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 #define KVM_REQ_ENABLE_IBS        23
 138 #define KVM_REQ_DISABLE_IBS       24
 139 #define KVM_REQ_APIC_PAGE_RELOAD  25
 140
 141 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 142 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 143
 144 extern struct kmem_cache *kvm_vcpu_cache;
 145
 146 extern spinlock_t kvm_lock;
 147 extern struct list_head vm_list;
 148
 149 struct kvm_io_range {
 150         gpa_t addr;
 151         int len;
 152         struct kvm_io_device *dev;
 153 };
 154
 155 #define NR_IOBUS_DEVS 1000
 156
 157 struct kvm_io_bus {
 158         int dev_count;
 159         int ioeventfd_count;
 160         struct kvm_io_range range[];
 161 };
 162
 163 enum kvm_bus {
 164         KVM_MMIO_BUS,
 165         KVM_PIO_BUS,
 166         KVM_VIRTIO_CCW_NOTIFY_BUS,
 167         KVM_FAST_MMIO_BUS,
 168         KVM_NR_BUSES
 169 };
 170
 171 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 172                      int len, const void *val);
 173 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 174                             int len, const void *val, long cookie);
 175 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 176                     void *val);
 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         bool   wakeup_all;
 193 };
 194
 195 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 196 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 197 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
 198                        struct kvm_arch_async_pf *arch);
 199 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 200 #endif
 201
 202 /*
 203  * Carry out a gup that requires IO. Allow the mm to relinquish the mmap
 204  * semaphore if the filemap/swap has to wait on a page lock. pagep == NULL
 205  * controls whether we retry the gup one more time to completion in that case.
 206  * Typically this is called after a FAULT_FLAG_RETRY_NOWAIT in the main tdp
 207  * handler.
 208  */
 209 int kvm_get_user_page_io(struct task_struct *tsk, struct mm_struct *mm,
 210                          unsigned long addr, bool write_fault,
 211                          struct page **pagep);
 212
 213 enum {
 214         OUTSIDE_GUEST_MODE,
 215         IN_GUEST_MODE,
 216         EXITING_GUEST_MODE,
 217         READING_SHADOW_PAGE_TABLES,
 218 };
 219
 220 /*
 221  * Sometimes a large or cross-page mmio needs to be broken up into separate
 222  * exits for userspace servicing.
 223  */
 224 struct kvm_mmio_fragment {
 225         gpa_t gpa;
 226         void *data;
 227         unsigned len;
 228 };
 229
 230 struct kvm_vcpu {
 231         struct kvm *kvm;
 232 #ifdef CONFIG_PREEMPT_NOTIFIERS
 233         struct preempt_notifier preempt_notifier;
 234 #endif
 235         int cpu;
 236         int vcpu_id;
 237         int srcu_idx;
 238         int mode;
 239         unsigned long requests;
 240         unsigned long guest_debug;
 241
 242         struct mutex mutex;
 243         struct kvm_run *run;
 244
 245         int fpu_active;
 246         int guest_fpu_loaded, guest_xcr0_loaded;
 247         wait_queue_head_t wq;
 248         struct pid *pid;
 249         int sigset_active;
 250         sigset_t sigset;
 251         struct kvm_vcpu_stat stat;
 252
 253 #ifdef CONFIG_HAS_IOMEM
 254         int mmio_needed;
 255         int mmio_read_completed;
 256         int mmio_is_write;
 257         int mmio_cur_fragment;
 258         int mmio_nr_fragments;
 259         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 260 #endif
 261
 262 #ifdef CONFIG_KVM_ASYNC_PF
 263         struct {
 264                 u32 queued;
 265                 struct list_head queue;
 266                 struct list_head done;
 267                 spinlock_t lock;
 268         } async_pf;
 269 #endif
 270
 271 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 272         /*
 273          * Cpu relax intercept or pause loop exit optimization
 274          * in_spin_loop: set when a vcpu does a pause loop exit
 275          *  or cpu relax intercepted.
