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 <asm/signal.h>
  26
  27 #include <linux/kvm.h>
  28 #include <linux/kvm_para.h>
  29
  30 #include <linux/kvm_types.h>
  31
  32 #include <asm/kvm_host.h>
  33
  34 #ifndef KVM_MMIO_SIZE
  35 #define KVM_MMIO_SIZE 8
  36 #endif
  37
  38 /*
  39  * If we support unaligned MMIO, at most one fragment will be split into two:
  40  */
  41 #ifdef KVM_UNALIGNED_MMIO
  42 #  define KVM_EXTRA_MMIO_FRAGMENTS 1
  43 #else
  44 #  define KVM_EXTRA_MMIO_FRAGMENTS 0
  45 #endif
  46
  47 #define KVM_USER_MMIO_SIZE 8
  48
  49 #define KVM_MAX_MMIO_FRAGMENTS \
  50         (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
  51
  52 /*
  53  * For the normal pfn, the highest 12 bits should be zero,
  54  * so we can mask these bits to indicate the error.
  55  */
  56 #define KVM_PFN_ERR_MASK        (0xfffULL << 52)
  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_BAD         (KVM_PFN_ERR_MASK + 2)
  61
  62 static inline bool is_error_pfn(pfn_t pfn)
  63 {
  64         return !!(pfn & KVM_PFN_ERR_MASK);
  65 }
  66
  67 static inline bool is_noslot_pfn(pfn_t pfn)
  68 {
  69         return pfn == KVM_PFN_ERR_BAD;
  70 }
  71
  72 static inline bool is_invalid_pfn(pfn_t pfn)
  73 {
  74         return !is_noslot_pfn(pfn) && is_error_pfn(pfn);
  75 }
  76
  77 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
  78
  79 static inline bool is_error_page(struct page *page)
  80 {
  81         return IS_ERR(page);
  82 }
  83
  84 /*
  85  * vcpu->requests bit members
  86  */
  87 #define KVM_REQ_TLB_FLUSH          0
  88 #define KVM_REQ_MIGRATE_TIMER      1
  89 #define KVM_REQ_REPORT_TPR_ACCESS  2
  90 #define KVM_REQ_MMU_RELOAD         3
  91 #define KVM_REQ_TRIPLE_FAULT       4
  92 #define KVM_REQ_PENDING_TIMER      5
  93 #define KVM_REQ_UNHALT             6
  94 #define KVM_REQ_MMU_SYNC           7
  95 #define KVM_REQ_CLOCK_UPDATE       8
  96 #define KVM_REQ_KICK               9
  97 #define KVM_REQ_DEACTIVATE_FPU    10
  98 #define KVM_REQ_EVENT             11
  99 #define KVM_REQ_APF_HALT          12
 100 #define KVM_REQ_STEAL_UPDATE      13
 101 #define KVM_REQ_NMI               14
 102 #define KVM_REQ_IMMEDIATE_EXIT    15
 103 #define KVM_REQ_PMU               16
 104 #define KVM_REQ_PMI               17
 105
 106 #define KVM_USERSPACE_IRQ_SOURCE_ID     0
 107
 108 struct kvm;
 109 struct kvm_vcpu;
 110 extern struct kmem_cache *kvm_vcpu_cache;
 111
 112 struct kvm_io_range {
 113         gpa_t addr;
 114         int len;
 115         struct kvm_io_device *dev;
 116 };
 117
 118 #define NR_IOBUS_DEVS 1000
 119
 120 struct kvm_io_bus {
 121         int                   dev_count;
 122         struct kvm_io_range range[];
 123 };
 124
 125 enum kvm_bus {
 126         KVM_MMIO_BUS,
 127         KVM_PIO_BUS,
 128         KVM_NR_BUSES
 129 };
 130
 131 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 132                      int len, const void *val);
 133 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 134                     void *val);
 135 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 136                             int len, struct kvm_io_device *dev);
 137 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 138                               struct kvm_io_device *dev);
 139
 140 #ifdef CONFIG_KVM_ASYNC_PF
 141 struct kvm_async_pf {
 142         struct work_struct work;
 143         struct list_head link;
 144         struct list_head queue;
 145         struct kvm_vcpu *vcpu;
 146         struct mm_struct *mm;
 147         gva_t gva;
 148         unsigned long addr;
 149         struct kvm_arch_async_pf arch;
 150         struct page *page;
 151         bool done;
 152 };
 153
 154 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 155 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 156 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 157                        struct kvm_arch_async_pf *arch);
 158 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 159 #endif
 160
 161 enum {
 162         OUTSIDE_GUEST_MODE,
 163         IN_GUEST_MODE,
 164         EXITING_GUEST_MODE,
 165         READING_SHADOW_PAGE_TABLES,
 166 };
 167
 168 /*
 169  * Sometimes a large or cross-page mmio needs to be broken up into separate
 170  * exits for userspace servicing.
