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 raw_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_read_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 176                            int len, void *val, long cookie);
 177 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 178                             int len, struct kvm_io_device *dev);
 179 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 180                               struct kvm_io_device *dev);
 181
 182 #ifdef CONFIG_KVM_ASYNC_PF
 183 struct kvm_async_pf {
 184         struct work_struct work;
 185         struct list_head link;
 186         struct list_head queue;
 187         struct kvm_vcpu *vcpu;
 188         struct mm_struct *mm;
 189         gva_t gva;
 190         unsigned long addr;
 191         struct kvm_arch_async_pf arch;
 192         struct page *page;
 193         bool done;
 194 };
 195
 196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 197 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 198 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 199                        struct kvm_arch_async_pf *arch);
 200 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 201 #endif
 202
 203 enum {
 204         OUTSIDE_GUEST_MODE,
 205         IN_GUEST_MODE,
 206         EXITING_GUEST_MODE,
 207         READING_SHADOW_PAGE_TABLES,
 208 };
 209
 210 /*
 211  * Sometimes a large or cross-page mmio needs to be broken up into separate
 212  * exits for userspace servicing.
 213  */
 214 struct kvm_mmio_fragment {
 215         gpa_t gpa;
 216         void *data;
 217         unsigned len;
 218 };
 219
 220 struct kvm_vcpu {
 221         struct kvm *kvm;
 222 #ifdef CONFIG_PREEMPT_NOTIFIERS
 223         struct preempt_notifier preempt_notifier;
 224 #endif
 225         int cpu;
 226         int vcpu_id;
 227         int srcu_idx;
 228         int mode;
 229         unsigned long requests;
 230         unsigned long guest_debug;
 231
 232         struct mutex mutex;
 233         struct kvm_run *run;
 234
 235         int fpu_active;
 236         int guest_fpu_loaded, guest_xcr0_loaded;
 237         wait_queue_head_t wq;
 238         struct pid *pid;
 239         int sigset_active;
 240         sigset_t sigset;
 241         struct kvm_vcpu_stat stat;
 242
 243 #ifdef CONFIG_HAS_IOMEM
 244         int mmio_needed;
 245         int mmio_read_completed;
 246         int mmio_is_write;
 247         int mmio_cur_fragment;
 248         int mmio_nr_fragments;
 249         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 250 #endif
 251
 252 #ifdef CONFIG_KVM_ASYNC_PF
 253         struct {
 254                 u32 queued;
 255                 struct list_head queue;
 256                 struct list_head done;
 257                 spinlock_t lock;
 258         } async_pf;
 259 #endif
 260
 261 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 262         /*
 263          * Cpu relax intercept or pause loop exit optimization
 264          * in_spin_loop: set when a vcpu does a pause loop exit
 265          *  or cpu relax intercepted.
 266          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 267          */
 268         struct {
 269                 bool in_spin_loop;
 270                 bool dy_eligible;
 271         } spin_loop;
 272 #endif
 273         bool preempted;
 274         struct kvm_vcpu_arch arch;
 275 };
 276
 277 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 278 {
 279         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 280 }
 281
 282 /*
 283  * Some of the bitops functions do not support too long bitmaps.
 284  * This number must be determined not to exceed such limits.
 285  */
 286 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 287
 288 struct kvm_memory_slot {
 289         gfn_t base_gfn;
 290         unsigned long npages;
 291         unsigned long *dirty_bitmap;
 292         struct kvm_arch_memory_slot arch;
 293         unsigned long userspace_addr;
 294         u32 flags;
 295         short id;
 296 };
 297
 298 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 299 {
 300         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 301 }
 302
 303 struct kvm_kernel_irq_routing_entry {
 304         u32 gsi;
 305         u32 type;
 306         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 307                    struct kvm *kvm, int irq_source_id, int level,
 308                    bool line_status);
 309         union {
 310                 struct {
 311                         unsigned irqchip;
 312                         unsigned pin;
 313                 } irqchip;
 314                 struct msi_msg msi;
 315         };
 316         struct hlist_node link;
 317 };
 318
 319 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 320
 321 struct kvm_irq_routing_table {
 322         int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
 323         struct kvm_kernel_irq_routing_entry *rt_entries;
 324         u32 nr_rt_entries;
 325         /*
 326          * Array indexed by gsi. Each entry contains list of irq chips
 327          * the gsi is connected to.
