include/linux/kvm_host.h

   1 #ifndef __KVM_HOST_H
   2 #define __KVM_HOST_H
   3
   4 /*
   5  * This work is licensed under the terms of the GNU GPL, version 2.  See
   6  * the COPYING file in the top-level directory.
   7  */
   8
   9 #include <linux/types.h>
  10 #include <linux/hardirq.h>
  11 #include <linux/list.h>
  12 #include <linux/mutex.h>
  13 #include <linux/spinlock.h>
  14 #include <linux/signal.h>
  15 #include <linux/sched.h>
  16 #include <linux/bug.h>
  17 #include <linux/mm.h>
  18 #include <linux/mmu_notifier.h>
  19 #include <linux/preempt.h>
  20 #include <linux/msi.h>
  21 #include <linux/slab.h>
  22 #include <linux/rcupdate.h>
  23 #include <linux/ratelimit.h>
  24 #include <linux/err.h>
  25 #include <linux/irqflags.h>
  26 #include <asm/signal.h>
  27
  28 #include <linux/kvm.h>
  29 #include <linux/kvm_para.h>
  30
  31 #include <linux/kvm_types.h>
  32
  33 #include <asm/kvm_host.h>
  34
  35 #ifndef KVM_MMIO_SIZE
  36 #define KVM_MMIO_SIZE 8
  37 #endif
  38
  39 /*
  40  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  41  * in kvm, other bits are visible for userspace which are defined in
  42  * include/linux/kvm_h.
  43  */
  44 #define KVM_MEMSLOT_INVALID     (1UL << 16)
  45
  46 /* Two fragments for cross MMIO pages. */
  47 #define KVM_MAX_MMIO_FRAGMENTS  2
  48
  49 /*
  50  * For the normal pfn, the highest 12 bits should be zero,
  51  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  52  * mask bit 63 to indicate the noslot pfn.
  53  */
  54 #define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  55 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  56 #define KVM_PFN_NOSLOT          (0x1ULL << 63)
  57
  58 #define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  59 #define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  60 #define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  61
  62 /*
  63  * error pfns indicate that the gfn is in slot but faild to
  64  * translate it to pfn on host.
  65  */
  66 static inline bool is_error_pfn(pfn_t pfn)
  67 {
  68         return !!(pfn & KVM_PFN_ERR_MASK);
  69 }
  70
  71 /*
  72  * error_noslot pfns indicate that the gfn can not be
  73  * translated to pfn - it is not in slot or failed to
  74  * translate it to pfn.
  75  */
  76 static inline bool is_error_noslot_pfn(pfn_t pfn)
  77 {
  78         return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
  79 }
  80
  81 /* noslot pfn indicates that the gfn is not in slot. */
  82 static inline bool is_noslot_pfn(pfn_t pfn)
  83 {
  84         return pfn == KVM_PFN_NOSLOT;
  85 }
  86
  87 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  88 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  89
  90 static inline bool kvm_is_error_hva(unsigned long addr)
  91 {
  92         return addr >= PAGE_OFFSET;
  93 }
  94
  95 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
  96
  97 static inline bool is_error_page(struct page *page)
  98 {
  99         return IS_ERR(page);
 100 }
 101
 102 /*
 103  * vcpu->requests bit members
 104  */
 105 #define KVM_REQ_TLB_FLUSH          0
 106 #define KVM_REQ_MIGRATE_TIMER      1
 107 #define KVM_REQ_REPORT_TPR_ACCESS  2
 108 #define KVM_REQ_MMU_RELOAD         3
 109 #define KVM_REQ_TRIPLE_FAULT       4
 110 #define KVM_REQ_PENDING_TIMER      5
 111 #define KVM_REQ_UNHALT             6
 112 #define KVM_REQ_MMU_SYNC           7
 113 #define KVM_REQ_CLOCK_UPDATE       8
 114 #define KVM_REQ_KICK               9
 115 #define KVM_REQ_DEACTIVATE_FPU    10
 116 #define KVM_REQ_EVENT             11
 117 #define KVM_REQ_APF_HALT          12
 118 #define KVM_REQ_STEAL_UPDATE      13
 119 #define KVM_REQ_NMI               14
 120 #define KVM_REQ_PMU               15
 121 #define KVM_REQ_PMI               16
 122 #define KVM_REQ_WATCHDOG          17
 123 #define KVM_REQ_MASTERCLOCK_UPDATE 18
 124 #define KVM_REQ_MCLOCK_INPROGRESS 19
