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 <asm/signal.h>
  25
  26 #include <linux/kvm.h>
  27 #include <linux/kvm_para.h>
  28
  29 #include <linux/kvm_types.h>
  30
  31 #include <asm/kvm_host.h>
  32
  33 #ifndef KVM_MMIO_SIZE
  34 #define KVM_MMIO_SIZE 8
  35 #endif
  36
  37 /*
  38  * If we support unaligned MMIO, at most one fragment will be split into two:
  39  */
  40 #ifdef KVM_UNALIGNED_MMIO
  41 #  define KVM_EXTRA_MMIO_FRAGMENTS 1
  42 #else
  43 #  define KVM_EXTRA_MMIO_FRAGMENTS 0
  44 #endif
  45
  46 #define KVM_USER_MMIO_SIZE 8
  47
  48 #define KVM_MAX_MMIO_FRAGMENTS \
  49         (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
  50
  51 /*
  52  * vcpu->requests bit members
  53  */
  54 #define KVM_REQ_TLB_FLUSH          0
  55 #define KVM_REQ_MIGRATE_TIMER      1
  56 #define KVM_REQ_REPORT_TPR_ACCESS  2
  57 #define KVM_REQ_MMU_RELOAD         3
  58 #define KVM_REQ_TRIPLE_FAULT       4
  59 #define KVM_REQ_PENDING_TIMER      5
  60 #define KVM_REQ_UNHALT             6
  61 #define KVM_REQ_MMU_SYNC           7
  62 #define KVM_REQ_CLOCK_UPDATE       8
  63 #define KVM_REQ_KICK               9
  64 #define KVM_REQ_DEACTIVATE_FPU    10
  65 #define KVM_REQ_EVENT             11
  66 #define KVM_REQ_APF_HALT          12
  67 #define KVM_REQ_STEAL_UPDATE      13
  68 #define KVM_REQ_NMI               14
  69 #define KVM_REQ_IMMEDIATE_EXIT    15
  70 #define KVM_REQ_PMU               16
  71 #define KVM_REQ_PMI               17
  72
  73 #define KVM_USERSPACE_IRQ_SOURCE_ID     0
  74
  75 struct kvm;
  76 struct kvm_vcpu;
  77 extern struct kmem_cache *kvm_vcpu_cache;
  78
  79 struct kvm_io_range {
  80         gpa_t addr;
  81         int len;
  82         struct kvm_io_device *dev;
  83 };
  84
  85 #define NR_IOBUS_DEVS 1000
  86
  87 struct kvm_io_bus {
  88         int                   dev_count;
  89         struct kvm_io_range range[];
  90 };
  91
  92 enum kvm_bus {
  93         KVM_MMIO_BUS,
  94         KVM_PIO_BUS,
  95         KVM_NR_BUSES
  96 };
  97
  98 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
  99                      int len, const void *val);
 100 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 101                     void *val);
 102 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 103                             int len, struct kvm_io_device *dev);
 104 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 105                               struct kvm_io_device *dev);
 106
 107 #ifdef CONFIG_KVM_ASYNC_PF
 108 struct kvm_async_pf {
 109         struct work_struct work;
 110         struct list_head link;
 111         struct list_head queue;
 112         struct kvm_vcpu *vcpu;
 113         struct mm_struct *mm;
 114         gva_t gva;
 115         unsigned long addr;
 116         struct kvm_arch_async_pf arch;
 117         struct page *page;
 118         bool done;
 119 };
 120
 121 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 122 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 123 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 124                        struct kvm_arch_async_pf *arch);
 125 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 126 #endif
 127
 128 enum {
 129         OUTSIDE_GUEST_MODE,
 130         IN_GUEST_MODE,
 131         EXITING_GUEST_MODE,
 132         READING_SHADOW_PAGE_TABLES,
 133 };
 134
 135 /*
 136  * Sometimes a large or cross-page mmio needs to be broken up into separate
 137  * exits for userspace servicing.