 171  */
 172 struct kvm_mmio_fragment {
 173         gpa_t gpa;
 174         void *data;
 175         unsigned len;
 176 };
 177
 178 struct kvm_vcpu {
 179         struct kvm *kvm;
 180 #ifdef CONFIG_PREEMPT_NOTIFIERS
 181         struct preempt_notifier preempt_notifier;
 182 #endif
 183         int cpu;
 184         int vcpu_id;
 185         int srcu_idx;
 186         int mode;
 187         unsigned long requests;
 188         unsigned long guest_debug;
 189
 190         struct mutex mutex;
 191         struct kvm_run *run;
 192
 193         int fpu_active;
 194         int guest_fpu_loaded, guest_xcr0_loaded;
 195         wait_queue_head_t wq;
 196         struct pid *pid;
 197         int sigset_active;
 198         sigset_t sigset;
 199         struct kvm_vcpu_stat stat;
 200
 201 #ifdef CONFIG_HAS_IOMEM
 202         int mmio_needed;
 203         int mmio_read_completed;
 204         int mmio_is_write;
 205         int mmio_cur_fragment;
 206         int mmio_nr_fragments;
 207         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 208 #endif
 209
 210 #ifdef CONFIG_KVM_ASYNC_PF
 211         struct {
 212                 u32 queued;
 213                 struct list_head queue;
 214                 struct list_head done;
 215                 spinlock_t lock;
 216         } async_pf;
 217 #endif
 218
 219 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 220         /*
 221          * Cpu relax intercept or pause loop exit optimization
 222          * in_spin_loop: set when a vcpu does a pause loop exit
 223          *  or cpu relax intercepted.
 224          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 225          */
 226         struct {
 227                 bool in_spin_loop;
 228                 bool dy_eligible;
 229         } spin_loop;
 230 #endif
 231         struct kvm_vcpu_arch arch;
 232 };
 233
 234 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 235 {
 236         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 237 }
 238
 239 /*
 240  * Some of the bitops functions do not support too long bitmaps.
 241  * This number must be determined not to exceed such limits.
 242  */
 243 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 244
 245 struct kvm_memory_slot {
 246         gfn_t base_gfn;
 247         unsigned long npages;
 248         unsigned long flags;
 249         unsigned long *dirty_bitmap;
 250         struct kvm_arch_memory_slot arch;
 251         unsigned long userspace_addr;
 252         int user_alloc;
 253         int id;
 254 };
 255
 256 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 257 {
 258         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 259 }
 260
 261 struct kvm_kernel_irq_routing_entry {
 262         u32 gsi;
 263         u32 type;
 264         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 265                    struct kvm *kvm, int irq_source_id, int level);
 266         union {
 267                 struct {
 268                         unsigned irqchip;
 269                         unsigned pin;
 270                 } irqchip;
 271                 struct msi_msg msi;
 272         };
 273         struct hlist_node link;
 274 };
 275
 276 #ifdef __KVM_HAVE_IOAPIC
 277
 278 struct kvm_irq_routing_table {
 279         int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
 280         struct kvm_kernel_irq_routing_entry *rt_entries;
 281         u32 nr_rt_entries;
 282         /*
 283          * Array indexed by gsi. Each entry contains list of irq chips
 284          * the gsi is connected to.