 125 #define KVM_REQ_EPR_EXIT          20
 126 #define KVM_REQ_SCAN_IOAPIC       21
 127 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
 128
 129 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 131
 132 struct kvm;
 133 struct kvm_vcpu;
 134 extern struct kmem_cache *kvm_vcpu_cache;
 135
 136 extern raw_spinlock_t kvm_lock;
 137 extern struct list_head vm_list;
 138
 139 struct kvm_io_range {
 140         gpa_t addr;
 141         int len;
 142         struct kvm_io_device *dev;
 143 };
 144
 145 #define NR_IOBUS_DEVS 1000
 146
 147 struct kvm_io_bus {
 148         int dev_count;
 149         int ioeventfd_count;
 150         struct kvm_io_range range[];
 151 };
 152
 153 enum kvm_bus {
 154         KVM_MMIO_BUS,
 155         KVM_PIO_BUS,
 156         KVM_VIRTIO_CCW_NOTIFY_BUS,
 157         KVM_NR_BUSES
 158 };
 159
 160 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 161                      int len, const void *val);
 162 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 163                     void *val);
 164 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 165                             int len, struct kvm_io_device *dev);
 166 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 167                               struct kvm_io_device *dev);
 168
 169 #ifdef CONFIG_KVM_ASYNC_PF
 170 struct kvm_async_pf {
 171         struct work_struct work;
 172         struct list_head link;
 173         struct list_head queue;
 174         struct kvm_vcpu *vcpu;
 175         struct mm_struct *mm;
 176         gva_t gva;
 177         unsigned long addr;
 178         struct kvm_arch_async_pf arch;
 179         struct page *page;
 180         bool done;
 181 };
 182
 183 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 184 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 185 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
 186                        struct kvm_arch_async_pf *arch);
 187 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 188 #endif
 189
 190 enum {
 191         OUTSIDE_GUEST_MODE,
 192         IN_GUEST_MODE,
 193         EXITING_GUEST_MODE,
 194         READING_SHADOW_PAGE_TABLES,
 195 };
 196
 197 /*
 198  * Sometimes a large or cross-page mmio needs to be broken up into separate
 199  * exits for userspace servicing.
 200  */
 201 struct kvm_mmio_fragment {
 202         gpa_t gpa;
 203         void *data;
 204         unsigned len;
 205 };
 206
 207 struct kvm_vcpu {
 208         struct kvm *kvm;
 209 #ifdef CONFIG_PREEMPT_NOTIFIERS
 210         struct preempt_notifier preempt_notifier;
 211 #endif
 212         int cpu;
 213         int vcpu_id;
 214         int srcu_idx;
 215         int mode;
 216         unsigned long requests;
 217         unsigned long guest_debug;
 218
 219         struct mutex mutex;
 220         struct kvm_run *run;
 221
 222         int fpu_active;
 223         int guest_fpu_loaded, guest_xcr0_loaded;
 224         wait_queue_head_t wq;
 225         struct pid *pid;
 226         int sigset_active;
 227         sigset_t sigset;
 228         struct kvm_vcpu_stat stat;
 229
 230 #ifdef CONFIG_HAS_IOMEM
 231         int mmio_needed;
 232         int mmio_read_completed;
 233         int mmio_is_write;
 234         int mmio_cur_fragment;
 235         int mmio_nr_fragments;
 236         struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 237 #endif
 238
 239 #ifdef CONFIG_KVM_ASYNC_PF
 240         struct {
 241                 u32 queued;
 242                 struct list_head queue;
 243                 struct list_head done;
 244                 spinlock_t lock;
 245         } async_pf;
 246 #endif
 247
 248 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 249         /*
 250          * Cpu relax intercept or pause loop exit optimization
 251          * in_spin_loop: set when a vcpu does a pause loop exit
 252          *  or cpu relax intercepted.
