863 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 864 {
 865         return (gfn_t)(gpa >> PAGE_SHIFT);
 866 }
 867
 868 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 869 {
 870         return (hpa_t)pfn << PAGE_SHIFT;
 871 }
 872
 873 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 874 {
 875         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 876 }
 877
 878 enum kvm_stat_kind {
 879         KVM_STAT_VM,
 880         KVM_STAT_VCPU,
 881 };
 882
 883 struct kvm_stats_debugfs_item {
 884         const char *name;
 885         int offset;
 886         enum kvm_stat_kind kind;
 887         struct dentry *dentry;
 888 };
 889 extern struct kvm_stats_debugfs_item debugfs_entries[];
 890 extern struct dentry *kvm_debugfs_dir;
 891
 892 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 893 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 894 {
 895         if (unlikely(kvm->mmu_notifier_count))
 896                 return 1;
 897         /*
 898          * Ensure the read of mmu_notifier_count happens before the read
 899          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 900          * mmu_notifier_invalidate_range_end to make sure that the caller
 901          * either sees the old (non-zero) value of mmu_notifier_count or
 902          * the new (incremented) value of mmu_notifier_seq.
 903          * PowerPC Book3s HV KVM calls this under a per-page lock
 904          * rather than under kvm->mmu_lock, for scalability, so
 905          * can't rely on kvm->mmu_lock to keep things ordered.
 906          */
 907         smp_rmb();
 908         if (kvm->mmu_notifier_seq != mmu_seq)
 909                 return 1;
 910         return 0;
 911 }
 912 #endif
 913
 914 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 915
 916 #define KVM_MAX_IRQ_ROUTES 1024
 917
 918 int kvm_setup_default_irq_routing(struct kvm *kvm);
 919 int kvm_set_irq_routing(struct kvm *kvm,
 920                         const struct kvm_irq_routing_entry *entries,
 921                         unsigned nr,
 922                         unsigned flags);
 923 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
 924                           struct kvm_kernel_irq_routing_entry *e,
 925                           const struct kvm_irq_routing_entry *ue);
 926 void kvm_free_irq_routing(struct kvm *kvm);
 927
 928 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 929
 930 #else
 931
 932 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 933
 934 #endif
 935
 936 #ifdef CONFIG_HAVE_KVM_EVENTFD
 937
 938 void kvm_eventfd_init(struct kvm *kvm);
 939 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 940
 941 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 942 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 943 void kvm_irqfd_release(struct kvm *kvm);
 944 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 945 #else
 946 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 947 {
 948         return -EINVAL;
 949 }
 950
 951 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 952 #endif
 953
 954 #else
 955
 956 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 957
 958 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 959 {
 960         return -EINVAL;
 961 }
 962
 963 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 964
 965 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 966 static inline void kvm_irq_routing_update(struct kvm *kvm,
 967                                           struct kvm_irq_routing_table *irq_rt)
 968 {
 969         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 970 }
 971 #endif
 972
 973 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 974 {
 975         return -ENOSYS;
 976 }
 977
 978 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 979
 980 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 981 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 982 {
 983         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 984 }
 985
 986 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 987
 988 #else
 989
 990 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 991
 992 #endif
 993
 994 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 995
 996 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 997                                   unsigned long arg);
 998
 999 void kvm_free_all_assigned_devices(struct kvm *kvm);
1000
1001 #else
1002
1003 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1004                                                 unsigned long arg)
1005 {
1006         return -ENOTTY;
1007 }
1008
1009 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1010
1011 #endif
1012
1013 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1014 {
1015         set_bit(req, &vcpu->requests);
1016 }
1017
1018 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1019 {
1020         if (test_bit(req, &vcpu->requests)) {
1021                 clear_bit(req, &vcpu->requests);
1022                 return true;
1023         } else {
1024                 return false;
1025         }
1026 }
1027
1028 extern bool kvm_rebooting;
1029
1030 struct kvm_device_ops;
1031
1032 struct kvm_device {
1033         struct kvm_device_ops *ops;
1034         struct kvm *kvm;
1035         void *private;
1036         struct list_head vm_node;
1037 };
1038
1039 /* create, destroy, and name are mandatory */
1040 struct kvm_device_ops {
1041         const char *name;
1042         int (*create)(struct kvm_device *dev, u32 type);
1043
1044         /*
1045          * Destroy is responsible for freeing dev.
1046          *
1047          * Destroy may be called before or after destructors are called
1048          * on emulated I/O regions, depending on whether a reference is
1049          * held by a vcpu or other kvm component that gets destroyed
1050          * after the emulated I/O.
1051          */
1052         void (*destroy)(struct kvm_device *dev);
1053
1054         int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1055         int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1056         int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1057         long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1058                       unsigned long arg);
1059 };
1060
1061 void kvm_device_get(struct kvm_device *dev);
1062 void kvm_device_put(struct kvm_device *dev);
1063 struct kvm_device *kvm_device_from_filp(struct file *filp);
1064
1065 extern struct kvm_device_ops kvm_mpic_ops;
1066 extern struct kvm_device_ops kvm_xics_ops;
1067 extern struct kvm_device_ops kvm_vfio_ops;
1068 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1069 extern struct kvm_device_ops kvm_flic_ops;
1070
1071 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1072
1073 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1074 {
1075         vcpu->spin_loop.in_spin_loop = val;
1076 }
1077 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1078 {
1079         vcpu->spin_loop.dy_eligible = val;
1080 }
1081
1082 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1083
1084 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1085 {
1086 }
1087
1088 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1089 {
1090 }
1091 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1092 #endif
1093