 285          */
 286         struct hlist_head map[0];
 287 };
 288
 289 #else
 290
 291 struct kvm_irq_routing_table {};
 292
 293 #endif
 294
 295 #ifndef KVM_MEM_SLOTS_NUM
 296 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 297 #endif
 298
 299 /*
 300  * Note:
 301  * memslots are not sorted by id anymore, please use id_to_memslot()
 302  * to get the memslot by its id.
 303  */
 304 struct kvm_memslots {
 305         u64 generation;
 306         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 307         /* The mapping table from slot id to the index in memslots[]. */
 308         int id_to_index[KVM_MEM_SLOTS_NUM];
 309 };
 310
 311 struct kvm {
 312         spinlock_t mmu_lock;
 313         struct mutex slots_lock;
 314         struct mm_struct *mm; /* userspace tied to this vm */
 315         struct kvm_memslots *memslots;
 316         struct srcu_struct srcu;
 317 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 318         u32 bsp_vcpu_id;
 319 #endif
 320         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 321         atomic_t online_vcpus;
 322         int last_boosted_vcpu;
 323         struct list_head vm_list;
 324         struct mutex lock;
 325         struct kvm_io_bus *buses[KVM_NR_BUSES];
 326 #ifdef CONFIG_HAVE_KVM_EVENTFD
 327         struct {
 328                 spinlock_t        lock;
 329                 struct list_head  items;
 330         } irqfds;
 331         struct list_head ioeventfds;
 332 #endif
 333         struct kvm_vm_stat stat;
 334         struct kvm_arch arch;
 335         atomic_t users_count;
 336 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 337         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 338         spinlock_t ring_lock;
 339         struct list_head coalesced_zones;
 340 #endif
 341
 342         struct mutex irq_lock;
 343 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 344         /*
 345          * Update side is protected by irq_lock and,
 346          * if configured, irqfds.lock.
 347          */
 348         struct kvm_irq_routing_table __rcu *irq_routing;
 349         struct hlist_head mask_notifier_list;
 350         struct hlist_head irq_ack_notifier_list;
 351 #endif
 352
 353 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 354         struct mmu_notifier mmu_notifier;
 355         unsigned long mmu_notifier_seq;
 356         long mmu_notifier_count;
 357 #endif
 358         long tlbs_dirty;
 359 };
 360
 361 #define kvm_err(fmt, ...) \
 362         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 363 #define kvm_info(fmt, ...) \
 364         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 365 #define kvm_debug(fmt, ...) \
 366         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 367 #define kvm_pr_unimpl(fmt, ...) \
 368         pr_err_ratelimited("kvm [%i]: " fmt, \
 369                            task_tgid_nr(current), ## __VA_ARGS__)
 370
 371 /* The guest did something we don't support. */
 372 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 373         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 374
 375 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 376 {
 377         smp_rmb();
 378         return kvm->vcpus[i];
 379 }
 380
 381 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 382         for (idx = 0; \
 383              idx < atomic_read(&kvm->online_vcpus) && \
 384              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 385              idx++)
 386
 387 #define kvm_for_each_memslot(memslot, slots)    \
 388         for (memslot = &slots->memslots[0];     \
 389               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 390                 memslot++)
 391
 392 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 393 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 394
 395 void vcpu_load(struct kvm_vcpu *vcpu);
 396 void vcpu_put(struct kvm_vcpu *vcpu);
 397
 398 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 399                   struct module *module);
 400 void kvm_exit(void);
 401
 402 void kvm_get_kvm(struct kvm *kvm);
 403 void kvm_put_kvm(struct kvm *kvm);
 404 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
 405
 406 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 407 {
 408         return rcu_dereference_check(kvm->memslots,
 409                         srcu_read_lock_held(&kvm->srcu)
 410                         || lockdep_is_held(&kvm->slots_lock));
 411 }
 412
 413 static inline struct kvm_memory_slot *