 253          * dy_eligible: indicates whether vcpu is eligible for directed yield.
 254          */
 255         struct {
 256                 bool in_spin_loop;
 257                 bool dy_eligible;
 258         } spin_loop;
 259 #endif
 260         bool preempted;
 261         struct kvm_vcpu_arch arch;
 262 };
 263
 264 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 265 {
 266         return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 267 }
 268
 269 /*
 270  * Some of the bitops functions do not support too long bitmaps.
 271  * This number must be determined not to exceed such limits.
 272  */
 273 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 274
 275 struct kvm_memory_slot {
 276         gfn_t base_gfn;
 277         unsigned long npages;
 278         unsigned long *dirty_bitmap;
 279         struct kvm_arch_memory_slot arch;
 280         unsigned long userspace_addr;
 281         u32 flags;
 282         short id;
 283 };
 284
 285 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 286 {
 287         return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 288 }
 289
 290 struct kvm_kernel_irq_routing_entry {
 291         u32 gsi;
 292         u32 type;
 293         int (*set)(struct kvm_kernel_irq_routing_entry *e,
 294                    struct kvm *kvm, int irq_source_id, int level,
 295                    bool line_status);
 296         union {
 297                 struct {
 298                         unsigned irqchip;
 299                         unsigned pin;
 300                 } irqchip;
 301                 struct msi_msg msi;
 302         };
 303         struct hlist_node link;
 304 };
 305
 306 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 307
 308 struct kvm_irq_routing_table {
 309         int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
 310         struct kvm_kernel_irq_routing_entry *rt_entries;
 311         u32 nr_rt_entries;
 312         /*
 313          * Array indexed by gsi. Each entry contains list of irq chips
 314          * the gsi is connected to.
 315          */
 316         struct hlist_head map[0];
 317 };
 318
 319 #else
 320
 321 struct kvm_irq_routing_table {};
 322
 323 #endif
 324
 325 #ifndef KVM_PRIVATE_MEM_SLOTS
 326 #define KVM_PRIVATE_MEM_SLOTS 0
 327 #endif
 328
 329 #ifndef KVM_MEM_SLOTS_NUM
 330 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 331 #endif
 332
 333 /*
 334  * Note:
 335  * memslots are not sorted by id anymore, please use id_to_memslot()
 336  * to get the memslot by its id.
 337  */
 338 struct kvm_memslots {
 339         u64 generation;
 340         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 341         /* The mapping table from slot id to the index in memslots[]. */
 342         short id_to_index[KVM_MEM_SLOTS_NUM];
 343 };
 344
 345 struct kvm {
 346         spinlock_t mmu_lock;
 347         struct mutex slots_lock;
 348         struct mm_struct *mm; /* userspace tied to this vm */
 349         struct kvm_memslots *memslots;
 350         struct srcu_struct srcu;
 351 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 352         u32 bsp_vcpu_id;
 353 #endif
 354         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 355         atomic_t online_vcpus;
 356         int last_boosted_vcpu;
 357         struct list_head vm_list;
 358         struct mutex lock;
 359         struct kvm_io_bus *buses[KVM_NR_BUSES];
 360 #ifdef CONFIG_HAVE_KVM_EVENTFD
 361         struct {
 362                 spinlock_t        lock;
 363                 struct list_head  items;
 364                 struct list_head  resampler_list;
 365                 struct mutex      resampler_lock;
 366         } irqfds;
 367         struct list_head ioeventfds;
 368 #endif
 369         struct kvm_vm_stat stat;
 370         struct kvm_arch arch;
 371         atomic_t users_count;
 372 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 373         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 374         spinlock_t ring_lock;
 375         struct list_head coalesced_zones;
 376 #endif
 377
 378         struct mutex irq_lock;
 379 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 380         /*
 381          * Update side is protected by irq_lock and,
 382          * if configured, irqfds.lock.
