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 <asm/signal.h>
  27
  28 #include <linux/kvm.h>
  29 #include <linux/kvm_para.h>
  30
  31 #include <linux/kvm_types.h>
  32
  33 #include <asm/kvm_host.h>
  34
  35 #ifndef KVM_MMIO_SIZE
  36 #define KVM_MMIO_SIZE 8
  37 #endif
  38
  39 /*
  40  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  41  * in kvm, other bits are visible for userspace which are defined in
  42  * include/linux/kvm_h.
  43  */
  44 #define KVM_MEMSLOT_INVALID     (1UL << 16)
  45
  46 /* Two fragments for cross MMIO pages. */
  47 #define KVM_MAX_MMIO_FRAGMENTS  2
  48
  49 /*
  50  * For the normal pfn, the highest 12 bits should be zero,
  51  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  52  * mask bit 63 to indicate the noslot pfn.
  53  */
  54 #define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  55 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  56 #define KVM_PFN_NOSLOT          (0x1ULL << 63)
  57
  58 #define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  59 #define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  60 #define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  61
  62 /*
  63  * error pfns indicate that the gfn is in slot but faild to
  64  * translate it to pfn on host.
  65  */
  66 static inline bool is_error_pfn(pfn_t pfn)
  67 {
  68         return !!(pfn & KVM_PFN_ERR_MASK);
  69 }
  70
  71 /*
  72  * error_noslot pfns indicate that the gfn can not be
  73  * translated to pfn - it is not in slot or failed to
  74  * translate it to pfn.
  75  */
  76 static inline bool is_error_noslot_pfn(pfn_t pfn)
  77 {
  78         return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
  79 }
  80
  81 /* noslot pfn indicates that the gfn is not in slot. */
  82 static inline bool is_noslot_pfn(pfn_t pfn)
  83 {
  84         return pfn == KVM_PFN_NOSLOT;
  85 }
  86
  87 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  88 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  89
  90 static inline bool kvm_is_error_hva(unsigned long addr)
  91 {
  92         return addr >= PAGE_OFFSET;
  93 }
  94
  95 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
  96
  97 static inline bool is_error_page(struct page *page)
  98 {
  99         return IS_ERR(page);
 100 }
 101
 102 /*
 103  * vcpu->requests bit members
 104  */
 105 #define KVM_REQ_TLB_FLUSH          0
 106 #define KVM_REQ_MIGRATE_TIMER      1
 107 #define KVM_REQ_REPORT_TPR_ACCESS  2
 108 #define KVM_REQ_MMU_RELOAD         3
 109 #define KVM_REQ_TRIPLE_FAULT       4
 110 #define KVM_REQ_PENDING_TIMER      5
 111 #define KVM_REQ_UNHALT             6
 112 #define KVM_REQ_MMU_SYNC           7
 113 #define KVM_REQ_CLOCK_UPDATE       8
 114 #define KVM_REQ_KICK               9
 115 #define KVM_REQ_DEACTIVATE_FPU    10
 116 #define KVM_REQ_EVENT             11
 117 #define KVM_REQ_APF_HALT          12
 118 #define KVM_REQ_STEAL_UPDATE      13
 119 #define KVM_REQ_NMI               14
 120 #define KVM_REQ_IMMEDIATE_EXIT    15
 121 #define KVM_REQ_PMU               16
 122 #define KVM_REQ_PMI               17
 123 #define KVM_REQ_WATCHDOG          18
 124 #define KVM_REQ_MASTERCLOCK_UPDATE 19
 125 #define KVM_REQ_MCLOCK_INPROGRESS 20
 126 #define KVM_REQ_EPR_EXIT          21
 127 #define KVM_REQ_EOIBITMAP         22
 128
 129 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 131
 132 struct kvm;
 133 struct kvm_vcpu;
 134 extern struct kmem_cache *kvm_vcpu_cache;
 135
 136 struct kvm_io_range {
 137         gpa_t addr;
 138         int len;
 139         struct kvm_io_device *dev;
 140 };
 141
 142 #define NR_IOBUS_DEVS 1000
 143
 144 struct kvm_io_bus {
 145         int                   dev_count;
 146         struct kvm_io_range range[];
 147 };
 148
 149 enum kvm_bus {
 150         KVM_MMIO_BUS,
 151         KVM_PIO_BUS,
 152         KVM_NR_BUSES
 153 };
 154
 155 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 156                      int len, const void *val);
 157 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 158                     void *val);
 159 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 160                             int len, struct kvm_io_device *dev);
 161 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 162                               struct kvm_io_device *dev);
 163
 164 #ifdef CONFIG_KVM_ASYNC_PF
 165 struct kvm_async_pf {
 166         struct work_struct work;