 328          */
 329         struct hlist_head map[0];
 330 };
 331
 332 #else
 333
 334 struct kvm_irq_routing_table {};
 335
 336 #endif
 337
 338 #ifndef KVM_PRIVATE_MEM_SLOTS
 339 #define KVM_PRIVATE_MEM_SLOTS 0
 340 #endif
 341
 342 #ifndef KVM_MEM_SLOTS_NUM
 343 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 344 #endif
 345
 346 /*
 347  * Note:
 348  * memslots are not sorted by id anymore, please use id_to_memslot()
 349  * to get the memslot by its id.
 350  */
 351 struct kvm_memslots {
 352         u64 generation;
 353         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 354         /* The mapping table from slot id to the index in memslots[]. */
 355         short id_to_index[KVM_MEM_SLOTS_NUM];
 356 };
 357
 358 struct kvm {
 359         spinlock_t mmu_lock;
 360         struct mutex slots_lock;
 361         struct mm_struct *mm; /* userspace tied to this vm */
 362         struct kvm_memslots *memslots;
 363         struct srcu_struct srcu;
 364 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 365         u32 bsp_vcpu_id;
 366 #endif
 367         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 368         atomic_t online_vcpus;
 369         int last_boosted_vcpu;
 370         struct list_head vm_list;
 371         struct mutex lock;
 372         struct kvm_io_bus *buses[KVM_NR_BUSES];
 373 #ifdef CONFIG_HAVE_KVM_EVENTFD
 374         struct {
 375                 spinlock_t        lock;
 376                 struct list_head  items;
 377                 struct list_head  resampler_list;
 378                 struct mutex      resampler_lock;
 379         } irqfds;
 380         struct list_head ioeventfds;
 381 #endif
 382         struct kvm_vm_stat stat;
 383         struct kvm_arch arch;
 384         atomic_t users_count;
 385 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 386         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 387         spinlock_t ring_lock;
 388         struct list_head coalesced_zones;
 389 #endif
 390
 391         struct mutex irq_lock;
 392 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 393         /*
 394          * Update side is protected by irq_lock and,
 395          * if configured, irqfds.lock.
 396          */
 397         struct kvm_irq_routing_table __rcu *irq_routing;
 398         struct hlist_head mask_notifier_list;
 399         struct hlist_head irq_ack_notifier_list;
 400 #endif
 401
 402 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 403         struct mmu_notifier mmu_notifier;
 404         unsigned long mmu_notifier_seq;
 405         long mmu_notifier_count;
 406 #endif
 407         long tlbs_dirty;
 408         struct list_head devices;
 409 };
 410
 411 #define kvm_err(fmt, ...) \
 412         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 413 #define kvm_info(fmt, ...) \
 414         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 415 #define kvm_debug(fmt, ...) \
 416         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 417 #define kvm_pr_unimpl(fmt, ...) \
 418         pr_err_ratelimited("kvm [%i]: " fmt, \
 419                            task_tgid_nr(current), ## __VA_ARGS__)
 420
 421 /* The guest did something we don't support. */
 422 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 423         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 424
 425 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 426 {
 427         smp_rmb();
 428         return kvm->vcpus[i];
 429 }
 430
 431 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 432         for (idx = 0; \
 433              idx < atomic_read(&kvm->online_vcpus) && \
 434              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 435              idx++)
 436
 437 #define kvm_for_each_memslot(memslot, slots)    \
 438         for (memslot = &slots->memslots[0];     \
 439               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 440                 memslot++)
 441
 442 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 443 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 444
 445 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 446 void vcpu_put(struct kvm_vcpu *vcpu);
 447
 448 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 449 int kvm_irqfd_init(void);
 450 void kvm_irqfd_exit(void);
 451 #else
 452 static inline int kvm_irqfd_init(void)
 453 {
 454         return 0;
 455 }
 456
 457 static inline void kvm_irqfd_exit(void)
 458 {
 459 }
 460 #endif
 461 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 462                   struct module *module);
 463 void kvm_exit(void);
 464
 465 void kvm_get_kvm(struct kvm *kvm);
 466 void kvm_put_kvm(struct kvm *kvm);
 467 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
 468                      u64 last_generation);
 469
 470 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 471 {
 472         return rcu_dereference_check(kvm->memslots,
 473                         srcu_read_lock_held(&kvm->srcu)
 474                         || lockdep_is_held(&kvm->slots_lock));
 475 }
 476
 477 static inline struct kvm_memory_slot *
 478 id_to_memslot(struct kvm_memslots *slots, int id)
 479 {
 480         int index = slots->id_to_index[id];
 481         struct kvm_memory_slot *slot;
 482
 483         slot = &slots->memslots[index];
 484
 485         WARN_ON(slot->id != id);
 486         return slot;
 487 }
 488
 489 /*
 490  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
 491  * - create a new memory slot
 492  * - delete an existing memory slot
 493  * - modify an existing memory slot