 383          */
 384         struct kvm_irq_routing_table __rcu *irq_routing;
 385         struct hlist_head mask_notifier_list;
 386         struct hlist_head irq_ack_notifier_list;
 387 #endif
 388
 389 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 390         struct mmu_notifier mmu_notifier;
 391         unsigned long mmu_notifier_seq;
 392         long mmu_notifier_count;
 393 #endif
 394         long tlbs_dirty;
 395         struct list_head devices;
 396 };
 397
 398 #define kvm_err(fmt, ...) \
 399         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 400 #define kvm_info(fmt, ...) \
 401         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 402 #define kvm_debug(fmt, ...) \
 403         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 404 #define kvm_pr_unimpl(fmt, ...) \
 405         pr_err_ratelimited("kvm [%i]: " fmt, \
 406                            task_tgid_nr(current), ## __VA_ARGS__)
 407
 408 /* The guest did something we don't support. */
 409 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 410         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 411
 412 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 413 {
 414         smp_rmb();
 415         return kvm->vcpus[i];
 416 }
 417
 418 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 419         for (idx = 0; \
 420              idx < atomic_read(&kvm->online_vcpus) && \
 421              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 422              idx++)
 423
 424 #define kvm_for_each_memslot(memslot, slots)    \
 425         for (memslot = &slots->memslots[0];     \
 426               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 427                 memslot++)
 428
 429 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 430 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 431
 432 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 433 void vcpu_put(struct kvm_vcpu *vcpu);
 434
 435 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 436 int kvm_irqfd_init(void);
 437 void kvm_irqfd_exit(void);
 438 #else
 439 static inline int kvm_irqfd_init(void)
 440 {
 441         return 0;
 442 }
 443
 444 static inline void kvm_irqfd_exit(void)
 445 {
 446 }
 447 #endif
 448 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 449                   struct module *module);
 450 void kvm_exit(void);
 451
 452 void kvm_get_kvm(struct kvm *kvm);
 453 void kvm_put_kvm(struct kvm *kvm);
 454 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new,
 455                      u64 last_generation);
 456
 457 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 458 {
 459         return rcu_dereference_check(kvm->memslots,
 460                         srcu_read_lock_held(&kvm->srcu)
 461                         || lockdep_is_held(&kvm->slots_lock));
 462 }
 463
 464 static inline struct kvm_memory_slot *
 465 id_to_memslot(struct kvm_memslots *slots, int id)
 466 {
 467         int index = slots->id_to_index[id];
 468         struct kvm_memory_slot *slot;
 469
 470         slot = &slots->memslots[index];
 471
 472         WARN_ON(slot->id != id);
 473         return slot;
 474 }
 475
 476 /*
 477  * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
 478  * - create a new memory slot
 479  * - delete an existing memory slot
 480  * - modify an existing memory slot
 481  *   -- move it in the guest physical memory space
 482  *   -- just change its flags
 483  *
 484  * Since flags can be changed by some of these operations, the following
 485  * differentiation is the best we can do for __kvm_set_memory_region():
 486  */
 487 enum kvm_mr_change {
 488         KVM_MR_CREATE,
 489         KVM_MR_DELETE,
 490         KVM_MR_MOVE,
 491         KVM_MR_FLAGS_ONLY,
 492 };
 493
 494 int kvm_set_memory_region(struct kvm *kvm,
 495                           struct kvm_userspace_memory_region *mem);
 496 int __kvm_set_memory_region(struct kvm *kvm,
 497                             struct kvm_userspace_memory_region *mem);
 498 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 499                            struct kvm_memory_slot *dont);
 500 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 501 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 502                                 struct kvm_memory_slot *memslot,
 503                                 struct kvm_userspace_memory_region *mem,
 504                                 enum kvm_mr_change change);
 505 void kvm_arch_commit_memory_region(struct kvm *kvm,
 506                                 struct kvm_userspace_memory_region *mem,
 507                                 const struct kvm_memory_slot *old,
 508                                 enum kvm_mr_change change);
 509 bool kvm_largepages_enabled(void);
 510 void kvm_disable_largepages(void);
 511 /* flush all memory translations */
 512 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 513 /* flush memory translations pointing to 'slot' */
 514 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 515                                    struct kvm_memory_slot *slot);
 516
 517 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 518                             int nr_pages);
 519
 520 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 521 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 522 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 523 void kvm_release_page_clean(struct page *page);