 276          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 277          */
 278         struct {
 279                 bool in_spin_loop;
 280                 bool dy_eligible;
 281         } spin_loop;
 282 #endif
 283         bool preempted;
 284         struct kvm_vcpu_arch arch;
 285 };
 286
 287 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 288 {
 289         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 290 }
 291
 292 /*
 293  * Some of the bitops functions do not support too long bitmaps.
 294  * This number must be determined not to exceed such limits.
 295  */
 296 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 297
 298 struct kvm_memory_slot {
 299         gfn_t base_gfn;
 300         unsigned long npages;
 301         unsigned long *dirty_bitmap;
 302         struct kvm_arch_memory_slot arch;
 303         unsigned long userspace_addr;
 304         u32 flags;
 305         short id;
 306 };
 307
 308 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 309 {
 310         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 311 }
 312
 313 struct kvm_s390_adapter_int {
 314         u64 ind_addr;
 315         u64 summary_addr;
 316         u64 ind_offset;
 317         u32 summary_offset;
 318         u32 adapter_id;
 319 };
 320
 321 struct kvm_kernel_irq_routing_entry {
 322         u32 gsi;
 323         u32 type;
 324         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 325                    struct kvm *kvm, int irq_source_id, int level,
 326                    bool line_status);
 327         union {
 328                 struct {
 329                         unsigned irqchip;
 330                         unsigned pin;
 331                 } irqchip;
 332                 struct msi_msg msi;
 333                 struct kvm_s390_adapter_int adapter;
 334         };
 335         struct hlist_node link;
 336 };
 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         struct srcu_struct irq_srcu;
 365 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 366         u32 bsp_vcpu_id;
 367 #endif
 368         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 369         atomic_t online_vcpus;
 370         int last_boosted_vcpu;
 371         struct list_head vm_list;
 372         struct mutex lock;
 373         struct kvm_io_bus *buses[KVM_NR_BUSES];
 374 #ifdef CONFIG_HAVE_KVM_EVENTFD
 375         struct {
 376                 spinlock_t        lock;
 377                 struct list_head  items;
 378                 struct list_head  resampler_list;
 379                 struct mutex      resampler_lock;
 380         } irqfds;
 381         struct list_head ioeventfds;
 382 #endif
 383         struct kvm_vm_stat stat;
 384         struct kvm_arch arch;
 385         atomic_t users_count;
 386 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 387         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 388         spinlock_t ring_lock;
 389         struct list_head coalesced_zones;
 390 #endif
 391
 392         struct mutex irq_lock;
 393 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 394         /*
 395          * Update side is protected by irq_lock.
 396          */
 397         struct kvm_irq_routing_table __rcu *irq_routing;
 398         struct hlist_head mask_notifier_list;
 399 #endif
 400 #ifdef CONFIG_HAVE_KVM_IRQFD
 401         struct hlist_head irq_ack_notifier_list;
 402 #endif
 403
 404 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 405         struct mmu_notifier mmu_notifier;
 406         unsigned long mmu_notifier_seq;
 407         long mmu_notifier_count;
 408 #endif
 409         long tlbs_dirty;
 410         struct list_head devices;
 411 };
 412
 413 #define kvm_err(fmt, ...) \
 414         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 415 #define kvm_info(fmt, ...) \
 416         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 417 #define kvm_debug(fmt, ...) \
 418         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 419 #define kvm_pr_unimpl(fmt, ...) \
 420         pr_err_ratelimited("kvm [%i]: " fmt, \
 421                            task_tgid_nr(current), ## __VA_ARGS__)
 422
 423 /* The guest did something we don't support. */
 424 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 425         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 426
 427 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 428 {
 429         smp_rmb();
 430         return kvm->vcpus[i];
 431 }
 432
 433 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 434         for (idx = 0; \
 435              idx < atomic_read(&kvm->online_vcpus) && \
 436              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 437              idx++)
 438
 439 #define kvm_for_each_memslot(memslot, slots)    \
 440         for (memslot = &slots->memslots[0];     \
 441               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 442                 memslot++)
 443
 444 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 445 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 446
 447 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 448 void vcpu_put(struct kvm_vcpu *vcpu);
 449
 450 #ifdef CONFIG_HAVE_KVM_IRQFD
 451 int kvm_irqfd_init(void);
 452 void kvm_irqfd_exit(void);
 453 #else
 454 static inline int kvm_irqfd_init(void)
 455 {
 456         return 0;
 457 }
 458
 459 static inline void kvm_irqfd_exit(void)
 460 {
 461 }
 462 #endif
 463 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 464                   struct module *module);
 465 void kvm_exit(void);
 466
 467 void kvm_get_kvm(struct kvm *kvm);
 468 void kvm_put_kvm(struct kvm *kvm);
 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 *kvm, struct kvm_memory_slot *free,
 512                            struct kvm_memory_slot *dont);
 513 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
 514                             unsigned long npages);
 515 void kvm_arch_memslots_updated(struct kvm *kvm);
 516 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 517                                 struct kvm_memory_slot *memslot,
 518                                 struct kvm_userspace_memory_region *mem,