 167         struct list_head link;
 168         struct list_head queue;
 169         struct kvm_vcpu *vcpu;
 170         struct mm_struct *mm;
 171         gva_t gva;
 172         unsigned long addr;
 173         struct kvm_arch_async_pf arch;
 174         struct page *page;
 175         bool done;
 176 };
 177
 178 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 179 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 180 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 181                        struct kvm_arch_async_pf *arch);
 182 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 183 #endif
 184
 185 enum {
 186         OUTSIDE_GUEST_MODE,
 187         IN_GUEST_MODE,
 188         EXITING_GUEST_MODE,
 189         READING_SHADOW_PAGE_TABLES,
 190 };
 191
 192 /*
 193  * Sometimes a large or cross-page mmio needs to be broken up into separate
 194  * exits for userspace servicing.
 195  */
 196 struct kvm_mmio_fragment {
 197         gpa_t gpa;
 198         void *data;
 199         unsigned len;
 200 };
 201
 202 struct kvm_vcpu {
 203         struct kvm *kvm;
 204 #ifdef CONFIG_PREEMPT_NOTIFIERS
 205         struct preempt_notifier preempt_notifier;
 206 #endif
 207         int cpu;
 208         int vcpu_id;
 209         int srcu_idx;
 210         int mode;
 211         unsigned long requests;
 212         unsigned long guest_debug;
 213
 214         struct mutex mutex;
 215         struct kvm_run *run;
 216
 217         int fpu_active;
 218         int guest_fpu_loaded, guest_xcr0_loaded;
 219         wait_queue_head_t wq;
 220         struct pid *pid;
 221         int sigset_active;
 222         sigset_t sigset;
 223         struct kvm_vcpu_stat stat;
 224
 225 #ifdef CONFIG_HAS_IOMEM
 226         int mmio_needed;
 227         int mmio_read_completed;
 228         int mmio_is_write;
 229         int mmio_cur_fragment;
 230         int mmio_nr_fragments;
 231         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 232 #endif
 233
 234 #ifdef CONFIG_KVM_ASYNC_PF
 235         struct {
 236                 u32 queued;
 237                 struct list_head queue;
 238                 struct list_head done;
 239                 spinlock_t lock;
 240         } async_pf;
 241 #endif
 242
 243 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 244         /*
 245          * Cpu relax intercept or pause loop exit optimization
 246          * in_spin_loop: set when a vcpu does a pause loop exit
 247          *  or cpu relax intercepted.
 248          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 249          */
 250         struct {
 251                 bool in_spin_loop;
 252                 bool dy_eligible;
 253         } spin_loop;
 254 #endif
 255         struct kvm_vcpu_arch arch;
 256 };
 257
 258 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 259 {
 260         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 261 }
 262
 263 /*
 264  * Some of the bitops functions do not support too long bitmaps.
 265  * This number must be determined not to exceed such limits.
 266  */
 267 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 268
 269 struct kvm_memory_slot {
 270         gfn_t base_gfn;
 271         unsigned long npages;
 272         unsigned long *dirty_bitmap;
 273         struct kvm_arch_memory_slot arch;
 274         unsigned long userspace_addr;
 275         u32 flags;
 276         short id;
 277 };
 278
 279 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 280 {
 281         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 282 }
 283
 284 struct kvm_kernel_irq_routing_entry {
 285         u32 gsi;
 286         u32 type;
 287         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 288                    struct kvm *kvm, int irq_source_id, int level);
 289         union {
 290                 struct {
 291                         unsigned irqchip;
 292                         unsigned pin;
 293                 } irqchip;
 294                 struct msi_msg msi;
 295         };
 296         struct hlist_node link;
 297 };
 298
 299 #ifdef __KVM_HAVE_IOAPIC
 300
 301 struct kvm_irq_routing_table {
 302         int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
 303         struct kvm_kernel_irq_routing_entry *rt_entries;
 304         u32 nr_rt_entries;
 305         /*
 306          * Array indexed by gsi. Each entry contains list of irq chips
 307          * the gsi is connected to.