 494  *   -- move it in the guest physical memory space
 495  *   -- just change its flags
 496  *
 497  * Since flags can be changed by some of these operations, the following
 498  * differentiation is the best we can do for __kvm_set_memory_region():
 499  */
 500 enum kvm_mr_change {
 501         KVM_MR_CREATE,
 502         KVM_MR_DELETE,
 503         KVM_MR_MOVE,
 504         KVM_MR_FLAGS_ONLY,
 505 };
 506
 507 int kvm_set_memory_region(struct kvm *kvm,
 508                           struct kvm_userspace_memory_region *mem);
 509 int __kvm_set_memory_region(struct kvm *kvm,
 510                             struct kvm_userspace_memory_region *mem);
 511 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 512                            struct kvm_memory_slot *dont);
 513 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 514 void kvm_arch_memslots_updated(struct kvm *kvm);
 515 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 516                                 struct kvm_memory_slot *memslot,
 517                                 struct kvm_userspace_memory_region *mem,
 518                                 enum kvm_mr_change change);
 519 void kvm_arch_commit_memory_region(struct kvm *kvm,
 520                                 struct kvm_userspace_memory_region *mem,
 521                                 const struct kvm_memory_slot *old,
 522                                 enum kvm_mr_change change);
 523 bool kvm_largepages_enabled(void);
 524 void kvm_disable_largepages(void);
 525 /* flush all memory translations */
 526 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 527 /* flush memory translations pointing to 'slot' */
 528 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 529                                    struct kvm_memory_slot *slot);
 530
 531 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 532                             int nr_pages);
 533
 534 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 535 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 536 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
 537 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 538 void kvm_release_page_clean(struct page *page);
 539 void kvm_release_page_dirty(struct page *page);
 540 void kvm_set_page_dirty(struct page *page);
 541 void kvm_set_page_accessed(struct page *page);
 542
 543 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 544 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 545                        bool write_fault, bool *writable);
 546 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 547 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 548                       bool *writable);
 549 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 550 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 551
 552 void kvm_release_pfn_dirty(pfn_t pfn);
 553 void kvm_release_pfn_clean(pfn_t pfn);
 554 void kvm_set_pfn_dirty(pfn_t pfn);
 555 void kvm_set_pfn_accessed(pfn_t pfn);
 556 void kvm_get_pfn(pfn_t pfn);
 557
 558 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 559                         int len);
 560 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 561                           unsigned long len);
 562 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 563 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 564                            void *data, unsigned long len);
 565 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 566                          int offset, int len);
 567 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 568                     unsigned long len);
 569 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 570                            void *data, unsigned long len);
 571 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 572                               gpa_t gpa, unsigned long len);
 573 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 574 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 575 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 576 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 577 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 578 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 579 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 580                              gfn_t gfn);
 581
 582 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 583 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 584 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 585 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 586 void kvm_resched(struct kvm_vcpu *vcpu);
 587 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 588 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 589
 590 void kvm_flush_remote_tlbs(struct kvm *kvm);
 591 void kvm_reload_remote_mmus(struct kvm *kvm);
 592 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 593 void kvm_make_scan_ioapic_request(struct kvm *kvm);
 594
 595 long kvm_arch_dev_ioctl(struct file *filp,
 596                         unsigned int ioctl, unsigned long arg);
 597 long kvm_arch_vcpu_ioctl(struct file *filp,
 598                          unsigned int ioctl, unsigned long arg);
 599 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 600
 601 int kvm_dev_ioctl_check_extension(long ext);
 602
 603 int kvm_get_dirty_log(struct kvm *kvm,
 604                         struct kvm_dirty_log *log, int *is_dirty);
 605 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 606                                 struct kvm_dirty_log *log);
 607