 414 id_to_memslot(struct kvm_memslots *slots, int id)
 415 {
 416         int index = slots->id_to_index[id];
 417         struct kvm_memory_slot *slot;
 418
 419         slot = &slots->memslots[index];
 420
 421         WARN_ON(slot->id != id);
 422         return slot;
 423 }
 424
 425 int kvm_is_error_hva(unsigned long addr);
 426 int kvm_set_memory_region(struct kvm *kvm,
 427                           struct kvm_userspace_memory_region *mem,
 428                           int user_alloc);
 429 int __kvm_set_memory_region(struct kvm *kvm,
 430                             struct kvm_userspace_memory_region *mem,
 431                             int user_alloc);
 432 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 433                            struct kvm_memory_slot *dont);
 434 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 435 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 436                                 struct kvm_memory_slot *memslot,
 437                                 struct kvm_memory_slot old,
 438                                 struct kvm_userspace_memory_region *mem,
 439                                 int user_alloc);
 440 void kvm_arch_commit_memory_region(struct kvm *kvm,
 441                                 struct kvm_userspace_memory_region *mem,
 442                                 struct kvm_memory_slot old,
 443                                 int user_alloc);
 444 bool kvm_largepages_enabled(void);
 445 void kvm_disable_largepages(void);
 446 void kvm_arch_flush_shadow(struct kvm *kvm);
 447
 448 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 449                             int nr_pages);
 450
 451 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 452 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 453 void kvm_release_page_clean(struct page *page);
 454 void kvm_release_page_dirty(struct page *page);
 455 void kvm_set_page_dirty(struct page *page);
 456 void kvm_set_page_accessed(struct page *page);
 457
 458 pfn_t hva_to_pfn_atomic(unsigned long addr);
 459 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 460 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 461                        bool write_fault, bool *writable);
 462 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 463 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 464                       bool *writable);
 465 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 466 void kvm_release_pfn_dirty(pfn_t pfn);
 467 void kvm_release_pfn_clean(pfn_t pfn);
 468 void kvm_set_pfn_dirty(pfn_t pfn);
 469 void kvm_set_pfn_accessed(pfn_t pfn);
 470 void kvm_get_pfn(pfn_t pfn);
 471
 472 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 473                         int len);
 474 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 475                           unsigned long len);
 476 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 477 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 478                            void *data, unsigned long len);
 479 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 480                          int offset, int len);
 481 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 482                     unsigned long len);
 483 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 484                            void *data, unsigned long len);
 485 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 486                               gpa_t gpa);
 487 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 488 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 489 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 490 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 491 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 492 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 493 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 494                              gfn_t gfn);
 495
 496 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 497 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 498 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 499 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 500 void kvm_resched(struct kvm_vcpu *vcpu);
 501 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 502 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 503
 504 void kvm_flush_remote_tlbs(struct kvm *kvm);
 505 void kvm_reload_remote_mmus(struct kvm *kvm);
 506
 507 long kvm_arch_dev_ioctl(struct file *filp,
 508                         unsigned int ioctl, unsigned long arg);