 308          */
 309         struct hlist_head map[0];
 310 };
 311
 312 #else
 313
 314 struct kvm_irq_routing_table {};
 315
 316 #endif
 317
 318 #ifndef KVM_PRIVATE_MEM_SLOTS
 319 #define KVM_PRIVATE_MEM_SLOTS 0
 320 #endif
 321
 322 #ifndef KVM_MEM_SLOTS_NUM
 323 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 324 #endif
 325
 326 /*
 327  * Note:
 328  * memslots are not sorted by id anymore, please use id_to_memslot()
 329  * to get the memslot by its id.
 330  */
 331 struct kvm_memslots {
 332         u64 generation;
 333         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 334         /* The mapping table from slot id to the index in memslots[]. */
 335         short id_to_index[KVM_MEM_SLOTS_NUM];
 336 };
 337
 338 struct kvm {
 339         spinlock_t mmu_lock;
 340         struct mutex slots_lock;
 341         struct mm_struct *mm; /* userspace tied to this vm */
 342         struct kvm_memslots *memslots;
 343         struct srcu_struct srcu;
 344 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 345         u32 bsp_vcpu_id;
 346 #endif
 347         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 348         atomic_t online_vcpus;
 349         int last_boosted_vcpu;
 350         struct list_head vm_list;
 351         struct mutex lock;
 352         struct kvm_io_bus *buses[KVM_NR_BUSES];
 353 #ifdef CONFIG_HAVE_KVM_EVENTFD
 354         struct {
 355                 spinlock_t        lock;
 356                 struct list_head  items;
 357                 struct list_head  resampler_list;
 358                 struct mutex      resampler_lock;
 359         } irqfds;
 360         struct list_head ioeventfds;
 361 #endif
 362         struct kvm_vm_stat stat;
 363         struct kvm_arch arch;
 364         atomic_t users_count;
 365 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 366         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 367         spinlock_t ring_lock;
 368         struct list_head coalesced_zones;
 369 #endif
 370
 371         struct mutex irq_lock;
 372 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 373         /*
 374          * Update side is protected by irq_lock and,
 375          * if configured, irqfds.lock.
 376          */
 377         struct kvm_irq_routing_table __rcu *irq_routing;
 378         struct hlist_head mask_notifier_list;
 379         struct hlist_head irq_ack_notifier_list;
 380 #endif
 381
 382 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 383         struct mmu_notifier mmu_notifier;
 384         unsigned long mmu_notifier_seq;
 385         long mmu_notifier_count;
 386 #endif
 387         long tlbs_dirty;
 388 };
 389
 390 #define kvm_err(fmt, ...) \
 391         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 392 #define kvm_info(fmt, ...) \
 393         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 394 #define kvm_debug(fmt, ...) \
 395         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 396 #define kvm_pr_unimpl(fmt, ...) \
 397         pr_err_ratelimited("kvm [%i]: " fmt, \
 398                            task_tgid_nr(current), ## __VA_ARGS__)
 399
 400 /* The guest did something we don't support. */
 401 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 402         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 403
 404 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 405 {
 406         smp_rmb();
 407         return kvm->vcpus[i];
 408 }
 409
 410 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 411         for (idx = 0; \
 412              idx < atomic_read(&kvm->online_vcpus) && \
 413              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 414              idx++)
 415
 416 #define kvm_for_each_memslot(memslot, slots)    \
 417         for (memslot = &slots->memslots[0];     \
 418               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 419                 memslot++)
 420
 421 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 422 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 423
 424 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 425 void vcpu_put(struct kvm_vcpu *vcpu);
 426
 427 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 428                   struct module *module);
 429 void kvm_exit(void);
 430
 431 void kvm_get_kvm(struct kvm *kvm);
 432 void kvm_put_kvm(struct kvm *kvm);
 433 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
 434                      u64 last_generation);
 435
 436 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 437 {
 438         return rcu_dereference_check(kvm->memslots,
 439                         srcu_read_lock_held(&kvm->srcu)
 440                         || lockdep_is_held(&kvm->slots_lock));
 441 }
 442
 443 static inline struct kvm_memory_slot *
 444 id_to_memslot(struct kvm_memslots *slots, int id)
 445 {
 446         int index = slots->id_to_index[id];
 447         struct kvm_memory_slot *slot;
 448
 449         slot = &slots->memslots[index];
 450
 451         WARN_ON(slot->id != id);
 452         return slot;
 453 }
 454
 455 int kvm_set_memory_region(struct kvm *kvm,
 456                           struct kvm_userspace_memory_region *mem,
 457                           bool user_alloc);
 458 int __kvm_set_memory_region(struct kvm *kvm,
 459                             struct kvm_userspace_memory_region *mem,
 460                             bool user_alloc);
 461 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 462                            struct kvm_memory_slot *dont);
 463 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 464 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 465                                 struct kvm_memory_slot *memslot,
 466                                 struct kvm_memory_slot old,
 467                                 struct kvm_userspace_memory_region *mem,
 468                                 bool user_alloc);
 469 void kvm_arch_commit_memory_region(struct kvm *kvm,