 138  */
 139 struct kvm_mmio_fragment {
 140         gpa_t gpa;
 141         void *data;
 142         unsigned len;
 143 };
 144
 145 struct kvm_vcpu {
 146         struct kvm *kvm;
 147 #ifdef CONFIG_PREEMPT_NOTIFIERS
 148         struct preempt_notifier preempt_notifier;
 149 #endif
 150         int cpu;
 151         int vcpu_id;
 152         int srcu_idx;
 153         int mode;
 154         unsigned long requests;
 155         unsigned long guest_debug;
 156
 157         struct mutex mutex;
 158         struct kvm_run *run;
 159
 160         int fpu_active;
 161         int guest_fpu_loaded, guest_xcr0_loaded;
 162         wait_queue_head_t wq;
 163         struct pid *pid;
 164         int sigset_active;
 165         sigset_t sigset;
 166         struct kvm_vcpu_stat stat;
 167
 168 #ifdef CONFIG_HAS_IOMEM
 169         int mmio_needed;
 170         int mmio_read_completed;
 171         int mmio_is_write;
 172         int mmio_cur_fragment;
 173         int mmio_nr_fragments;
 174         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 175 #endif
 176
 177 #ifdef CONFIG_KVM_ASYNC_PF
 178         struct {
 179                 u32 queued;
 180                 struct list_head queue;
 181                 struct list_head done;
 182                 spinlock_t lock;
 183         } async_pf;
 184 #endif
 185
 186 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 187         /*
 188          * Cpu relax intercept or pause loop exit optimization
 189          * in_spin_loop: set when a vcpu does a pause loop exit
 190          *  or cpu relax intercepted.
 191          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 192          */
 193         struct {
 194                 bool in_spin_loop;
 195                 bool dy_eligible;
 196         } spin_loop;
 197 #endif
 198         struct kvm_vcpu_arch arch;
 199 };
 200
 201 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 202 {
 203         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 204 }
 205
 206 /*
 207  * Some of the bitops functions do not support too long bitmaps.
 208  * This number must be determined not to exceed such limits.
 209  */
 210 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 211
 212 struct kvm_memory_slot {
 213         gfn_t base_gfn;
 214         unsigned long npages;
 215         unsigned long flags;
 216         unsigned long *rmap;
 217         unsigned long *dirty_bitmap;
 218         struct kvm_arch_memory_slot arch;
 219         unsigned long userspace_addr;
 220         int user_alloc;
 221         int id;
 222 };
 223
 224 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 225 {
 226         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 227 }
 228
 229 struct kvm_kernel_irq_routing_entry {
 230         u32 gsi;
 231         u32 type;
 232         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 233                    struct kvm *kvm, int irq_source_id, int level);
 234         union {
 235                 struct {
 236                         unsigned irqchip;
 237                         unsigned pin;
 238                 } irqchip;
 239                 struct msi_msg msi;
 240         };
 241         struct hlist_node link;
 242 };
 243
 244 #ifdef __KVM_HAVE_IOAPIC
 245
 246 struct kvm_irq_routing_table {
 247         int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
 248         struct kvm_kernel_irq_routing_entry *rt_entries;
 249         u32 nr_rt_entries;
 250         /*
 251          * Array indexed by gsi. Each entry contains list of irq chips
 252          * the gsi is connected to.
 253          */
 254         struct hlist_head map[0];
 255 };
 256
 257 #else
 258
 259 struct kvm_irq_routing_table {};
 260
 261 #endif
 262
 263 #ifndef KVM_MEM_SLOTS_NUM
 264 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 265 #endif
 266
 267 /*
 268  * Note:
 269  * memslots are not sorted by id anymore, please use id_to_memslot()
 270  * to get the memslot by its id.
 271  */
 272 struct kvm_memslots {
 273         u64 generation;
 274         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 275         /* The mapping table from slot id to the index in memslots[]. */
 276         int id_to_index[KVM_MEM_SLOTS_NUM];
 277 };
 278
 279 struct kvm {
 280         spinlock_t mmu_lock;
 281         struct mutex slots_lock;
 282         struct mm_struct *mm; /* userspace tied to this vm */
 283         struct kvm_memslots *memslots;
 284         struct srcu_struct srcu;
 285 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 286         u32 bsp_vcpu_id;
 287 #endif
 288         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 289         atomic_t online_vcpus;
 290         int last_boosted_vcpu;
 291         struct list_head vm_list;
 292         struct mutex lock;
 293         struct kvm_io_bus *buses[KVM_NR_BUSES];
 294 #ifdef CONFIG_HAVE_KVM_EVENTFD
 295         struct {
 296                 spinlock_t        lock;
 297                 struct list_head  items;
 298         } irqfds;
 299         struct list_head ioeventfds;
 300 #endif
 301         struct kvm_vm_stat stat;
 302         struct kvm_arch arch;
 303         atomic_t users_count;
 304 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 305         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 306         spinlock_t ring_lock;
 307         struct list_head coalesced_zones;
 308 #endif
 309
 310         struct mutex irq_lock;
 311 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 312         /*
 313          * Update side is protected by irq_lock and,
 314          * if configured, irqfds.lock.