 608 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 609                                    struct kvm_userspace_memory_region *mem);
 610 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 611                         bool line_status);
 612 long kvm_arch_vm_ioctl(struct file *filp,
 613                        unsigned int ioctl, unsigned long arg);
 614
 615 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 616 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 617
 618 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 619                                     struct kvm_translation *tr);
 620
 621 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 622 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 623 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 624                                   struct kvm_sregs *sregs);
 625 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 626                                   struct kvm_sregs *sregs);
 627 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 628                                     struct kvm_mp_state *mp_state);
 629 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 630                                     struct kvm_mp_state *mp_state);
 631 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 632                                         struct kvm_guest_debug *dbg);
 633 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 634
 635 int kvm_arch_init(void *opaque);
 636 void kvm_arch_exit(void);
 637
 638 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 639 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 640
 641 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 642 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 643 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 644 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 645 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 646 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 647 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 648
 649 int kvm_arch_hardware_enable(void *garbage);
 650 void kvm_arch_hardware_disable(void *garbage);
 651 int kvm_arch_hardware_setup(void);
 652 void kvm_arch_hardware_unsetup(void);
 653 void kvm_arch_check_processor_compat(void *rtn);
 654 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 655 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 656
 657 void kvm_free_physmem(struct kvm *kvm);
 658
 659 void *kvm_kvzalloc(unsigned long size);
 660 void kvm_kvfree(const void *addr);
 661
 662 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 663 static inline struct kvm *kvm_arch_alloc_vm(void)
 664 {
 665         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 666 }
 667
 668 static inline void kvm_arch_free_vm(struct kvm *kvm)
 669 {
 670         kfree(kvm);
 671 }
 672 #endif
 673
 674 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 675 {
 676 #ifdef __KVM_HAVE_ARCH_WQP
 677         return vcpu->arch.wqp;
 678 #else
 679         return &vcpu->wq;
 680 #endif
 681 }
 682
 683 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 684 void kvm_arch_destroy_vm(struct kvm *kvm);
 685 void kvm_arch_sync_events(struct kvm *kvm);
 686
 687 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 688 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 689
 690 bool kvm_is_mmio_pfn(pfn_t pfn);
 691
 692 struct kvm_irq_ack_notifier {
 693         struct hlist_node link;
 694         unsigned gsi;
 695         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 696 };
 697
 698 struct kvm_assigned_dev_kernel {
 699         struct kvm_irq_ack_notifier ack_notifier;
 700         struct list_head list;
 701         int assigned_dev_id;
 702         int host_segnr;
 703         int host_busnr;
 704         int host_devfn;
 705         unsigned int entries_nr;
 706         int host_irq;
 707         bool host_irq_disabled;
 708         bool pci_2_3;
 709         struct msix_entry *host_msix_entries;
 710         int guest_irq;
 711         struct msix_entry *guest_msix_entries;
 712         unsigned long irq_requested_type;
 713         int irq_source_id;
 714         int flags;
 715         struct pci_dev *dev;
 716         struct kvm *kvm;
 717         spinlock_t intx_lock;
 718         spinlock_t intx_mask_lock;
 719         char irq_name[32];
 720         struct pci_saved_state *pci_saved_state;
 721 };
 722
 723 struct kvm_irq_mask_notifier {
 724         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 725         int irq;
 726         struct hlist_node link;
 727 };
 728
 729 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 730                                     struct kvm_irq_mask_notifier *kimn);
 731 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 732                                       struct kvm_irq_mask_notifier *kimn);
 733 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 734                              bool mask);
 735
 736 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 737                 bool line_status);
 738 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 739 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 740                 int irq_source_id, int level, bool line_status);
 741 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 742 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 743 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 744                                    struct kvm_irq_ack_notifier *kian);
 745 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 746                                    struct kvm_irq_ack_notifier *kian);