 524 void kvm_release_page_dirty(struct page *page);
 525 void kvm_set_page_dirty(struct page *page);
 526 void kvm_set_page_accessed(struct page *page);
 527
 528 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 529 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 530                        bool write_fault, bool *writable);
 531 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 532 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 533                       bool *writable);
 534 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 535 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 536
 537 void kvm_release_pfn_dirty(pfn_t pfn);
 538 void kvm_release_pfn_clean(pfn_t pfn);
 539 void kvm_set_pfn_dirty(pfn_t pfn);
 540 void kvm_set_pfn_accessed(pfn_t pfn);
 541 void kvm_get_pfn(pfn_t pfn);
 542
 543 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 544                         int len);
 545 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 546                           unsigned long len);
 547 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 548 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 549                            void *data, unsigned long len);
 550 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 551                          int offset, int len);
 552 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 553                     unsigned long len);
 554 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 555                            void *data, unsigned long len);
 556 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 557                               gpa_t gpa, unsigned long len);
 558 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 559 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 560 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 561 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 562 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 563 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 564 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 565                              gfn_t gfn);
 566
 567 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 568 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 569 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 570 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 571 void kvm_resched(struct kvm_vcpu *vcpu);
 572 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 573 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 574
 575 void kvm_flush_remote_tlbs(struct kvm *kvm);
 576 void kvm_reload_remote_mmus(struct kvm *kvm);
 577 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 578 void kvm_make_scan_ioapic_request(struct kvm *kvm);
 579
 580 long kvm_arch_dev_ioctl(struct file *filp,
 581                         unsigned int ioctl, unsigned long arg);
 582 long kvm_arch_vcpu_ioctl(struct file *filp,
 583                          unsigned int ioctl, unsigned long arg);
 584 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 585
 586 int kvm_dev_ioctl_check_extension(long ext);
 587
 588 int kvm_get_dirty_log(struct kvm *kvm,
 589                         struct kvm_dirty_log *log, int *is_dirty);
 590 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 591                                 struct kvm_dirty_log *log);
 592
 593 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 594                                    struct kvm_userspace_memory_region *mem);
 595 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 596                         bool line_status);
 597 long kvm_arch_vm_ioctl(struct file *filp,
 598                        unsigned int ioctl, unsigned long arg);
 599
 600 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 601 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 602
 603 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 604                                     struct kvm_translation *tr);
 605
 606 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 607 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 608 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 609                                   struct kvm_sregs *sregs);
 610 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 611                                   struct kvm_sregs *sregs);
 612 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 613                                     struct kvm_mp_state *mp_state);
 614 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 615                                     struct kvm_mp_state *mp_state);
 616 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 617                                         struct kvm_guest_debug *dbg);
 618 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 619
 620 int kvm_arch_init(void *opaque);
 621 void kvm_arch_exit(void);
 622
 623 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 624 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 625
 626 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 627 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 628 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 629 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 630 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 631 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