 470                                 struct kvm_userspace_memory_region *mem,
 471                                 struct kvm_memory_slot old,
 472                                 bool user_alloc);
 473 bool kvm_largepages_enabled(void);
 474 void kvm_disable_largepages(void);
 475 /* flush all memory translations */
 476 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 477 /* flush memory translations pointing to 'slot' */
 478 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 479                                    struct kvm_memory_slot *slot);
 480
 481 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 482                             int nr_pages);
 483
 484 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 485 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 486 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 487 void kvm_release_page_clean(struct page *page);
 488 void kvm_release_page_dirty(struct page *page);
 489 void kvm_set_page_dirty(struct page *page);
 490 void kvm_set_page_accessed(struct page *page);
 491
 492 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 493 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 494                        bool write_fault, bool *writable);
 495 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 496 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 497                       bool *writable);
 498 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 499 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 500
 501 void kvm_release_pfn_dirty(pfn_t pfn);
 502 void kvm_release_pfn_clean(pfn_t pfn);
 503 void kvm_set_pfn_dirty(pfn_t pfn);
 504 void kvm_set_pfn_accessed(pfn_t pfn);
 505 void kvm_get_pfn(pfn_t pfn);
 506
 507 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 508                         int len);
 509 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 510                           unsigned long len);
 511 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 512 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 513                            void *data, unsigned long len);
 514 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 515                          int offset, int len);
 516 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 517                     unsigned long len);
 518 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 519                            void *data, unsigned long len);
 520 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 521                               gpa_t gpa);
 522 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 523 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 524 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 525 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 526 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 527 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 528 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 529                              gfn_t gfn);
 530
 531 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 532 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 533 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 534 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 535 void kvm_resched(struct kvm_vcpu *vcpu);
 536 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 537 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 538
 539 void kvm_flush_remote_tlbs(struct kvm *kvm);
 540 void kvm_reload_remote_mmus(struct kvm *kvm);
 541 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 542 void kvm_make_update_eoibitmap_request(struct kvm *kvm);
 543
 544 long kvm_arch_dev_ioctl(struct file *filp,
 545                         unsigned int ioctl, unsigned long arg);
 546 long kvm_arch_vcpu_ioctl(struct file *filp,
 547                          unsigned int ioctl, unsigned long arg);
 548 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 549
 550 int kvm_dev_ioctl_check_extension(long ext);
 551
 552 int kvm_get_dirty_log(struct kvm *kvm,
 553                         struct kvm_dirty_log *log, int *is_dirty);
 554 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 555                                 struct kvm_dirty_log *log);
 556
 557 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 558                                    struct
 559                                    kvm_userspace_memory_region *mem,
 560                                    bool user_alloc);
 561 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
 562 long kvm_arch_vm_ioctl(struct file *filp,
 563                        unsigned int ioctl, unsigned long arg);
 564
 565 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 566 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 567
 568 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 569                                     struct kvm_translation *tr);
 570
 571 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 572 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 573 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 574                                   struct kvm_sregs *sregs);
 575 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 576                                   struct kvm_sregs *sregs);
 577 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 578                                     struct kvm_mp_state *mp_state);
 579 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 580                                     struct kvm_mp_state *mp_state);
 581 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 582                                         struct kvm_guest_debug *dbg);
 583 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 584
 585 int kvm_arch_init(void *opaque);
 586 void kvm_arch_exit(void);
 587
 588 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 589 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 590