 519                                 enum kvm_mr_change change);
 520 void kvm_arch_commit_memory_region(struct kvm *kvm,
 521                                 struct kvm_userspace_memory_region *mem,
 522                                 const struct kvm_memory_slot *old,
 523                                 enum kvm_mr_change change);
 524 bool kvm_largepages_enabled(void);
 525 void kvm_disable_largepages(void);
 526 /* flush all memory translations */
 527 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 528 /* flush memory translations pointing to 'slot' */
 529 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 530                                    struct kvm_memory_slot *slot);
 531
 532 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 533                             int nr_pages);
 534
 535 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 536 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 537 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
 538 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 539 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
 540                                       bool *writable);
 541 void kvm_release_page_clean(struct page *page);
 542 void kvm_release_page_dirty(struct page *page);
 543 void kvm_set_page_accessed(struct page *page);
 544
 545 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 546 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 547                        bool write_fault, bool *writable);
 548 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 549 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 550                       bool *writable);
 551 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 552 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 553
 554 void kvm_release_pfn_clean(pfn_t pfn);
 555 void kvm_set_pfn_dirty(pfn_t pfn);
 556 void kvm_set_pfn_accessed(pfn_t pfn);
 557 void kvm_get_pfn(pfn_t pfn);
 558
 559 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 560                         int len);
 561 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 562                           unsigned long len);
 563 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 564 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 565                            void *data, unsigned long len);
 566 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 567                          int offset, int len);
 568 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 569                     unsigned long len);
 570 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 571                            void *data, unsigned long len);
 572 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 573                               gpa_t gpa, unsigned long len);
 574 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 575 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 576 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 577 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 578 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 579 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 580
 581 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 582 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 583 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
 584 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 585 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 586 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 587
 588 void kvm_flush_remote_tlbs(struct kvm *kvm);
 589 void kvm_reload_remote_mmus(struct kvm *kvm);
 590 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 591 void kvm_make_scan_ioapic_request(struct kvm *kvm);
 592 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
 593
 594 long kvm_arch_dev_ioctl(struct file *filp,
 595                         unsigned int ioctl, unsigned long arg);
 596 long kvm_arch_vcpu_ioctl(struct file *filp,
 597                          unsigned int ioctl, unsigned long arg);
 598 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 599
 600 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
 601
 602 int kvm_get_dirty_log(struct kvm *kvm,
 603                         struct kvm_dirty_log *log, int *is_dirty);
 604 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 605                                 struct kvm_dirty_log *log);
 606
 607 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 608                         bool line_status);
 609 long kvm_arch_vm_ioctl(struct file *filp,
 610                        unsigned int ioctl, unsigned long arg);
 611
 612 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 613 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 614
 615 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 616                                     struct kvm_translation *tr);
 617
 618 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 619 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 620 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 621                                   struct kvm_sregs *sregs);
 622 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 623                                   struct kvm_sregs *sregs);
 624 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 625                                     struct kvm_mp_state *mp_state);
 626 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 627                                     struct kvm_mp_state *mp_state);
 628 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 629                                         struct kvm_guest_debug *dbg);
 630 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 631
 632 int kvm_arch_init(void *opaque);
 633 void kvm_arch_exit(void);
 634
 635 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 636 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 637
 638 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
 639
 640 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 641 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 642 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 643 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 644 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 645 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 646 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 647