 632 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 633
 634 int kvm_arch_hardware_enable(void *garbage);
 635 void kvm_arch_hardware_disable(void *garbage);
 636 int kvm_arch_hardware_setup(void);
 637 void kvm_arch_hardware_unsetup(void);
 638 void kvm_arch_check_processor_compat(void *rtn);
 639 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 640 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 641
 642 void kvm_free_physmem(struct kvm *kvm);
 643
 644 void *kvm_kvzalloc(unsigned long size);
 645 void kvm_kvfree(const void *addr);
 646
 647 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 648 static inline struct kvm *kvm_arch_alloc_vm(void)
 649 {
 650         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 651 }
 652
 653 static inline void kvm_arch_free_vm(struct kvm *kvm)
 654 {
 655         kfree(kvm);
 656 }
 657 #endif
 658
 659 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 660 {
 661 #ifdef __KVM_HAVE_ARCH_WQP
 662         return vcpu->arch.wqp;
 663 #else
 664         return &vcpu->wq;
 665 #endif
 666 }
 667
 668 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 669 void kvm_arch_destroy_vm(struct kvm *kvm);
 670 void kvm_arch_sync_events(struct kvm *kvm);
 671
 672 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 673 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 674
 675 bool kvm_is_mmio_pfn(pfn_t pfn);
 676
 677 struct kvm_irq_ack_notifier {
 678         struct hlist_node link;
 679         unsigned gsi;
 680         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 681 };
 682
 683 struct kvm_assigned_dev_kernel {
 684         struct kvm_irq_ack_notifier ack_notifier;
 685         struct list_head list;
 686         int assigned_dev_id;
 687         int host_segnr;
 688         int host_busnr;
 689         int host_devfn;
 690         unsigned int entries_nr;
 691         int host_irq;
 692         bool host_irq_disabled;
 693         bool pci_2_3;
 694         struct msix_entry *host_msix_entries;
 695         int guest_irq;
 696         struct msix_entry *guest_msix_entries;
 697         unsigned long irq_requested_type;
 698         int irq_source_id;
 699         int flags;
 700         struct pci_dev *dev;
 701         struct kvm *kvm;
 702         spinlock_t intx_lock;
 703         spinlock_t intx_mask_lock;
 704         char irq_name[32];
 705         struct pci_saved_state *pci_saved_state;
 706 };
 707
 708 struct kvm_irq_mask_notifier {
 709         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 710         int irq;
 711         struct hlist_node link;
 712 };
 713
 714 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 715                                     struct kvm_irq_mask_notifier *kimn);
 716 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 717                                       struct kvm_irq_mask_notifier *kimn);
 718 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 719                              bool mask);
 720
 721 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
 722                 bool line_status);
 723 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 724 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 725                 int irq_source_id, int level, bool line_status);
 726 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
 727 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 728 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 729                                    struct kvm_irq_ack_notifier *kian);
 730 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 731                                    struct kvm_irq_ack_notifier *kian);
 732 int kvm_request_irq_source_id(struct kvm *kvm);
 733 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 734
 735 /* For vcpu->arch.iommu_flags */
 736 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 737
 738 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
 739 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 740 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 741 int kvm_iommu_map_guest(struct kvm *kvm);
 742 int kvm_iommu_unmap_guest(struct kvm *kvm);
 743 int kvm_assign_device(struct kvm *kvm,
 744                       struct kvm_assigned_dev_kernel *assigned_dev);
 745 int kvm_deassign_device(struct kvm *kvm,
 746                         struct kvm_assigned_dev_kernel *assigned_dev);
 747 #else
 748 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 749                                       struct kvm_memory_slot *slot)
 750 {
 751         return 0;
 752 }
 753
 754 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 755                                          struct kvm_memory_slot *slot)
 756 {
 757 }
 758
 759 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 760 {
 761         return 0;
 762 }
 763 #endif
 764
 765 static inline void __guest_enter(void)
 766 {
 767         /*
 768          * This is running in ioctl context so we can avoid
 769          * the call to vtime_account() with its unnecessary idle check.
 770          */
 771         vtime_account_system(current);
 772         current->flags |= PF_VCPU;
 773 }
 774
 775 static inline void __guest_exit(void)
 776 {
 777         /*
 778          * This is running in ioctl context so we can avoid
 779          * the call to vtime_account() with its unnecessary idle check.