 591 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 592 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 593 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 594 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 595 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 596 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 597 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 598
 599 int kvm_arch_hardware_enable(void *garbage);
 600 void kvm_arch_hardware_disable(void *garbage);
 601 int kvm_arch_hardware_setup(void);
 602 void kvm_arch_hardware_unsetup(void);
 603 void kvm_arch_check_processor_compat(void *rtn);
 604 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 605 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 606
 607 void kvm_free_physmem(struct kvm *kvm);
 608
 609 void *kvm_kvzalloc(unsigned long size);
 610 void kvm_kvfree(const void *addr);
 611
 612 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 613 static inline struct kvm *kvm_arch_alloc_vm(void)
 614 {
 615         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 616 }
 617
 618 static inline void kvm_arch_free_vm(struct kvm *kvm)
 619 {
 620         kfree(kvm);
 621 }
 622 #endif
 623
 624 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 625 {
 626 #ifdef __KVM_HAVE_ARCH_WQP
 627         return vcpu->arch.wqp;
 628 #else
 629         return &vcpu->wq;
 630 #endif
 631 }
 632
 633 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 634 void kvm_arch_destroy_vm(struct kvm *kvm);
 635 void kvm_free_all_assigned_devices(struct kvm *kvm);
 636 void kvm_arch_sync_events(struct kvm *kvm);
 637
 638 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 639 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 640
 641 bool kvm_is_mmio_pfn(pfn_t pfn);
 642
 643 struct kvm_irq_ack_notifier {
 644         struct hlist_node link;
 645         unsigned gsi;
 646         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 647 };
 648
 649 struct kvm_assigned_dev_kernel {
 650         struct kvm_irq_ack_notifier ack_notifier;
 651         struct list_head list;
 652         int assigned_dev_id;
 653         int host_segnr;
 654         int host_busnr;
 655         int host_devfn;
 656         unsigned int entries_nr;
 657         int host_irq;
 658         bool host_irq_disabled;
 659         bool pci_2_3;
 660         struct msix_entry *host_msix_entries;
 661         int guest_irq;
 662         struct msix_entry *guest_msix_entries;
 663         unsigned long irq_requested_type;
 664         int irq_source_id;
 665         int flags;
 666         struct pci_dev *dev;
 667         struct kvm *kvm;
 668         spinlock_t intx_lock;
 669         spinlock_t intx_mask_lock;
 670         char irq_name[32];
 671         struct pci_saved_state *pci_saved_state;
 672 };
 673
 674 struct kvm_irq_mask_notifier {
 675         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 676         int irq;
 677         struct hlist_node link;
 678 };
 679
 680 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 681                                     struct kvm_irq_mask_notifier *kimn);
 682 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 683                                       struct kvm_irq_mask_notifier *kimn);
 684 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 685                              bool mask);
 686
 687 #ifdef __KVM_HAVE_IOAPIC
 688 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
 689                                    union kvm_ioapic_redirect_entry *entry,
 690                                    unsigned long *deliver_bitmask);
 691 #endif
 692 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 693 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 694 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 695                 int irq_source_id, int level);
 696 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 697 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 698 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 699                                    struct kvm_irq_ack_notifier *kian);
 700 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 701                                    struct kvm_irq_ack_notifier *kian);
 702 int kvm_request_irq_source_id(struct kvm *kvm);
 703 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 704
 705 /* For vcpu->arch.iommu_flags */
 706 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 707
 708 #ifdef CONFIG_IOMMU_API
 709 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 710 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 711 int kvm_iommu_map_guest(struct kvm *kvm);
 712 int kvm_iommu_unmap_guest(struct kvm *kvm);
 713 int kvm_assign_device(struct kvm *kvm,
 714                       struct kvm_assigned_dev_kernel *assigned_dev);
 715 int kvm_deassign_device(struct kvm *kvm,
 716                         struct kvm_assigned_dev_kernel *assigned_dev);
 717 #else /* CONFIG_IOMMU_API */
 718 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 719                                       struct kvm_memory_slot *slot)
 720 {
 721         return 0;
 722 }
 723
 724 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 725                                          struct kvm_memory_slot *slot)
 726 {
 727 }
 728
 729 static inline int kvm_iommu_map_guest(struct kvm *kvm)
 730 {
 731         return -ENODEV;
 732 }
 733
 734 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 735 {
 736         return 0;
 737 }
 738
 739 static inline int kvm_assign_device(struct kvm *kvm,
 740                 struct kvm_assigned_dev_kernel *assigned_dev)
 741 {
 742         return 0;
 743 }
 744
 745 static inline int kvm_deassign_device(struct kvm *kvm,
 746                 struct kvm_assigned_dev_kernel *assigned_dev)
 747 {
 748         return 0;
 749 }
 750 #endif /* CONFIG_IOMMU_API */
 751
 752 static inline void __guest_enter(void)
 753 {
 754         /*
 755          * This is running in ioctl context so we can avoid
 756          * the call to vtime_account() with its unnecessary idle check.