 315          */
 316         struct kvm_irq_routing_table __rcu *irq_routing;
 317         struct hlist_head mask_notifier_list;
 318         struct hlist_head irq_ack_notifier_list;
 319 #endif
 320
 321 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 322         struct mmu_notifier mmu_notifier;
 323         unsigned long mmu_notifier_seq;
 324         long mmu_notifier_count;
 325 #endif
 326         long tlbs_dirty;
 327 };
 328
 329 #define kvm_err(fmt, ...) \
 330         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 331 #define kvm_info(fmt, ...) \
 332         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 333 #define kvm_debug(fmt, ...) \
 334         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 335 #define kvm_pr_unimpl(fmt, ...) \
 336         pr_err_ratelimited("kvm [%i]: " fmt, \
 337                            task_tgid_nr(current), ## __VA_ARGS__)
 338
 339 /* The guest did something we don't support. */
 340 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 341         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 342
 343 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 344 {
 345         smp_rmb();
 346         return kvm->vcpus[i];
 347 }
 348
 349 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 350         for (idx = 0; \
 351              idx < atomic_read(&kvm->online_vcpus) && \
 352              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 353              idx++)
 354
 355 #define kvm_for_each_memslot(memslot, slots)    \
 356         for (memslot = &slots->memslots[0];     \
 357               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 358                 memslot++)
 359
 360 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 361 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 362
 363 void vcpu_load(struct kvm_vcpu *vcpu);
 364 void vcpu_put(struct kvm_vcpu *vcpu);
 365
 366 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 367                   struct module *module);
 368 void kvm_exit(void);
 369
 370 void kvm_get_kvm(struct kvm *kvm);
 371 void kvm_put_kvm(struct kvm *kvm);
 372 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
 373
 374 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 375 {
 376         return rcu_dereference_check(kvm->memslots,
 377                         srcu_read_lock_held(&kvm->srcu)
 378                         || lockdep_is_held(&kvm->slots_lock));
 379 }
 380
 381 static inline struct kvm_memory_slot *
 382 id_to_memslot(struct kvm_memslots *slots, int id)
 383 {
 384         int index = slots->id_to_index[id];
 385         struct kvm_memory_slot *slot;
 386
 387         slot = &slots->memslots[index];
 388
 389         WARN_ON(slot->id != id);
 390         return slot;
 391 }
 392
 393 extern struct page *bad_page;
 394
 395 int is_error_page(struct page *page);
 396 int is_error_pfn(pfn_t pfn);
 397 int is_hwpoison_pfn(pfn_t pfn);
 398 int is_noslot_pfn(pfn_t pfn);
 399 int is_invalid_pfn(pfn_t pfn);
 400 int kvm_is_error_hva(unsigned long addr);
 401 int kvm_set_memory_region(struct kvm *kvm,
 402                           struct kvm_userspace_memory_region *mem,
 403                           int user_alloc);
 404 int __kvm_set_memory_region(struct kvm *kvm,
 405                             struct kvm_userspace_memory_region *mem,
 406                             int user_alloc);
 407 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 408                            struct kvm_memory_slot *dont);
 409 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 410 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 411                                 struct kvm_memory_slot *memslot,
 412                                 struct kvm_memory_slot old,
 413                                 struct kvm_userspace_memory_region *mem,
 414                                 int user_alloc);
 415 void kvm_arch_commit_memory_region(struct kvm *kvm,
 416                                 struct kvm_userspace_memory_region *mem,
 417                                 struct kvm_memory_slot old,
 418                                 int user_alloc);
 419 bool kvm_largepages_enabled(void);
 420 void kvm_disable_largepages(void);
 421 void kvm_arch_flush_shadow(struct kvm *kvm);
 422
 423 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 424                             int nr_pages);
 425
 426 struct page *get_bad_page(void);
 427 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 428 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 429 void kvm_release_page_clean(struct page *page);
 430 void kvm_release_page_dirty(struct page *page);
 431 void kvm_set_page_dirty(struct page *page);
 432 void kvm_set_page_accessed(struct page *page);
 433
 434 pfn_t hva_to_pfn_atomic(unsigned long addr);
 435 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 436 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 437                        bool write_fault, bool *writable);
 438 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 439 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 440                       bool *writable);
 441 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 442 void kvm_release_pfn_dirty(pfn_t);
 443 void kvm_release_pfn_clean(pfn_t pfn);
 444 void kvm_set_pfn_dirty(pfn_t pfn);
 445 void kvm_set_pfn_accessed(pfn_t pfn);
 446 void kvm_get_pfn(pfn_t pfn);
 447 pfn_t get_fault_pfn(void);
 448