 780          */
 781         vtime_account_system(current);
 782         current->flags &= ~PF_VCPU;
 783 }
 784
 785 #ifdef CONFIG_CONTEXT_TRACKING
 786 extern void guest_enter(void);
 787 extern void guest_exit(void);
 788
 789 #else /* !CONFIG_CONTEXT_TRACKING */
 790 static inline void guest_enter(void)
 791 {
 792         __guest_enter();
 793 }
 794
 795 static inline void guest_exit(void)
 796 {
 797         __guest_exit();
 798 }
 799 #endif /* !CONFIG_CONTEXT_TRACKING */
 800
 801 static inline void kvm_guest_enter(void)
 802 {
 803         unsigned long flags;
 804
 805         BUG_ON(preemptible());
 806
 807         local_irq_save(flags);
 808         guest_enter();
 809         local_irq_restore(flags);
 810
 811         /* KVM does not hold any references to rcu protected data when it
 812          * switches CPU into a guest mode. In fact switching to a guest mode
 813          * is very similar to exiting to userspase from rcu point of view. In
 814          * addition CPU may stay in a guest mode for quite a long time (up to
 815          * one time slice). Lets treat guest mode as quiescent state, just like
 816          * we do with user-mode execution.
 817          */
 818         rcu_virt_note_context_switch(smp_processor_id());
 819 }
 820
 821 static inline void kvm_guest_exit(void)
 822 {
 823         unsigned long flags;
 824
 825         local_irq_save(flags);
 826         guest_exit();
 827         local_irq_restore(flags);
 828 }
 829
 830 /*
 831  * search_memslots() and __gfn_to_memslot() are here because they are
 832  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 833  * gfn_to_memslot() itself isn't here as an inline because that would
 834  * bloat other code too much.
 835  */
 836 static inline struct kvm_memory_slot *
 837 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 838 {
 839         struct kvm_memory_slot *memslot;
 840
 841         kvm_for_each_memslot(memslot, slots)
 842                 if (gfn >= memslot->base_gfn &&
 843                       gfn < memslot->base_gfn + memslot->npages)
 844                         return memslot;
 845
 846         return NULL;
 847 }
 848
 849 static inline struct kvm_memory_slot *
 850 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 851 {
 852         return search_memslots(slots, gfn);
 853 }
 854
 855 static inline unsigned long
 856 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 857 {
 858         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 859 }
 860
 861 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 862 {
 863         return gfn_to_memslot(kvm, gfn)->id;
 864 }
 865
 866 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 867 {
 868         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 869         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 870                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 871 }
 872
 873 static inline gfn_t
 874 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 875 {
 876         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 877
 878         return slot->base_gfn + gfn_offset;
 879 }
 880
 881 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 882 {
 883         return (gpa_t)gfn << PAGE_SHIFT;
 884 }
 885
 886 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 887 {
 888         return (gfn_t)(gpa >> PAGE_SHIFT);
 889 }
 890
 891 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 892 {
 893         return (hpa_t)pfn << PAGE_SHIFT;
 894 }
 895
 896 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 897 {
 898         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 899 }
 900
 901 enum kvm_stat_kind {
 902         KVM_STAT_VM,
 903         KVM_STAT_VCPU,
 904 };
 905
 906 struct kvm_stats_debugfs_item {
 907         const char *name;
 908         int offset;
 909         enum kvm_stat_kind kind;
 910         struct dentry *dentry;
 911 };
 912 extern struct kvm_stats_debugfs_item debugfs_entries[];
 913 extern struct dentry *kvm_debugfs_dir;
 914
 915 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 916 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 917 {
 918         if (unlikely(kvm->mmu_notifier_count))
 919                 return 1;
 920         /*
 921          * Ensure the read of mmu_notifier_count happens before the read
 922          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 923          * mmu_notifier_invalidate_range_end to make sure that the caller
 924          * either sees the old (non-zero) value of mmu_notifier_count or
 925          * the new (incremented) value of mmu_notifier_seq.
 926          * PowerPC Book3s HV KVM calls this under a per-page lock
 927          * rather than under kvm->mmu_lock, for scalability, so
 928          * can't rely on kvm->mmu_lock to keep things ordered.
