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