 449 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 450                         int len);
 451 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 452                           unsigned long len);
 453 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 454 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 455                            void *data, unsigned long len);
 456 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 457                          int offset, int len);
 458 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 459                     unsigned long len);
 460 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 461                            void *data, unsigned long len);
 462 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 463                               gpa_t gpa);
 464 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 465 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 466 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 467 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 468 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 469 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 470 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 471                              gfn_t gfn);
 472
 473 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 474 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 475 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 476 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 477 void kvm_resched(struct kvm_vcpu *vcpu);
 478 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 479 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 480
 481 void kvm_flush_remote_tlbs(struct kvm *kvm);
 482 void kvm_reload_remote_mmus(struct kvm *kvm);
 483
 484 long kvm_arch_dev_ioctl(struct file *filp,
 485                         unsigned int ioctl, unsigned long arg);
 486 long kvm_arch_vcpu_ioctl(struct file *filp,
 487                          unsigned int ioctl, unsigned long arg);
 488 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 489
 490 int kvm_dev_ioctl_check_extension(long ext);
 491
 492 int kvm_get_dirty_log(struct kvm *kvm,
 493                         struct kvm_dirty_log *log, int *is_dirty);
 494 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 495                                 struct kvm_dirty_log *log);
 496
 497 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 498                                    struct
 499                                    kvm_userspace_memory_region *mem,
 500                                    int user_alloc);
 501 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
 502 long kvm_arch_vm_ioctl(struct file *filp,
 503                        unsigned int ioctl, unsigned long arg);
 504
 505 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 506 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 507
 508 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 509                                     struct kvm_translation *tr);
 510
 511 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 512 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 513 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 514                                   struct kvm_sregs *sregs);
 515 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 516                                   struct kvm_sregs *sregs);
 517 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 518                                     struct kvm_mp_state *mp_state);
 519 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 520                                     struct kvm_mp_state *mp_state);
 521 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 522                                         struct kvm_guest_debug *dbg);
 523 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 524
 525 int kvm_arch_init(void *opaque);
 526 void kvm_arch_exit(void);
 527
 528 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 529 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 530
 531 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 532 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 533 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 534 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 535 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 536 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 537
 538 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
 539 int kvm_arch_hardware_enable(void *garbage);
 540 void kvm_arch_hardware_disable(void *garbage);
 541 int kvm_arch_hardware_setup(void);
 542 void kvm_arch_hardware_unsetup(void);
 543 void kvm_arch_check_processor_compat(void *rtn);
 544 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 545 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 546
 547 void kvm_free_physmem(struct kvm *kvm);
 548
 549 void *kvm_kvzalloc(unsigned long size);
 550 void kvm_kvfree(const void *addr);
 551
 552 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 553 static inline struct kvm *kvm_arch_alloc_vm(void)
 554 {
 555         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 556 }
 557
 558 static inline void kvm_arch_free_vm(struct kvm *kvm)
 559 {
 560         kfree(kvm);
 561 }
 562 #endif
 563
 564 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 565 {
 566 #ifdef __KVM_HAVE_ARCH_WQP
 567         return vcpu->arch.wqp;
 568 #else
 569         return &vcpu->wq;
 570 #endif
 571 }
 572