 509 long kvm_arch_vcpu_ioctl(struct file *filp,
 510                          unsigned int ioctl, unsigned long arg);
 511 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 512
 513 int kvm_dev_ioctl_check_extension(long ext);
 514
 515 int kvm_get_dirty_log(struct kvm *kvm,
 516                         struct kvm_dirty_log *log, int *is_dirty);
 517 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 518                                 struct kvm_dirty_log *log);
 519
 520 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 521                                    struct
 522                                    kvm_userspace_memory_region *mem,
 523                                    int user_alloc);
 524 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
 525 long kvm_arch_vm_ioctl(struct file *filp,
 526                        unsigned int ioctl, unsigned long arg);
 527
 528 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 529 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 530
 531 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 532                                     struct kvm_translation *tr);
 533
 534 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 535 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 536 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 537                                   struct kvm_sregs *sregs);
 538 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 539                                   struct kvm_sregs *sregs);
 540 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 541                                     struct kvm_mp_state *mp_state);
 542 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 543                                     struct kvm_mp_state *mp_state);
 544 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 545                                         struct kvm_guest_debug *dbg);
 546 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 547
 548 int kvm_arch_init(void *opaque);
 549 void kvm_arch_exit(void);
 550
 551 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 552 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 553
 554 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 555 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 556 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 557 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 558 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 559 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 560
 561 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
 562 int kvm_arch_hardware_enable(void *garbage);
 563 void kvm_arch_hardware_disable(void *garbage);
 564 int kvm_arch_hardware_setup(void);
 565 void kvm_arch_hardware_unsetup(void);
 566 void kvm_arch_check_processor_compat(void *rtn);
 567 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 568 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 569
 570 void kvm_free_physmem(struct kvm *kvm);
 571
 572 void *kvm_kvzalloc(unsigned long size);
 573 void kvm_kvfree(const void *addr);
 574
 575 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 576 static inline struct kvm *kvm_arch_alloc_vm(void)
 577 {
 578         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 579 }
 580
 581 static inline void kvm_arch_free_vm(struct kvm *kvm)
 582 {
 583         kfree(kvm);
 584 }
 585 #endif
 586
 587 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 588 {
 589 #ifdef __KVM_HAVE_ARCH_WQP
 590         return vcpu->arch.wqp;
 591 #else
 592         return &vcpu->wq;
 593 #endif
 594 }
 595
 596 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 597 void kvm_arch_destroy_vm(struct kvm *kvm);
 598 void kvm_free_all_assigned_devices(struct kvm *kvm);
 599 void kvm_arch_sync_events(struct kvm *kvm);
 600
 601 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 602 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 603
 604 bool kvm_is_mmio_pfn(pfn_t pfn);
 605
 606 struct kvm_irq_ack_notifier {
 607         struct hlist_node link;
 608         unsigned gsi;
 609         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 610 };
 611
 612 struct kvm_assigned_dev_kernel {
 613         struct kvm_irq_ack_notifier ack_notifier;
 614         struct list_head list;
 615         int assigned_dev_id;
 616         int host_segnr;
 617         int host_busnr;
 618         int host_devfn;
 619         unsigned int entries_nr;
 620         int host_irq;
 621         bool host_irq_disabled;
 622         bool pci_2_3;