 648 int kvm_arch_hardware_enable(void);
 649 void kvm_arch_hardware_disable(void);
 650 int kvm_arch_hardware_setup(void);
 651 void kvm_arch_hardware_unsetup(void);
 652 void kvm_arch_check_processor_compat(void *rtn);
 653 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 654 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 655
 656 void *kvm_kvzalloc(unsigned long size);
 657 void kvm_kvfree(const void *addr);
 658
 659 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 660 static inline struct kvm *kvm_arch_alloc_vm(void)
 661 {
 662         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 663 }
 664
 665 static inline void kvm_arch_free_vm(struct kvm *kvm)
 666 {
 667         kfree(kvm);
 668 }
 669 #endif
 670
 671 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
 672 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
 673 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
 674 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
 675 #else
 676 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
 677 {
 678 }
 679
 680 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
 681 {
 682 }
 683
 684 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
 685 {
 686         return false;
 687 }
 688 #endif
 689
 690 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 691 {
 692 #ifdef __KVM_HAVE_ARCH_WQP
 693         return vcpu->arch.wqp;
 694 #else
 695         return &vcpu->wq;
 696 #endif
 697 }
 698
 699 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 700 void kvm_arch_destroy_vm(struct kvm *kvm);
 701 void kvm_arch_sync_events(struct kvm *kvm);
 702
 703 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 704 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 705
 706 bool kvm_is_mmio_pfn(pfn_t pfn);
 707
 708 struct kvm_irq_ack_notifier {
 709         struct hlist_node link;
 710         unsigned gsi;
 711         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 712 };
 713
 714 struct kvm_assigned_dev_kernel {
 715         struct kvm_irq_ack_notifier ack_notifier;
 716         struct list_head list;
 717         int assigned_dev_id;
 718         int host_segnr;
 719         int host_busnr;
 720         int host_devfn;
 721         unsigned int entries_nr;
 722         int host_irq;
 723         bool host_irq_disabled;
 724         bool pci_2_3;
 725         struct msix_entry *host_msix_entries;
 726         int guest_irq;
 727         struct msix_entry *guest_msix_entries;
 728         unsigned long irq_requested_type;
 729         int irq_source_id;
 730         int flags;
 731         struct pci_dev *dev;
 732         struct kvm *kvm;
 733         spinlock_t intx_lock;
 734         spinlock_t intx_mask_lock;
 735         char irq_name[32];
 736         struct pci_saved_state *pci_saved_state;
 737 };
 738
 739 struct kvm_irq_mask_notifier {
 740         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 741         int irq;
 742         struct hlist_node link;
 743 };
 744
 745 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 746                                     struct kvm_irq_mask_notifier *kimn);
 747 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 748                                       struct kvm_irq_mask_notifier *kimn);
 749 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 750                              bool mask);
 751
 752 int kvm_irq_map_gsi(struct kvm *kvm,
 753                     struct kvm_kernel_irq_routing_entry *entries, int gsi);
 754 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
 755
 756 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 757                 bool line_status);
 758 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 759 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 760                 int irq_source_id, int level, bool line_status);
 761 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 762 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 763 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 764                                    struct kvm_irq_ack_notifier *kian);
 765 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 766                                    struct kvm_irq_ack_notifier *kian);
 767 int kvm_request_irq_source_id(struct kvm *kvm);
 768 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 769
 770 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 771 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 772 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 773 int kvm_iommu_map_guest(struct kvm *kvm);
 774 int kvm_iommu_unmap_guest(struct kvm *kvm);
 775 int kvm_assign_device(struct kvm *kvm,
 776                       struct kvm_assigned_dev_kernel *assigned_dev);
 777 int kvm_deassign_device(struct kvm *kvm,
 778                         struct kvm_assigned_dev_kernel *assigned_dev);
 779 #else
 780 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 781                                       struct kvm_memory_slot *slot)
 782 {
 783         return 0;
 784 }
 785
 786 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 787                                          struct kvm_memory_slot *slot)
 788 {
 789 }
 790
 791 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 792 {
 793         return 0;
 794 }
 795 #endif
 796
 797 static inline void kvm_guest_enter(void)
 798 {
 799         unsigned long flags;
 800
 801         BUG_ON(preemptible());
 802
 803         local_irq_save(flags);
 804         guest_enter();
 805         local_irq_restore(flags);
 806
 807         /* KVM does not hold any references to rcu protected data when it
 808          * switches CPU into a guest mode. In fact switching to a guest mode
 809          * is very similar to exiting to userspace from rcu point of view. In
 810          * addition CPU may stay in a guest mode for quite a long time (up to
 811          * one time slice). Lets treat guest mode as quiescent state, just like
 812          * we do with user-mode execution.
