include/linux/kvm_host.h

   1 #ifndef __KVM_HOST_H
   2 #define __KVM_HOST_H
   3
   4 /*
   5  * This work is licensed under the terms of the GNU GPL, version 2.  See
   6  * the COPYING file in the top-level directory.
   7  */
   8
   9 #include <linux/types.h>
  10 #include <linux/hardirq.h>
  11 #include <linux/list.h>
  12 #include <linux/mutex.h>
  13 #include <linux/spinlock.h>
  14 #include <linux/signal.h>
  15 #include <linux/sched.h>
  16 #include <linux/bug.h>
  17 #include <linux/mm.h>
  18 #include <linux/mmu_notifier.h>
  19 #include <linux/preempt.h>
  20 #include <linux/msi.h>
  21 #include <linux/slab.h>
  22 #include <linux/rcupdate.h>
  23 #include <linux/ratelimit.h>
  24 #include <linux/err.h>
  25 #include <asm/signal.h>
  26
  27 #include <linux/kvm.h>
  28 #include <linux/kvm_para.h>
  29
  30 #include <linux/kvm_types.h>
  31
  32 #include <asm/kvm_host.h>
  33
  34 #ifndef KVM_MMIO_SIZE
  35 #define KVM_MMIO_SIZE 8
  36 #endif
  37
  38 /*
  39  * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  40  * in kvm, other bits are visible for userspace which are defined in
  41  * include/linux/kvm_h.
  42  */
  43 #define KVM_MEMSLOT_INVALID     (1UL << 16)
  44
  45 /*
  46  * If we support unaligned MMIO, at most one fragment will be split into two:
  47  */
  48 #ifdef KVM_UNALIGNED_MMIO
  49 #  define KVM_EXTRA_MMIO_FRAGMENTS 1
  50 #else
  51 #  define KVM_EXTRA_MMIO_FRAGMENTS 0
  52 #endif
  53
  54 #define KVM_USER_MMIO_SIZE 8
  55
  56 #define KVM_MAX_MMIO_FRAGMENTS \
  57         (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
  58
  59 /*
  60  * For the normal pfn, the highest 12 bits should be zero,
  61  * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  62  * mask bit 63 to indicate the noslot pfn.
  63  */
  64 #define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  65 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  66 #define KVM_PFN_NOSLOT          (0x1ULL << 63)
  67
  68 #define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  69 #define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  70 #define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  71
  72 /*
  73  * error pfns indicate that the gfn is in slot but faild to
  74  * translate it to pfn on host.
  75  */
  76 static inline bool is_error_pfn(pfn_t pfn)
  77 {
  78         return !!(pfn & KVM_PFN_ERR_MASK);
  79 }
  80
  81 /*
  82  * error_noslot pfns indicate that the gfn can not be
  83  * translated to pfn - it is not in slot or failed to
  84  * translate it to pfn.
  85  */
  86 static inline bool is_error_noslot_pfn(pfn_t pfn)
  87 {
  88         return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
  89 }
  90
  91 /* noslot pfn indicates that the gfn is not in slot. */
  92 static inline bool is_noslot_pfn(pfn_t pfn)
  93 {
  94         return pfn == KVM_PFN_NOSLOT;
  95 }
  96
  97 #define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
  98 #define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
  99
 100 static inline bool kvm_is_error_hva(unsigned long addr)
 101 {
 102         return addr >= PAGE_OFFSET;
 103 }
 104
 105 #define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
 106
 107 static inline bool is_error_page(struct page *page)
 108 {
 109         return IS_ERR(page);
 110 }
 111
 112 /*
 113  * vcpu->requests bit members
 114  */
 115 #define KVM_REQ_TLB_FLUSH          0
 116 #define KVM_REQ_MIGRATE_TIMER      1
 117 #define KVM_REQ_REPORT_TPR_ACCESS  2
 118 #define KVM_REQ_MMU_RELOAD         3
 119 #define KVM_REQ_TRIPLE_FAULT       4
 120 #define KVM_REQ_PENDING_TIMER      5
 121 #define KVM_REQ_UNHALT             6