 757          */
 758         vtime_account_system(current);
 759         current->flags |= PF_VCPU;
 760 }
 761
 762 static inline void __guest_exit(void)
 763 {
 764         /*
 765          * This is running in ioctl context so we can avoid
 766          * the call to vtime_account() with its unnecessary idle check.
 767          */
 768         vtime_account_system(current);
 769         current->flags &= ~PF_VCPU;
 770 }
 771
 772 #ifdef CONFIG_CONTEXT_TRACKING
 773 extern void guest_enter(void);
 774 extern void guest_exit(void);
 775
 776 #else /* !CONFIG_CONTEXT_TRACKING */
 777 static inline void guest_enter(void)
 778 {
 779         __guest_enter();
 780 }
 781
 782 static inline void guest_exit(void)
 783 {
 784         __guest_exit();
 785 }
 786 #endif /* !CONFIG_CONTEXT_TRACKING */
 787
 788 static inline void kvm_guest_enter(void)
 789 {
 790         unsigned long flags;
 791
 792         BUG_ON(preemptible());
 793
 794         local_irq_save(flags);
 795         guest_enter();
 796         local_irq_restore(flags);
 797
 798         /* KVM does not hold any references to rcu protected data when it
 799          * switches CPU into a guest mode. In fact switching to a guest mode
 800          * is very similar to exiting to userspase from rcu point of view. In
 801          * addition CPU may stay in a guest mode for quite a long time (up to
 802          * one time slice). Lets treat guest mode as quiescent state, just like
 803          * we do with user-mode execution.
 804          */
 805         rcu_virt_note_context_switch(smp_processor_id());
 806 }
 807
 808 static inline void kvm_guest_exit(void)
 809 {
 810         unsigned long flags;
 811
 812         local_irq_save(flags);
 813         guest_exit();
 814         local_irq_restore(flags);
 815 }
 816
 817 /*
 818  * search_memslots() and __gfn_to_memslot() are here because they are
 819  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 820  * gfn_to_memslot() itself isn't here as an inline because that would
 821  * bloat other code too much.
