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