 573 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 574 void kvm_arch_destroy_vm(struct kvm *kvm);
 575 void kvm_free_all_assigned_devices(struct kvm *kvm);
 576 void kvm_arch_sync_events(struct kvm *kvm);
 577
 578 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 579 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 580
 581 bool kvm_is_mmio_pfn(pfn_t pfn);
 582
 583 struct kvm_irq_ack_notifier {
 584         struct hlist_node link;
 585         unsigned gsi;
 586         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 587 };
 588
 589 struct kvm_assigned_dev_kernel {
 590         struct kvm_irq_ack_notifier ack_notifier;
 591         struct list_head list;
 592         int assigned_dev_id;
 593         int host_segnr;
 594         int host_busnr;
 595         int host_devfn;
 596         unsigned int entries_nr;
 597         int host_irq;
 598         bool host_irq_disabled;
 599         bool pci_2_3;
 600         struct msix_entry *host_msix_entries;
 601         int guest_irq;
 602         struct msix_entry *guest_msix_entries;
 603         unsigned long irq_requested_type;
 604         int irq_source_id;
 605         int flags;
 606         struct pci_dev *dev;
 607         struct kvm *kvm;
 608         spinlock_t intx_lock;
 609         spinlock_t intx_mask_lock;
 610         char irq_name[32];
 611         struct pci_saved_state *pci_saved_state;
 612 };
 613
 614 struct kvm_irq_mask_notifier {
 615         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 616         int irq;
 617         struct hlist_node link;
 618 };
 619
 620 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 621                                     struct kvm_irq_mask_notifier *kimn);
 622 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 623                                       struct kvm_irq_mask_notifier *kimn);
 624 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 625                              bool mask);
 626
 627 #ifdef __KVM_HAVE_IOAPIC
 628 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
 629                                    union kvm_ioapic_redirect_entry *entry,
 630                                    unsigned long *deliver_bitmask);
 631 #endif
 632 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 633 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 634                 int irq_source_id, int level);
 635 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 636 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 637                                    struct kvm_irq_ack_notifier *kian);
 638 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 639                                    struct kvm_irq_ack_notifier *kian);
 640 int kvm_request_irq_source_id(struct kvm *kvm);
 641 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 642
 643 /* For vcpu->arch.iommu_flags */
 644 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 645
 646 #ifdef CONFIG_IOMMU_API
 647 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 648 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 649 int kvm_iommu_map_guest(struct kvm *kvm);
 650 int kvm_iommu_unmap_guest(struct kvm *kvm);
 651 int kvm_assign_device(struct kvm *kvm,
 652                       struct kvm_assigned_dev_kernel *assigned_dev);
 653 int kvm_deassign_device(struct kvm *kvm,
 654                         struct kvm_assigned_dev_kernel *assigned_dev);
 655 #else /* CONFIG_IOMMU_API */
 656 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 657                                       struct kvm_memory_slot *slot)
 658 {
 659         return 0;
 660 }
 661
 662 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 663                                          struct kvm_memory_slot *slot)
 664 {
 665 }
 666
 667 static inline int kvm_iommu_map_guest(struct kvm *kvm)
 668 {
 669         return -ENODEV;
 670 }
 671
 672 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 673 {
 674         return 0;
 675 }
 676
 677 static inline int kvm_assign_device(struct kvm *kvm,
 678                 struct kvm_assigned_dev_kernel *assigned_dev)
 679 {
 680         return 0;
 681 }
 682
 683 static inline int kvm_deassign_device(struct kvm *kvm,
 684                 struct kvm_assigned_dev_kernel *assigned_dev)
 685 {
 686         return 0;
 687 }
 688 #endif /* CONFIG_IOMMU_API */
 689
 690 static inline void kvm_guest_enter(void)
 691 {
 692         BUG_ON(preemptible());
 693         account_system_vtime(current);
 694         current->flags |= PF_VCPU;
 695         /* KVM does not hold any references to rcu protected data when it
 696          * switches CPU into a guest mode. In fact switching to a guest mode
 697          * is very similar to exiting to userspase from rcu point of view. In
 698          * addition CPU may stay in a guest mode for quite a long time (up to
 699          * one time slice). Lets treat guest mode as quiescent state, just like
 700          * we do with user-mode execution.
 701          */
 702         rcu_virt_note_context_switch(smp_processor_id());
 703 }
 704
 705 static inline void kvm_guest_exit(void)
 706 {
 707         account_system_vtime(current);
 708         current->flags &= ~PF_VCPU;
 709 }
 710
 711 /*
 712  * search_memslots() and __gfn_to_memslot() are here because they are
 713  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 714  * gfn_to_memslot() itself isn't here as an inline because that would
 715  * bloat other code too much.