 929          */
 930         smp_rmb();
 931         if (kvm->mmu_notifier_seq != mmu_seq)
 932                 return 1;
 933         return 0;
 934 }
 935 #endif
 936
 937 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 938
 939 #define KVM_MAX_IRQ_ROUTES 1024
 940
 941 int kvm_setup_default_irq_routing(struct kvm *kvm);
 942 int kvm_set_irq_routing(struct kvm *kvm,
 943                         const struct kvm_irq_routing_entry *entries,
 944                         unsigned nr,
 945                         unsigned flags);
 946 int kvm_set_routing_entry(struct kvm_irq_routing_table *rt,
 947                           struct kvm_kernel_irq_routing_entry *e,
 948                           const struct kvm_irq_routing_entry *ue);
 949 void kvm_free_irq_routing(struct kvm *kvm);
 950
 951 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 952
 953 #else
 954
 955 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 956
 957 #endif
 958
 959 #ifdef CONFIG_HAVE_KVM_EVENTFD
 960
 961 void kvm_eventfd_init(struct kvm *kvm);
 962 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 963
 964 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 965 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 966 void kvm_irqfd_release(struct kvm *kvm);
 967 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 968 #else
 969 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 970 {
 971         return -EINVAL;
 972 }
 973
 974 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 975 #endif
 976
 977 #else
 978
 979 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 980
 981 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 982 {
 983         return -EINVAL;
 984 }
 985
 986 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 987
 988 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 989 static inline void kvm_irq_routing_update(struct kvm *kvm,
 990                                           struct kvm_irq_routing_table *irq_rt)
 991 {
 992         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 993 }
 994 #endif
 995
 996 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 997 {
 998         return -ENOSYS;
 999 }
1000
1001 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1002
1003 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1004 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
1005 {
1006         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
1007 }
1008
1009 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1010
1011 #else
1012
1013 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1014
1015 #endif
1016
1017 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1018
1019 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1020                                   unsigned long arg);
1021
1022 void kvm_free_all_assigned_devices(struct kvm *kvm);
1023
1024 #else
1025
1026 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1027                                                 unsigned long arg)
1028 {
1029         return -ENOTTY;
1030 }
1031
1032 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1033
1034 #endif
1035
1036 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1037 {
1038         set_bit(req, &vcpu->requests);
1039 }
1040
1041 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1042 {
1043         if (test_bit(req, &vcpu->requests)) {
1044                 clear_bit(req, &vcpu->requests);
1045                 return true;
1046         } else {
1047                 return false;
1048         }
1049 }
1050
1051 extern bool kvm_rebooting;
1052
1053 struct kvm_device_ops;
1054
1055 struct kvm_device {
1056         struct kvm_device_ops *ops;
1057         struct kvm *kvm;
1058         void *private;
1059         struct list_head vm_node;
1060 };
1061
1062 /* create, destroy, and name are mandatory */
1063 struct kvm_device_ops {
1064         const char *name;
1065         int (*create)(struct kvm_device *dev, u32 type);
1066
1067         /*
1068          * Destroy is responsible for freeing dev.
1069          *
1070          * Destroy may be called before or after destructors are called
1071          * on emulated I/O regions, depending on whether a reference is
1072          * held by a vcpu or other kvm component that gets destroyed
1073          * after the emulated I/O.
1074          */
1075         void (*destroy)(struct kvm_device *dev);
1076
1077         int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1078         int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1079         int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1080         long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1081                       unsigned long arg);
1082 };
1083
1084 void kvm_device_get(struct kvm_device *dev);
1085 void kvm_device_put(struct kvm_device *dev);
1086 struct kvm_device *kvm_device_from_filp(struct file *filp);
1087
1088 extern struct kvm_device_ops kvm_mpic_ops;
1089 extern struct kvm_device_ops kvm_xics_ops;
1090
1091 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1092
1093 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1094 {
1095         vcpu->spin_loop.in_spin_loop = val;
1096 }
1097 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1098 {
1099         vcpu->spin_loop.dy_eligible = val;
1100 }
1101
1102 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1103
1104 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1105 {
1106 }
1107
1108 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1109 {
1110 }
1111
1112 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
1113 {
1114         return true;
1115 }
1116
1117 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1118 #endif
1119