 822  */
 823 static inline struct kvm_memory_slot *
 824 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 825 {
 826         struct kvm_memory_slot *memslot;
 827
 828         kvm_for_each_memslot(memslot, slots)
 829                 if (gfn >= memslot->base_gfn &&
 830                       gfn < memslot->base_gfn + memslot->npages)
 831                         return memslot;
 832
 833         return NULL;
 834 }
 835
 836 static inline struct kvm_memory_slot *
 837 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 838 {
 839         return search_memslots(slots, gfn);
 840 }
 841
 842 static inline unsigned long
 843 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 844 {
 845         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 846 }
 847
 848 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 849 {
 850         return gfn_to_memslot(kvm, gfn)->id;
 851 }
 852
 853 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 854 {
 855         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 856         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 857                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 858 }
 859
 860 static inline gfn_t
 861 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 862 {
 863         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 864
 865         return slot->base_gfn + gfn_offset;
 866 }
 867
 868 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 869 {
 870         return (gpa_t)gfn << PAGE_SHIFT;
 871 }
 872
 873 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 874 {
 875         return (gfn_t)(gpa >> PAGE_SHIFT);
 876 }
 877
 878 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 879 {
 880         return (hpa_t)pfn << PAGE_SHIFT;
 881 }
 882
 883 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 884 {
 885         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 886 }
 887
 888 enum kvm_stat_kind {
 889         KVM_STAT_VM,
 890         KVM_STAT_VCPU,
 891 };
 892
 893 struct kvm_stats_debugfs_item {
 894         const char *name;
 895         int offset;
 896         enum kvm_stat_kind kind;
 897         struct dentry *dentry;
 898 };
 899 extern struct kvm_stats_debugfs_item debugfs_entries[];
 900 extern struct dentry *kvm_debugfs_dir;
 901
 902 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 903 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 904 {
 905         if (unlikely(kvm->mmu_notifier_count))
 906                 return 1;
 907         /*
 908          * Ensure the read of mmu_notifier_count happens before the read
 909          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 910          * mmu_notifier_invalidate_range_end to make sure that the caller
 911          * either sees the old (non-zero) value of mmu_notifier_count or
 912          * the new (incremented) value of mmu_notifier_seq.
 913          * PowerPC Book3s HV KVM calls this under a per-page lock
 914          * rather than under kvm->mmu_lock, for scalability, so
 915          * can't rely on kvm->mmu_lock to keep things ordered.
 916          */
 917         smp_rmb();
 918         if (kvm->mmu_notifier_seq != mmu_seq)
 919                 return 1;
 920         return 0;
 921 }
 922 #endif
 923
 924 #ifdef KVM_CAP_IRQ_ROUTING
 925
 926 #define KVM_MAX_IRQ_ROUTES 1024
 927
 928 int kvm_setup_default_irq_routing(struct kvm *kvm);
 929 int kvm_set_irq_routing(struct kvm *kvm,
 930                         const struct kvm_irq_routing_entry *entries,
 931                         unsigned nr,
 932                         unsigned flags);
 933 void kvm_free_irq_routing(struct kvm *kvm);
 934
 935 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 936
 937 #else
 938
 939 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 940
 941 #endif
 942
 943 #ifdef CONFIG_HAVE_KVM_EVENTFD
 944
 945 void kvm_eventfd_init(struct kvm *kvm);
 946 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 947
 948 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 949 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 950 void kvm_irqfd_release(struct kvm *kvm);
 951 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 952 #else
 953 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 954 {
 955         return -EINVAL;
 956 }
 957
 958 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 959 #endif
 960
 961 #else
 962
 963 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 964
 965 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 966 {
 967         return -EINVAL;
 968 }
 969
 970 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 971
 972 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 973 static inline void kvm_irq_routing_update(struct kvm *kvm,
 974                                           struct kvm_irq_routing_table *irq_rt)
 975 {
 976         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 977 }
 978 #endif
 979
 980 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 981 {
 982         return -ENOSYS;
 983 }
 984
 985 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 986
 987 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 988 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 989 {
 990         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 991 }
 992
 993 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 994
 995 #else
 996
 997 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 998
 999 #endif
1000
1001 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
1002
1003 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1004                                   unsigned long arg);
1005
1006 #else
1007
1008 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1009                                                 unsigned long arg)
1010 {
1011         return -ENOTTY;
1012 }
1013
1014 #endif
1015
1016 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1017 {
1018         set_bit(req, &vcpu->requests);
1019 }
1020
1021 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1022 {
1023         if (test_bit(req, &vcpu->requests)) {
1024                 clear_bit(req, &vcpu->requests);
1025                 return true;
1026         } else {
1027                 return false;
1028         }
1029 }
1030
1031 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1032
1033 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1034 {
1035         vcpu->spin_loop.in_spin_loop = val;
1036 }
1037 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1038 {
1039         vcpu->spin_loop.dy_eligible = val;
1040 }
1041
1042 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1043
1044 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1045 {
1046 }
1047
1048 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1049 {
1050 }
1051
1052 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1053 {
1054         return true;
1055 }
1056
1057 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1058 #endif
1059