 623         struct msix_entry *host_msix_entries;
 624         int guest_irq;
 625         struct msix_entry *guest_msix_entries;
 626         unsigned long irq_requested_type;
 627         int irq_source_id;
 628         int flags;
 629         struct pci_dev *dev;
 630         struct kvm *kvm;
 631         spinlock_t intx_lock;
 632         spinlock_t intx_mask_lock;
 633         char irq_name[32];
 634         struct pci_saved_state *pci_saved_state;
 635 };
 636
 637 struct kvm_irq_mask_notifier {
 638         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 639         int irq;
 640         struct hlist_node link;
 641 };
 642
 643 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 644                                     struct kvm_irq_mask_notifier *kimn);
 645 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 646                                       struct kvm_irq_mask_notifier *kimn);
 647 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 648                              bool mask);
 649
 650 #ifdef __KVM_HAVE_IOAPIC
 651 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
 652                                    union kvm_ioapic_redirect_entry *entry,
 653                                    unsigned long *deliver_bitmask);
 654 #endif
 655 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 656 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 657                 int irq_source_id, int level);
 658 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 659 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 660                                    struct kvm_irq_ack_notifier *kian);
 661 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 662                                    struct kvm_irq_ack_notifier *kian);
 663 int kvm_request_irq_source_id(struct kvm *kvm);
 664 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 665
 666 /* For vcpu->arch.iommu_flags */
 667 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 668
 669 #ifdef CONFIG_IOMMU_API
 670 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 671 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 672 int kvm_iommu_map_guest(struct kvm *kvm);
 673 int kvm_iommu_unmap_guest(struct kvm *kvm);
 674 int kvm_assign_device(struct kvm *kvm,
 675                       struct kvm_assigned_dev_kernel *assigned_dev);
 676 int kvm_deassign_device(struct kvm *kvm,
 677                         struct kvm_assigned_dev_kernel *assigned_dev);
 678 #else /* CONFIG_IOMMU_API */
 679 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 680                                       struct kvm_memory_slot *slot)
 681 {
 682         return 0;
 683 }
 684
 685 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 686                                          struct kvm_memory_slot *slot)
 687 {
 688 }
 689
 690 static inline int kvm_iommu_map_guest(struct kvm *kvm)
 691 {
 692         return -ENODEV;
 693 }
 694
 695 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 696 {
 697         return 0;
 698 }
 699
 700 static inline int kvm_assign_device(struct kvm *kvm,
 701                 struct kvm_assigned_dev_kernel *assigned_dev)
 702 {
 703         return 0;
 704 }
 705
 706 static inline int kvm_deassign_device(struct kvm *kvm,
 707                 struct kvm_assigned_dev_kernel *assigned_dev)
 708 {
 709         return 0;
 710 }
 711 #endif /* CONFIG_IOMMU_API */
 712
 713 static inline void kvm_guest_enter(void)
 714 {
 715         BUG_ON(preemptible());
 716         account_system_vtime(current);
 717         current->flags |= PF_VCPU;
 718         /* KVM does not hold any references to rcu protected data when it
 719          * switches CPU into a guest mode. In fact switching to a guest mode
 720          * is very similar to exiting to userspase from rcu point of view. In
 721          * addition CPU may stay in a guest mode for quite a long time (up to
 722          * one time slice). Lets treat guest mode as quiescent state, just like
 723          * we do with user-mode execution.
 724          */
 725         rcu_virt_note_context_switch(smp_processor_id());
 726 }
 727
 728 static inline void kvm_guest_exit(void)
 729 {
 730         account_system_vtime(current);
 731         current->flags &= ~PF_VCPU;
 732 }
 733
 734 /*
 735  * search_memslots() and __gfn_to_memslot() are here because they are
 736  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 737  * gfn_to_memslot() itself isn't here as an inline because that would
 738  * bloat other code too much.