 122 #define KVM_REQ_MMU_SYNC           7
 123 #define KVM_REQ_CLOCK_UPDATE       8
 124 #define KVM_REQ_KICK               9
 125 #define KVM_REQ_DEACTIVATE_FPU    10
 126 #define KVM_REQ_EVENT             11
 127 #define KVM_REQ_APF_HALT          12
 128 #define KVM_REQ_STEAL_UPDATE      13
 129 #define KVM_REQ_NMI               14
 130 #define KVM_REQ_IMMEDIATE_EXIT    15
 131 #define KVM_REQ_PMU               16
 132 #define KVM_REQ_PMI               17
 133 #define KVM_REQ_WATCHDOG          18
 134
 135 #define KVM_USERSPACE_IRQ_SOURCE_ID             0
 136 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 137
 138 struct kvm;
 139 struct kvm_vcpu;
 140 extern struct kmem_cache *kvm_vcpu_cache;
 141
 142 struct kvm_io_range {
 143         gpa_t addr;
 144         int len;
 145         struct kvm_io_device *dev;
 146 };
 147
 148 #define NR_IOBUS_DEVS 1000
 149
 150 struct kvm_io_bus {
 151         int                   dev_count;
 152         struct kvm_io_range range[];
 153 };
 154
 155 enum kvm_bus {
 156         KVM_MMIO_BUS,
 157         KVM_PIO_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         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 flags;
 279         unsigned long *dirty_bitmap;
 280         struct kvm_arch_memory_slot arch;
 281         unsigned long userspace_addr;
 282         int user_alloc;
 283         int 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         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 __KVM_HAVE_IOAPIC
 307
 308 struct kvm_irq_routing_table {
 309         int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_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_MEM_SLOTS_NUM
 326 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
 327 #endif
 328
 329 /*
 330  * Note:
 331  * memslots are not sorted by id anymore, please use id_to_memslot()
 332  * to get the memslot by its id.
 333  */
 334 struct kvm_memslots {
 335         u64 generation;
 336         struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
 337         /* The mapping table from slot id to the index in memslots[]. */
 338         int id_to_index[KVM_MEM_SLOTS_NUM];
 339 };
 340
 341 struct kvm {
 342         spinlock_t mmu_lock;
 343         struct mutex slots_lock;
 344         struct mm_struct *mm; /* userspace tied to this vm */
 345         struct kvm_memslots *memslots;
 346         struct srcu_struct srcu;
 347 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 348         u32 bsp_vcpu_id;
 349 #endif
 350         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 351         atomic_t online_vcpus;
 352         int last_boosted_vcpu;
 353         struct list_head vm_list;
 354         struct mutex lock;
 355         struct kvm_io_bus *buses[KVM_NR_BUSES];
 356 #ifdef CONFIG_HAVE_KVM_EVENTFD
 357         struct {
 358                 spinlock_t        lock;
 359                 struct list_head  items;
 360                 struct list_head  resampler_list;
 361                 struct mutex      resampler_lock;
 362         } irqfds;
 363         struct list_head ioeventfds;
 364 #endif
 365         struct kvm_vm_stat stat;
 366         struct kvm_arch arch;
 367         atomic_t users_count;
 368 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 369         struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 370         spinlock_t ring_lock;
 371         struct list_head coalesced_zones;
 372 #endif
 373
 374         struct mutex irq_lock;
 375 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 376         /*
 377          * Update side is protected by irq_lock and,
 378          * if configured, irqfds.lock.