 716  */
 717 static inline struct kvm_memory_slot *
 718 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 719 {
 720         struct kvm_memory_slot *memslot;
 721
 722         kvm_for_each_memslot(memslot, slots)
 723                 if (gfn >= memslot->base_gfn &&
 724                       gfn < memslot->base_gfn + memslot->npages)
 725                         return memslot;
 726
 727         return NULL;
 728 }
 729
 730 static inline struct kvm_memory_slot *
 731 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 732 {
 733         return search_memslots(slots, gfn);
 734 }
 735
 736 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 737 {
 738         return gfn_to_memslot(kvm, gfn)->id;
 739 }
 740
 741 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 742 {
 743         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 744         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 745                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 746 }
 747
 748 static inline gfn_t
 749 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 750 {
 751         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 752
 753         return slot->base_gfn + gfn_offset;
 754 }
 755
 756 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
 757                                                gfn_t gfn)
 758 {
 759         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 760 }
 761
 762 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 763 {
 764         return (gpa_t)gfn << PAGE_SHIFT;
 765 }
 766
 767 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 768 {
 769         return (gfn_t)(gpa >> PAGE_SHIFT);
 770 }
 771
 772 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 773 {
 774         return (hpa_t)pfn << PAGE_SHIFT;
 775 }
 776
 777 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 778 {
 779         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 780 }
 781
 782 enum kvm_stat_kind {
 783         KVM_STAT_VM,
 784         KVM_STAT_VCPU,
 785 };
 786
 787 struct kvm_stats_debugfs_item {
 788         const char *name;
 789         int offset;
 790         enum kvm_stat_kind kind;
 791         struct dentry *dentry;
 792 };
 793 extern struct kvm_stats_debugfs_item debugfs_entries[];
 794 extern struct dentry *kvm_debugfs_dir;
 795
 796 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 797 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
 798 {
 799         if (unlikely(vcpu->kvm->mmu_notifier_count))
 800                 return 1;
 801         /*
 802          * Ensure the read of mmu_notifier_count happens before the read
 803          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 804          * mmu_notifier_invalidate_range_end to make sure that the caller
 805          * either sees the old (non-zero) value of mmu_notifier_count or
 806          * the new (incremented) value of mmu_notifier_seq.
 807          * PowerPC Book3s HV KVM calls this under a per-page lock
 808          * rather than under kvm->mmu_lock, for scalability, so
 809          * can't rely on kvm->mmu_lock to keep things ordered.
 810          */
 811         smp_rmb();
 812         if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
 813                 return 1;
 814         return 0;
 815 }
 816 #endif
 817
 818 #ifdef KVM_CAP_IRQ_ROUTING
 819
 820 #define KVM_MAX_IRQ_ROUTES 1024
 821
 822 int kvm_setup_default_irq_routing(struct kvm *kvm);
 823 int kvm_set_irq_routing(struct kvm *kvm,
 824                         const struct kvm_irq_routing_entry *entries,
 825                         unsigned nr,
 826                         unsigned flags);
 827 void kvm_free_irq_routing(struct kvm *kvm);
 828
 829 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 830
 831 #else
 832
 833 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 834
 835 #endif
 836
 837 #ifdef CONFIG_HAVE_KVM_EVENTFD
 838
 839 void kvm_eventfd_init(struct kvm *kvm);
 840 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 841 void kvm_irqfd_release(struct kvm *kvm);
 842 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 843 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 844
 845 #else
 846
 847 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 848
 849 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 850 {
 851         return -EINVAL;
 852 }
 853
 854 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 855
 856 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 857 static inline void kvm_irq_routing_update(struct kvm *kvm,
 858                                           struct kvm_irq_routing_table *irq_rt)
 859 {
 860         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 861 }
 862 #endif
 863
 864 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 865 {
 866         return -ENOSYS;
 867 }
 868
 869 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 870
 871 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 872 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 873 {
 874         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 875 }
 876
 877 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 878
 879 #else
 880
 881 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 882
 883 #endif
 884
 885 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
 886
 887 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 888                                   unsigned long arg);
 889
 890 #else
 891
 892 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 893                                                 unsigned long arg)
 894 {
 895         return -ENOTTY;
 896 }
 897
 898 #endif
 899
 900 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
 901 {
 902         set_bit(req, &vcpu->requests);
 903 }
 904
 905 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
 906 {
 907         if (test_bit(req, &vcpu->requests)) {
 908                 clear_bit(req, &vcpu->requests);
 909                 return true;
 910         } else {
 911                 return false;
 912         }
 913 }
 914
 915 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 916
 917 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 918 {
 919         vcpu->spin_loop.in_spin_loop = val;
 920 }
 921 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 922 {
 923         vcpu->spin_loop.dy_eligible = val;
 924 }
 925
 926 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 927
 928 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 929 {
 930 }
 931
 932 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 933 {
 934 }
 935
 936 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
 937 {
 938         return true;
 939 }
 940
 941 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 942 #endif
 943