 813          */
 814         rcu_virt_note_context_switch(smp_processor_id());
 815 }
 816
 817 static inline void kvm_guest_exit(void)
 818 {
 819         unsigned long flags;
 820
 821         local_irq_save(flags);
 822         guest_exit();
 823         local_irq_restore(flags);
 824 }
 825
 826 /*
 827  * search_memslots() and __gfn_to_memslot() are here because they are
 828  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 829  * gfn_to_memslot() itself isn't here as an inline because that would
 830  * bloat other code too much.
 831  */
 832 static inline struct kvm_memory_slot *
 833 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 834 {
 835         struct kvm_memory_slot *memslot;
 836
 837         kvm_for_each_memslot(memslot, slots)
 838                 if (gfn >= memslot->base_gfn &&
 839                       gfn < memslot->base_gfn + memslot->npages)
 840                         return memslot;
 841
 842         return NULL;
 843 }
 844
 845 static inline struct kvm_memory_slot *
 846 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 847 {
 848         return search_memslots(slots, gfn);
 849 }
 850
 851 static inline unsigned long
 852 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 853 {
 854         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 855 }
 856
 857 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 858 {
 859         return gfn_to_memslot(kvm, gfn)->id;
 860 }
 861
 862 static inline gfn_t
 863 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 864 {
 865         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 866
 867         return slot->base_gfn + gfn_offset;
 868 }
 869
 870 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 871 {
 872         return (gpa_t)gfn << PAGE_SHIFT;
 873 }
 874
 875 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 876 {
 877         return (gfn_t)(gpa >> PAGE_SHIFT);
 878 }
 879
 880 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 881 {
 882         return (hpa_t)pfn << PAGE_SHIFT;
 883 }
 884
 885 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
 886 {
 887         unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
 888
 889         return kvm_is_error_hva(hva);
 890 }
 891
 892 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 893 {
 894         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 895 }
 896
 897 enum kvm_stat_kind {
 898         KVM_STAT_VM,
 899         KVM_STAT_VCPU,
 900 };
 901
 902 struct kvm_stats_debugfs_item {
 903         const char *name;
 904         int offset;
 905         enum kvm_stat_kind kind;
 906         struct dentry *dentry;
 907 };
 908 extern struct kvm_stats_debugfs_item debugfs_entries[];
 909 extern struct dentry *kvm_debugfs_dir;
 910
 911 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 912 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 913 {
 914         if (unlikely(kvm->mmu_notifier_count))
 915                 return 1;
 916         /*
 917          * Ensure the read of mmu_notifier_count happens before the read
 918          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 919          * mmu_notifier_invalidate_range_end to make sure that the caller
 920          * either sees the old (non-zero) value of mmu_notifier_count or
 921          * the new (incremented) value of mmu_notifier_seq.
 922          * PowerPC Book3s HV KVM calls this under a per-page lock
 923          * rather than under kvm->mmu_lock, for scalability, so
 924          * can't rely on kvm->mmu_lock to keep things ordered.