 747 int kvm_request_irq_source_id(struct kvm *kvm);
 748 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 749
 750 /* For vcpu->arch.iommu_flags */
 751 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 752
 753 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 754 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 755 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 756 int kvm_iommu_map_guest(struct kvm *kvm);
 757 int kvm_iommu_unmap_guest(struct kvm *kvm);
 758 int kvm_assign_device(struct kvm *kvm,
 759                       struct kvm_assigned_dev_kernel *assigned_dev);
 760 int kvm_deassign_device(struct kvm *kvm,
 761                         struct kvm_assigned_dev_kernel *assigned_dev);
 762 #else
 763 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 764                                       struct kvm_memory_slot *slot)
 765 {
 766         return 0;
 767 }
 768
 769 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 770                                          struct kvm_memory_slot *slot)
 771 {
 772 }
 773
 774 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 775 {
 776         return 0;
 777 }
 778 #endif
 779
 780 static inline void kvm_guest_enter(void)
 781 {
 782         unsigned long flags;
 783
 784         BUG_ON(preemptible());
 785
 786         local_irq_save(flags);
 787         guest_enter();
 788         local_irq_restore(flags);
 789
 790         /* KVM does not hold any references to rcu protected data when it
 791          * switches CPU into a guest mode. In fact switching to a guest mode
 792          * is very similar to exiting to userspase from rcu point of view. In
 793          * addition CPU may stay in a guest mode for quite a long time (up to
 794          * one time slice). Lets treat guest mode as quiescent state, just like
 795          * we do with user-mode execution.
 796          */
 797         rcu_virt_note_context_switch(smp_processor_id());
 798 }
 799
 800 static inline void kvm_guest_exit(void)
 801 {
 802         unsigned long flags;
 803
 804         local_irq_save(flags);
 805         guest_exit();
 806         local_irq_restore(flags);
 807 }
 808
 809 /*
 810  * search_memslots() and __gfn_to_memslot() are here because they are
 811  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 812  * gfn_to_memslot() itself isn't here as an inline because that would
 813  * bloat other code too much.
 814  */
 815 static inline struct kvm_memory_slot *
 816 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 817 {
 818         struct kvm_memory_slot *memslot;
 819
 820         kvm_for_each_memslot(memslot, slots)
 821                 if (gfn >= memslot->base_gfn &&
 822                       gfn < memslot->base_gfn + memslot->npages)
 823                         return memslot;
 824
 825         return NULL;
 826 }
 827
 828 static inline struct kvm_memory_slot *
 829 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 830 {
 831         return search_memslots(slots, gfn);
 832 }
 833
 834 static inline unsigned long
 835 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 836 {
 837         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 838 }
 839
 840 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 841 {
 842         return gfn_to_memslot(kvm, gfn)->id;
 843 }
 844
 845 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 846 {
 847         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 848         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 849                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 850 }
 851
 852 static inline gfn_t
 853 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 854 {
 855         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 856
 857         return slot->base_gfn + gfn_offset;
 858 }
 859
 860 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 861 {
 862         return (gpa_t)gfn << PAGE_SHIFT;
 863 }
 864
 865 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 866 {
 867         return (gfn_t)(gpa >> PAGE_SHIFT);
 868 }
 869
 870 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 871 {
 872         return (hpa_t)pfn << PAGE_SHIFT;
 873 }
 874
 875 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 876 {
 877         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 878 }
 879
 880 enum kvm_stat_kind {
 881         KVM_STAT_VM,
 882         KVM_STAT_VCPU,
 883 };
 884
 885 struct kvm_stats_debugfs_item {
 886         const char *name;
 887         int offset;
 888         enum kvm_stat_kind kind;
 889         struct dentry *dentry;
 890 };
 891 extern struct kvm_stats_debugfs_item debugfs_entries[];
 892 extern struct dentry *kvm_debugfs_dir;
 893
 894 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 895 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 896 {
 897         if (unlikely(kvm->mmu_notifier_count))
 898                 return 1;
 899         /*
 900          * Ensure the read of mmu_notifier_count happens before the read
 901          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 902          * mmu_notifier_invalidate_range_end to make sure that the caller
 903          * either sees the old (non-zero) value of mmu_notifier_count or
 904          * the new (incremented) value of mmu_notifier_seq.
 905          * PowerPC Book3s HV KVM calls this under a per-page lock
 906          * rather than under kvm->mmu_lock, for scalability, so
 907          * can't rely on kvm->mmu_lock to keep things ordered.