 379          */
 380         struct kvm_irq_routing_table __rcu *irq_routing;
 381         struct hlist_head mask_notifier_list;
 382         struct hlist_head irq_ack_notifier_list;
 383 #endif
 384
 385 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 386         struct mmu_notifier mmu_notifier;
 387         unsigned long mmu_notifier_seq;
 388         long mmu_notifier_count;
 389 #endif
 390         long tlbs_dirty;
 391 };
 392
 393 #define kvm_err(fmt, ...) \
 394         pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 395 #define kvm_info(fmt, ...) \
 396         pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 397 #define kvm_debug(fmt, ...) \
 398         pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 399 #define kvm_pr_unimpl(fmt, ...) \
 400         pr_err_ratelimited("kvm [%i]: " fmt, \
 401                            task_tgid_nr(current), ## __VA_ARGS__)
 402
 403 /* The guest did something we don't support. */
 404 #define vcpu_unimpl(vcpu, fmt, ...)                                     \
 405         kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 406
 407 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 408 {
 409         smp_rmb();
 410         return kvm->vcpus[i];
 411 }
 412
 413 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
 414         for (idx = 0; \
 415              idx < atomic_read(&kvm->online_vcpus) && \
 416              (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 417              idx++)
 418
 419 #define kvm_for_each_memslot(memslot, slots)    \
 420         for (memslot = &slots->memslots[0];     \
 421               memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
 422                 memslot++)
 423
 424 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 425 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 426
 427 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
 428 void vcpu_put(struct kvm_vcpu *vcpu);
 429
 430 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 431                   struct module *module);
 432 void kvm_exit(void);
 433
 434 void kvm_get_kvm(struct kvm *kvm);
 435 void kvm_put_kvm(struct kvm *kvm);
 436 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
 437
 438 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 439 {
 440         return rcu_dereference_check(kvm->memslots,
 441                         srcu_read_lock_held(&kvm->srcu)
 442                         || lockdep_is_held(&kvm->slots_lock));
 443 }
 444
 445 static inline struct kvm_memory_slot *
 446 id_to_memslot(struct kvm_memslots *slots, int id)
 447 {
 448         int index = slots->id_to_index[id];
 449         struct kvm_memory_slot *slot;
 450
 451         slot = &slots->memslots[index];
 452
 453         WARN_ON(slot->id != id);
 454         return slot;
 455 }
 456
 457 int kvm_set_memory_region(struct kvm *kvm,
 458                           struct kvm_userspace_memory_region *mem,
 459                           int user_alloc);
 460 int __kvm_set_memory_region(struct kvm *kvm,
 461                             struct kvm_userspace_memory_region *mem,
 462                             int user_alloc);
 463 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 464                            struct kvm_memory_slot *dont);
 465 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
 466 int kvm_arch_prepare_memory_region(struct kvm *kvm,
 467                                 struct kvm_memory_slot *memslot,
 468                                 struct kvm_memory_slot old,
 469                                 struct kvm_userspace_memory_region *mem,
 470                                 int user_alloc);
 471 void kvm_arch_commit_memory_region(struct kvm *kvm,
 472                                 struct kvm_userspace_memory_region *mem,
 473                                 struct kvm_memory_slot old,
 474                                 int user_alloc);
 475 bool kvm_largepages_enabled(void);
 476 void kvm_disable_largepages(void);
 477 /* flush all memory translations */
 478 void kvm_arch_flush_shadow_all(struct kvm *kvm);
 479 /* flush memory translations pointing to 'slot' */
 480 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 481                                    struct kvm_memory_slot *slot);
 482
 483 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
 484                             int nr_pages);
 485
 486 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 487 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 488 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 489 void kvm_release_page_clean(struct page *page);
 490 void kvm_release_page_dirty(struct page *page);
 491 void kvm_set_page_dirty(struct page *page);
 492 void kvm_set_page_accessed(struct page *page);
 493
 494 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
 495 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
 496                        bool write_fault, bool *writable);
 497 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 498 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 499                       bool *writable);
 500 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 501 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 502
 503 void kvm_release_pfn_dirty(pfn_t pfn);
 504 void kvm_release_pfn_clean(pfn_t pfn);
 505 void kvm_set_pfn_dirty(pfn_t pfn);
 506 void kvm_set_pfn_accessed(pfn_t pfn);
 507 void kvm_get_pfn(pfn_t pfn);
 508
 509 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 510                         int len);
 511 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
 512                           unsigned long len);
 513 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 514 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 515                            void *data, unsigned long len);
 516 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 517                          int offset, int len);
 518 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 519                     unsigned long len);
 520 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 521                            void *data, unsigned long len);
 522 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 523                               gpa_t gpa);
