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