 908          */
 909         smp_rmb();
 910         if (kvm->mmu_notifier_seq != mmu_seq)
 911                 return 1;
 912         return 0;
 913 }
 914 #endif
 915
 916 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 917
 918 #define KVM_MAX_IRQ_ROUTES 1024
 919
 920 int kvm_setup_default_irq_routing(struct kvm *kvm);
 921 int kvm_set_irq_routing(struct kvm *kvm,
 922                         const struct kvm_irq_routing_entry *entries,
 923                         unsigned nr,
 924                         unsigned flags);
 925 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
 926                           struct kvm_kernel_irq_routing_entry *e,
 927                           const struct kvm_irq_routing_entry *ue);
 928 void kvm_free_irq_routing(struct kvm *kvm);
 929
 930 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 931
 932 #else
 933
 934 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 935
 936 #endif
 937
 938 #ifdef CONFIG_HAVE_KVM_EVENTFD
 939
 940 void kvm_eventfd_init(struct kvm *kvm);
 941 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 942
 943 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 944 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 945 void kvm_irqfd_release(struct kvm *kvm);
 946 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 947 #else
 948 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 949 {
 950         return -EINVAL;
 951 }
 952
 953 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 954 #endif
 955
 956 #else
 957
 958 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 959
 960 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 961 {
 962         return -EINVAL;
 963 }
 964
 965 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 966
 967 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 968 static inline void kvm_irq_routing_update(struct kvm *kvm,
 969                                           struct kvm_irq_routing_table *irq_rt)
 970 {
 971         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 972 }
 973 #endif
 974
 975 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 976 {
 977         return -ENOSYS;
 978 }
 979
 980 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 981
 982 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 983 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 984 {
 985         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 986 }
 987
 988 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 989
 990 #else
 991
 992 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 993
 994 #endif
 995
 996 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 997
 998 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 999                                   unsigned long arg);
1000
1001 void kvm_free_all_assigned_devices(struct kvm *kvm);
1002
1003 #else
1004
1005 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1006                                                 unsigned long arg)
1007 {
1008         return -ENOTTY;
1009 }
1010
1011 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1012
1013 #endif
1014
1015 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1016 {
1017         set_bit(req, &vcpu->requests);
1018 }
1019
1020 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1021 {
1022         if (test_bit(req, &vcpu->requests)) {
1023                 clear_bit(req, &vcpu->requests);
1024                 return true;
1025         } else {
1026                 return false;
1027         }
1028 }
1029
1030 extern bool kvm_rebooting;
1031
1032 struct kvm_device_ops;
1033
1034 struct kvm_device {
1035         struct kvm_device_ops *ops;
1036         struct kvm *kvm;
1037         void *private;
1038         struct list_head vm_node;
1039 };
1040
1041 /* create, destroy, and name are mandatory */
1042 struct kvm_device_ops {
1043         const char *name;
1044         int (*create)(struct kvm_device *dev, u32 type);
1045
1046         /*
1047          * Destroy is responsible for freeing dev.
1048          *
1049          * Destroy may be called before or after destructors are called
1050          * on emulated I/O regions, depending on whether a reference is
1051          * held by a vcpu or other kvm component that gets destroyed
1052          * after the emulated I/O.
1053          */
1054         void (*destroy)(struct kvm_device *dev);
1055
1056         int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1057         int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1058         int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1059         long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1060                       unsigned long arg);
1061 };
1062
1063 void kvm_device_get(struct kvm_device *dev);
1064 void kvm_device_put(struct kvm_device *dev);
1065 struct kvm_device *kvm_device_from_filp(struct file *filp);
1066
1067 extern struct kvm_device_ops kvm_mpic_ops;
1068 extern struct kvm_device_ops kvm_xics_ops;
1069
1070 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1071
1072 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1073 {
1074         vcpu->spin_loop.in_spin_loop = val;
1075 }
1076 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1077 {
1078         vcpu->spin_loop.dy_eligible = val;
1079 }
1080
1081 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1082
1083 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1084 {
1085 }
1086
1087 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1088 {
1089 }
1090
1091 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1092 {
1093         return true;
1094 }
1095
1096 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1097 #endif
1098