 524 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 525 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 526 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 527 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 528 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
 529 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 530 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
 531                              gfn_t gfn);
 532
 533 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 534 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 535 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
 536 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
 537 void kvm_resched(struct kvm_vcpu *vcpu);
 538 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 539 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 540
 541 void kvm_flush_remote_tlbs(struct kvm *kvm);
 542 void kvm_reload_remote_mmus(struct kvm *kvm);
 543
 544 long kvm_arch_dev_ioctl(struct file *filp,
 545                         unsigned int ioctl, unsigned long arg);
 546 long kvm_arch_vcpu_ioctl(struct file *filp,
 547                          unsigned int ioctl, unsigned long arg);
 548 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 549
 550 int kvm_dev_ioctl_check_extension(long ext);
 551
 552 int kvm_get_dirty_log(struct kvm *kvm,
 553                         struct kvm_dirty_log *log, int *is_dirty);
 554 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 555                                 struct kvm_dirty_log *log);
 556
 557 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 558                                    struct
 559                                    kvm_userspace_memory_region *mem,
 560                                    int user_alloc);
 561 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
 562 long kvm_arch_vm_ioctl(struct file *filp,
 563                        unsigned int ioctl, unsigned long arg);
 564
 565 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 566 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 567
 568 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 569                                     struct kvm_translation *tr);
 570
 571 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 572 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 573 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 574                                   struct kvm_sregs *sregs);
 575 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 576                                   struct kvm_sregs *sregs);
 577 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
 578                                     struct kvm_mp_state *mp_state);
 579 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
 580                                     struct kvm_mp_state *mp_state);
 581 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
 582                                         struct kvm_guest_debug *dbg);
 583 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 584
 585 int kvm_arch_init(void *opaque);
 586 void kvm_arch_exit(void);
 587
 588 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 589 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 590
 591 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 592 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 593 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 594 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 595 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 596 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 597
 598 int kvm_arch_hardware_enable(void *garbage);
 599 void kvm_arch_hardware_disable(void *garbage);
 600 int kvm_arch_hardware_setup(void);
 601 void kvm_arch_hardware_unsetup(void);
 602 void kvm_arch_check_processor_compat(void *rtn);
 603 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 604 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 605
 606 void kvm_free_physmem(struct kvm *kvm);
 607
 608 void *kvm_kvzalloc(unsigned long size);
 609 void kvm_kvfree(const void *addr);
 610
 611 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 612 static inline struct kvm *kvm_arch_alloc_vm(void)
 613 {
 614         return kzalloc(sizeof(struct kvm), GFP_KERNEL);
 615 }
 616
 617 static inline void kvm_arch_free_vm(struct kvm *kvm)
 618 {
 619         kfree(kvm);
 620 }
 621 #endif
 622
 623 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 624 {
 625 #ifdef __KVM_HAVE_ARCH_WQP
 626         return vcpu->arch.wqp;
 627 #else
 628         return &vcpu->wq;
 629 #endif
 630 }
 631
 632 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 633 void kvm_arch_destroy_vm(struct kvm *kvm);
 634 void kvm_free_all_assigned_devices(struct kvm *kvm);
 635 void kvm_arch_sync_events(struct kvm *kvm);
 636
 637 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 638 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 639
 640 bool kvm_is_mmio_pfn(pfn_t pfn);
 641
 642 struct kvm_irq_ack_notifier {
 643         struct hlist_node link;
 644         unsigned gsi;
 645         void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
 646 };
 647
 648 struct kvm_assigned_dev_kernel {
 649         struct kvm_irq_ack_notifier ack_notifier;
 650         struct list_head list;
 651         int assigned_dev_id;
 652         int host_segnr;
 653         int host_busnr;
 654         int host_devfn;
 655         unsigned int entries_nr;
 656         int host_irq;
 657         bool host_irq_disabled;
 658         bool pci_2_3;
 659         struct msix_entry *host_msix_entries;
 660         int guest_irq;
 661         struct msix_entry *guest_msix_entries;
 662         unsigned long irq_requested_type;
 663         int irq_source_id;
 664         int flags;
 665         struct pci_dev *dev;
 666         struct kvm *kvm;
 667         spinlock_t intx_lock;
 668         spinlock_t intx_mask_lock;
 669         char irq_name[32];
 670         struct pci_saved_state *pci_saved_state;
 671 };
 672
 673 struct kvm_irq_mask_notifier {
 674         void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
 675         int irq;
 676         struct hlist_node link;
 677 };
 678
 679 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
 680                                     struct kvm_irq_mask_notifier *kimn);
 681 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
 682                                       struct kvm_irq_mask_notifier *kimn);
 683 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
 684                              bool mask);