 739  */
 740 static inline struct kvm_memory_slot *
 741 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 742 {
 743         struct kvm_memory_slot *memslot;
 744
 745         kvm_for_each_memslot(memslot, slots)
 746                 if (gfn >= memslot->base_gfn &&
 747                       gfn < memslot->base_gfn + memslot->npages)
 748                         return memslot;
 749
 750         return NULL;
 751 }
 752
 753 static inline struct kvm_memory_slot *
 754 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 755 {
 756         return search_memslots(slots, gfn);
 757 }
 758
 759 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 760 {
 761         return gfn_to_memslot(kvm, gfn)->id;
 762 }
 763
 764 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 765 {
 766         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 767         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 768                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 769 }
 770
 771 static inline gfn_t
 772 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 773 {
 774         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 775
 776         return slot->base_gfn + gfn_offset;
 777 }
 778
 779 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
 780                                                gfn_t gfn)
 781 {
 782         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 783 }
 784
 785 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 786 {
 787         return (gpa_t)gfn << PAGE_SHIFT;
 788 }
 789
 790 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 791 {
 792         return (gfn_t)(gpa >> PAGE_SHIFT);
 793 }
 794
 795 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 796 {
 797         return (hpa_t)pfn << PAGE_SHIFT;
 798 }
 799
 800 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 801 {
 802         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 803 }
 804
 805 enum kvm_stat_kind {
 806         KVM_STAT_VM,
 807         KVM_STAT_VCPU,
 808 };
 809
 810 struct kvm_stats_debugfs_item {
 811         const char *name;
 812         int offset;
 813         enum kvm_stat_kind kind;
 814         struct dentry *dentry;
 815 };
 816 extern struct kvm_stats_debugfs_item debugfs_entries[];
 817 extern struct dentry *kvm_debugfs_dir;
 818
 819 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 820 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
 821 {
 822         if (unlikely(vcpu->kvm->mmu_notifier_count))
 823                 return 1;
 824         /*
 825          * Ensure the read of mmu_notifier_count happens before the read
 826          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 827          * mmu_notifier_invalidate_range_end to make sure that the caller
 828          * either sees the old (non-zero) value of mmu_notifier_count or
 829          * the new (incremented) value of mmu_notifier_seq.
 830          * PowerPC Book3s HV KVM calls this under a per-page lock
 831          * rather than under kvm->mmu_lock, for scalability, so
 832          * can't rely on kvm->mmu_lock to keep things ordered.
 833          */
 834         smp_rmb();
 835         if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
 836                 return 1;
 837         return 0;
 838 }
 839 #endif
 840
 841 #ifdef KVM_CAP_IRQ_ROUTING
 842
 843 #define KVM_MAX_IRQ_ROUTES 1024
 844
 845 int kvm_setup_default_irq_routing(struct kvm *kvm);
 846 int kvm_set_irq_routing(struct kvm *kvm,
 847                         const struct kvm_irq_routing_entry *entries,
 848                         unsigned nr,
 849                         unsigned flags);
 850 void kvm_free_irq_routing(struct kvm *kvm);
 851
 852 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 853
 854 #else
 855
 856 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 857
 858 #endif
 859
 860 #ifdef CONFIG_HAVE_KVM_EVENTFD
 861
 862 void kvm_eventfd_init(struct kvm *kvm);
 863 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 864 void kvm_irqfd_release(struct kvm *kvm);
 865 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 866 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 867
 868 #else
 869
 870 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 871
 872 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 873 {
 874         return -EINVAL;
 875 }
 876
 877 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 878
 879 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 880 static inline void kvm_irq_routing_update(struct kvm *kvm,
 881                                           struct kvm_irq_routing_table *irq_rt)
 882 {
 883         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 884 }
 885 #endif
 886
 887 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 888 {
 889         return -ENOSYS;
 890 }
 891
 892 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 893
 894 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 895 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 896 {
 897         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 898 }
 899
 900 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 901
 902 #else
 903
 904 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 905
 906 #endif
 907
 908 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
 909
 910 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 911                                   unsigned long arg);
 912
 913 #else
 914
 915 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 916                                                 unsigned long arg)
 917 {
 918         return -ENOTTY;
 919 }
 920
 921 #endif
 922
 923 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
 924 {
 925         set_bit(req, &vcpu->requests);
 926 }
 927
 928 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
 929 {
 930         if (test_bit(req, &vcpu->requests)) {
 931                 clear_bit(req, &vcpu->requests);
 932                 return true;
 933         } else {
 934                 return false;
 935         }
 936 }
 937
 938 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 939
 940 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 941 {
 942         vcpu->spin_loop.in_spin_loop = val;
 943 }
 944 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 945 {
 946         vcpu->spin_loop.dy_eligible = val;
 947 }
 948
 949 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 950
 951 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 952 {
 953 }
 954
 955 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 956 {
 957 }
 958
 959 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
 960 {
 961         return true;
 962 }
 963
 964 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 965 #endif
 966