 925          */
 926         smp_rmb();
 927         if (kvm->mmu_notifier_seq != mmu_seq)
 928                 return 1;
 929         return 0;
 930 }
 931 #endif
 932
 933 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 934
 935 #ifdef CONFIG_S390
 936 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
 937 #else
 938 #define KVM_MAX_IRQ_ROUTES 1024
 939 #endif
 940
 941 int kvm_setup_default_irq_routing(struct kvm *kvm);
 942 int kvm_set_irq_routing(struct kvm *kvm,
 943                         const struct kvm_irq_routing_entry *entries,
 944                         unsigned nr,
 945                         unsigned flags);
 946 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
 947                           const struct kvm_irq_routing_entry *ue);
 948 void kvm_free_irq_routing(struct kvm *kvm);
 949
 950 #else
 951
 952 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 953
 954 #endif
 955
 956 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 957
 958 #ifdef CONFIG_HAVE_KVM_EVENTFD
 959
 960 void kvm_eventfd_init(struct kvm *kvm);
 961 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 962
 963 #ifdef CONFIG_HAVE_KVM_IRQFD
 964 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 965 void kvm_irqfd_release(struct kvm *kvm);
 966 void kvm_irq_routing_update(struct kvm *);
 967 #else
 968 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 969 {
 970         return -EINVAL;
 971 }
 972
 973 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 974 #endif
 975
 976 #else
 977
 978 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 979
 980 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 981 {
 982         return -EINVAL;
 983 }
 984
 985 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 986
 987 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 988 static inline void kvm_irq_routing_update(struct kvm *kvm)
 989 {
 990 }
 991 #endif
 992
 993 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 994 {
 995         return -ENOSYS;
 996 }
 997
 998 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 999
1000 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1001 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
1002 {
1003         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
1004 }
1005
1006 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1007
1008 #else
1009
1010 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1011
1012 #endif
1013
1014 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1015
1016 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1017                                   unsigned long arg);
1018
1019 void kvm_free_all_assigned_devices(struct kvm *kvm);
1020
1021 #else
1022
1023 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1024                                                 unsigned long arg)
1025 {
1026         return -ENOTTY;
1027 }
1028
1029 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1030
1031 #endif
1032
1033 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1034 {
1035         set_bit(req, &vcpu->requests);
1036 }
1037
1038 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1039 {
1040         if (test_bit(req, &vcpu->requests)) {
1041                 clear_bit(req, &vcpu->requests);
1042                 return true;
1043         } else {
1044                 return false;
1045         }
1046 }
1047
1048 extern bool kvm_rebooting;
1049
1050 struct kvm_device {
1051         struct kvm_device_ops *ops;
1052         struct kvm *kvm;
1053         void *private;
1054         struct list_head vm_node;
1055 };
1056
1057 /* create, destroy, and name are mandatory */
1058 struct kvm_device_ops {
1059         const char *name;
1060         int (*create)(struct kvm_device *dev, u32 type);
1061
1062         /*
1063          * Destroy is responsible for freeing dev.
1064          *
1065          * Destroy may be called before or after destructors are called
1066          * on emulated I/O regions, depending on whether a reference is
1067          * held by a vcpu or other kvm component that gets destroyed
1068          * after the emulated I/O.
1069          */
1070         void (*destroy)(struct kvm_device *dev);
1071
1072         int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1073         int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1074         int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1075         long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1076                       unsigned long arg);
1077 };
1078
1079 void kvm_device_get(struct kvm_device *dev);
1080 void kvm_device_put(struct kvm_device *dev);
1081 struct kvm_device *kvm_device_from_filp(struct file *filp);
1082 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1083 void kvm_unregister_device_ops(u32 type);
1084
1085 extern struct kvm_device_ops kvm_mpic_ops;
1086 extern struct kvm_device_ops kvm_xics_ops;
1087
1088 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1089
1090 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1091 {
1092         vcpu->spin_loop.in_spin_loop = val;
1093 }
1094 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1095 {
1096         vcpu->spin_loop.dy_eligible = val;
1097 }
1098
1099 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1100
1101 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1102 {
1103 }
1104
1105 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1106 {
1107 }
1108 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1109 #endif
1110