 685
 686 #ifdef __KVM_HAVE_IOAPIC
 687 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
 688                                    union kvm_ioapic_redirect_entry *entry,
 689                                    unsigned long *deliver_bitmask);
 690 #endif
 691 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
 692 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
 693                 int irq_source_id, int level);
 694 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
 695 void kvm_register_irq_ack_notifier(struct kvm *kvm,
 696                                    struct kvm_irq_ack_notifier *kian);
 697 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
 698                                    struct kvm_irq_ack_notifier *kian);
 699 int kvm_request_irq_source_id(struct kvm *kvm);
 700 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
 701
 702 /* For vcpu->arch.iommu_flags */
 703 #define KVM_IOMMU_CACHE_COHERENCY       0x1
 704
 705 #ifdef CONFIG_IOMMU_API
 706 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 707 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 708 int kvm_iommu_map_guest(struct kvm *kvm);
 709 int kvm_iommu_unmap_guest(struct kvm *kvm);
 710 int kvm_assign_device(struct kvm *kvm,
 711                       struct kvm_assigned_dev_kernel *assigned_dev);
 712 int kvm_deassign_device(struct kvm *kvm,
 713                         struct kvm_assigned_dev_kernel *assigned_dev);
 714 #else /* CONFIG_IOMMU_API */
 715 static inline int kvm_iommu_map_pages(struct kvm *kvm,
 716                                       struct kvm_memory_slot *slot)
 717 {
 718         return 0;
 719 }
 720
 721 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
 722                                          struct kvm_memory_slot *slot)
 723 {
 724 }
 725
 726 static inline int kvm_iommu_map_guest(struct kvm *kvm)
 727 {
 728         return -ENODEV;
 729 }
 730
 731 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
 732 {
 733         return 0;
 734 }
 735
 736 static inline int kvm_assign_device(struct kvm *kvm,
 737                 struct kvm_assigned_dev_kernel *assigned_dev)
 738 {
 739         return 0;
 740 }
 741
 742 static inline int kvm_deassign_device(struct kvm *kvm,
 743                 struct kvm_assigned_dev_kernel *assigned_dev)
 744 {
 745         return 0;
 746 }
 747 #endif /* CONFIG_IOMMU_API */
 748
 749 static inline void kvm_guest_enter(void)
 750 {
 751         BUG_ON(preemptible());
 752         vtime_account(current);
 753         current->flags |= PF_VCPU;
 754         /* KVM does not hold any references to rcu protected data when it
 755          * switches CPU into a guest mode. In fact switching to a guest mode
 756          * is very similar to exiting to userspase from rcu point of view. In
 757          * addition CPU may stay in a guest mode for quite a long time (up to
 758          * one time slice). Lets treat guest mode as quiescent state, just like
 759          * we do with user-mode execution.
 760          */
 761         rcu_virt_note_context_switch(smp_processor_id());
 762 }
 763
 764 static inline void kvm_guest_exit(void)
 765 {
 766         vtime_account(current);
 767         current->flags &= ~PF_VCPU;
 768 }
 769
 770 /*
 771  * search_memslots() and __gfn_to_memslot() are here because they are
 772  * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
 773  * gfn_to_memslot() itself isn't here as an inline because that would
 774  * bloat other code too much.
 775  */
 776 static inline struct kvm_memory_slot *
 777 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
 778 {
 779         struct kvm_memory_slot *memslot;
 780
 781         kvm_for_each_memslot(memslot, slots)
 782                 if (gfn >= memslot->base_gfn &&
 783                       gfn < memslot->base_gfn + memslot->npages)
 784                         return memslot;
 785
 786         return NULL;
 787 }
 788
 789 static inline struct kvm_memory_slot *
 790 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
 791 {
 792         return search_memslots(slots, gfn);
 793 }
 794
 795 static inline unsigned long
 796 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
 797 {
 798         return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
 799 }
 800
 801 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
 802 {
 803         return gfn_to_memslot(kvm, gfn)->id;
 804 }
 805
 806 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
 807 {
 808         /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
 809         return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
 810                 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
 811 }
 812
 813 static inline gfn_t
 814 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
 815 {
 816         gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
 817
 818         return slot->base_gfn + gfn_offset;
 819 }
 820
 821 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 822 {
 823         return (gpa_t)gfn << PAGE_SHIFT;
 824 }
 825
 826 static inline gfn_t gpa_to_gfn(gpa_t gpa)
 827 {
 828         return (gfn_t)(gpa >> PAGE_SHIFT);
 829 }
 830
 831 static inline hpa_t pfn_to_hpa(pfn_t pfn)
 832 {
 833         return (hpa_t)pfn << PAGE_SHIFT;
 834 }
 835
 836 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
 837 {
 838         set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
 839 }
 840
 841 enum kvm_stat_kind {
 842         KVM_STAT_VM,
 843         KVM_STAT_VCPU,
 844 };
 845
 846 struct kvm_stats_debugfs_item {
 847         const char *name;
 848         int offset;
 849         enum kvm_stat_kind kind;
 850         struct dentry *dentry;
 851 };
 852 extern struct kvm_stats_debugfs_item debugfs_entries[];
 853 extern struct dentry *kvm_debugfs_dir;
 854
 855 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 856 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
 857 {
 858         if (unlikely(kvm->mmu_notifier_count))
 859                 return 1;
 860         /*
 861          * Ensure the read of mmu_notifier_count happens before the read
 862          * of mmu_notifier_seq.  This interacts with the smp_wmb() in
 863          * mmu_notifier_invalidate_range_end to make sure that the caller
 864          * either sees the old (non-zero) value of mmu_notifier_count or
 865          * the new (incremented) value of mmu_notifier_seq.
 866          * PowerPC Book3s HV KVM calls this under a per-page lock
 867          * rather than under kvm->mmu_lock, for scalability, so
 868          * can't rely on kvm->mmu_lock to keep things ordered.
 869          */
 870         smp_rmb();
 871         if (kvm->mmu_notifier_seq != mmu_seq)
 872                 return 1;
 873         return 0;
 874 }
 875 #endif
 876
 877 #ifdef KVM_CAP_IRQ_ROUTING
 878
 879 #define KVM_MAX_IRQ_ROUTES 1024
 880
 881 int kvm_setup_default_irq_routing(struct kvm *kvm);
 882 int kvm_set_irq_routing(struct kvm *kvm,
 883                         const struct kvm_irq_routing_entry *entries,
 884                         unsigned nr,
 885                         unsigned flags);
 886 void kvm_free_irq_routing(struct kvm *kvm);
 887
 888 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
 889
 890 #else
 891
 892 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 893
 894 #endif
 895
 896 #ifdef CONFIG_HAVE_KVM_EVENTFD
 897
 898 void kvm_eventfd_init(struct kvm *kvm);
 899 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
 900 void kvm_irqfd_release(struct kvm *kvm);
 901 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
 902 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
 903
 904 #else
 905
 906 static inline void kvm_eventfd_init(struct kvm *kvm) {}
 907
 908 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
 909 {
 910         return -EINVAL;
 911 }
 912
 913 static inline void kvm_irqfd_release(struct kvm *kvm) {}
 914
 915 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 916 static inline void kvm_irq_routing_update(struct kvm *kvm,
 917                                           struct kvm_irq_routing_table *irq_rt)
 918 {
 919         rcu_assign_pointer(kvm->irq_routing, irq_rt);
 920 }
 921 #endif
 922
 923 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 924 {
 925         return -ENOSYS;
 926 }
 927
 928 #endif /* CONFIG_HAVE_KVM_EVENTFD */
 929
 930 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 931 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
 932 {
 933         return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
 934 }
 935
 936 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
 937
 938 #else
 939
 940 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
 941
 942 #endif
 943
 944 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
 945
 946 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 947                                   unsigned long arg);
 948
 949 #else
 950
 951 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
 952                                                 unsigned long arg)
 953 {
 954         return -ENOTTY;
 955 }
 956
 957 #endif
 958
 959 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
 960 {
 961         set_bit(req, &vcpu->requests);
 962 }
 963
 964 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
 965 {
 966         if (test_bit(req, &vcpu->requests)) {
 967                 clear_bit(req, &vcpu->requests);
 968                 return true;
 969         } else {
 970                 return false;
 971         }
 972 }
 973
 974 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 975
 976 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 977 {
 978         vcpu->spin_loop.in_spin_loop = val;
 979 }
 980 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 981 {
 982         vcpu->spin_loop.dy_eligible = val;
 983 }
 984
 985 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
 986
 987 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
 988 {
 989 }
 990
 991 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
 992 {
 993 }
 994
 995 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
 996 {
 997         return true;
 998 }
 999
1000 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1001 #endif
1002