linux/include/linux/kvm_host.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0-only */
   2#ifndef __KVM_HOST_H
   3#define __KVM_HOST_H
   4
   5
   6#include <linux/types.h>
   7#include <linux/hardirq.h>
   8#include <linux/list.h>
   9#include <linux/mutex.h>
  10#include <linux/spinlock.h>
  11#include <linux/signal.h>
  12#include <linux/sched.h>
  13#include <linux/sched/stat.h>
  14#include <linux/bug.h>
  15#include <linux/minmax.h>
  16#include <linux/mm.h>
  17#include <linux/mmu_notifier.h>
  18#include <linux/preempt.h>
  19#include <linux/msi.h>
  20#include <linux/slab.h>
  21#include <linux/vmalloc.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 <linux/irqbypass.h>
  28#include <linux/rcuwait.h>
  29#include <linux/refcount.h>
  30#include <linux/nospec.h>
  31#include <linux/notifier.h>
  32#include <asm/signal.h>
  33
  34#include <linux/kvm.h>
  35#include <linux/kvm_para.h>
  36
  37#include <linux/kvm_types.h>
  38
  39#include <asm/kvm_host.h>
  40#include <linux/kvm_dirty_ring.h>
  41
  42#ifndef KVM_MAX_VCPU_ID
  43#define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
  44#endif
  45
  46/*
  47 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
  48 * in kvm, other bits are visible for userspace which are defined in
  49 * include/linux/kvm_h.
  50 */
  51#define KVM_MEMSLOT_INVALID     (1UL << 16)
  52
  53/*
  54 * Bit 63 of the memslot generation number is an "update in-progress flag",
  55 * e.g. is temporarily set for the duration of install_new_memslots().
  56 * This flag effectively creates a unique generation number that is used to
  57 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
  58 * i.e. may (or may not) have come from the previous memslots generation.
  59 *
  60 * This is necessary because the actual memslots update is not atomic with
  61 * respect to the generation number update.  Updating the generation number
  62 * first would allow a vCPU to cache a spte from the old memslots using the
  63 * new generation number, and updating the generation number after switching
  64 * to the new memslots would allow cache hits using the old generation number
  65 * to reference the defunct memslots.
  66 *
  67 * This mechanism is used to prevent getting hits in KVM's caches while a
  68 * memslot update is in-progress, and to prevent cache hits *after* updating
  69 * the actual generation number against accesses that were inserted into the
  70 * cache *before* the memslots were updated.
  71 */
  72#define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS      BIT_ULL(63)
  73
  74/* Two fragments for cross MMIO pages. */
  75#define KVM_MAX_MMIO_FRAGMENTS  2
  76
  77#ifndef KVM_ADDRESS_SPACE_NUM
  78#define KVM_ADDRESS_SPACE_NUM   1
  79#endif
  80
  81/*
  82 * For the normal pfn, the highest 12 bits should be zero,
  83 * so we can mask bit 62 ~ bit 52  to indicate the error pfn,
  84 * mask bit 63 to indicate the noslot pfn.
  85 */
  86#define KVM_PFN_ERR_MASK        (0x7ffULL << 52)
  87#define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
  88#define KVM_PFN_NOSLOT          (0x1ULL << 63)
  89
  90#define KVM_PFN_ERR_FAULT       (KVM_PFN_ERR_MASK)
  91#define KVM_PFN_ERR_HWPOISON    (KVM_PFN_ERR_MASK + 1)
  92#define KVM_PFN_ERR_RO_FAULT    (KVM_PFN_ERR_MASK + 2)
  93
  94/*
  95 * error pfns indicate that the gfn is in slot but faild to
  96 * translate it to pfn on host.
  97 */
  98static inline bool is_error_pfn(kvm_pfn_t pfn)
  99{
 100        return !!(pfn & KVM_PFN_ERR_MASK);
 101}
 102
 103/*
 104 * error_noslot pfns indicate that the gfn can not be
 105 * translated to pfn - it is not in slot or failed to
 106 * translate it to pfn.
 107 */
 108static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
 109{
 110        return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
 111}
 112
 113/* noslot pfn indicates that the gfn is not in slot. */
 114static inline bool is_noslot_pfn(kvm_pfn_t pfn)
 115{
 116        return pfn == KVM_PFN_NOSLOT;
 117}
 118
 119/*
 120 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
 121 * provide own defines and kvm_is_error_hva
 122 */
 123#ifndef KVM_HVA_ERR_BAD
 124
 125#define KVM_HVA_ERR_BAD         (PAGE_OFFSET)
 126#define KVM_HVA_ERR_RO_BAD      (PAGE_OFFSET + PAGE_SIZE)
 127
 128static inline bool kvm_is_error_hva(unsigned long addr)
 129{
 130        return addr >= PAGE_OFFSET;
 131}
 132
 133#endif
 134
 135#define KVM_ERR_PTR_BAD_PAGE    (ERR_PTR(-ENOENT))
 136
 137static inline bool is_error_page(struct page *page)
 138{
 139        return IS_ERR(page);
 140}
 141
 142#define KVM_REQUEST_MASK           GENMASK(7,0)
 143#define KVM_REQUEST_NO_WAKEUP      BIT(8)
 144#define KVM_REQUEST_WAIT           BIT(9)
 145/*
 146 * Architecture-independent vcpu->requests bit members
 147 * Bits 4-7 are reserved for more arch-independent bits.
 148 */
 149#define KVM_REQ_TLB_FLUSH         (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 150#define KVM_REQ_MMU_RELOAD        (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 151#define KVM_REQ_UNBLOCK           2
 152#define KVM_REQ_UNHALT            3
 153#define KVM_REQ_VM_BUGGED         (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 154#define KVM_REQUEST_ARCH_BASE     8
 155
 156#define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
 157        BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
 158        (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
 159})
 160#define KVM_ARCH_REQ(nr)           KVM_ARCH_REQ_FLAGS(nr, 0)
 161
 162bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
 163                                 struct kvm_vcpu *except,
 164                                 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
 165bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
 166bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
 167                                      struct kvm_vcpu *except);
 168bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
 169                                unsigned long *vcpu_bitmap);
 170
 171#define KVM_USERSPACE_IRQ_SOURCE_ID             0
 172#define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID        1
 173
 174extern struct mutex kvm_lock;
 175extern struct list_head vm_list;
 176
 177struct kvm_io_range {
 178        gpa_t addr;
 179        int len;
 180        struct kvm_io_device *dev;
 181};
 182
 183#define NR_IOBUS_DEVS 1000
 184
 185struct kvm_io_bus {
 186        int dev_count;
 187        int ioeventfd_count;
 188        struct kvm_io_range range[];
 189};
 190
 191enum kvm_bus {
 192        KVM_MMIO_BUS,
 193        KVM_PIO_BUS,
 194        KVM_VIRTIO_CCW_NOTIFY_BUS,
 195        KVM_FAST_MMIO_BUS,
 196        KVM_NR_BUSES
 197};
 198
 199int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
 200                     int len, const void *val);
 201int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
 202                            gpa_t addr, int len, const void *val, long cookie);
 203int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
 204                    int len, void *val);
 205int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 206                            int len, struct kvm_io_device *dev);
 207int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 208                              struct kvm_io_device *dev);
 209struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 210                                         gpa_t addr);
 211
 212#ifdef CONFIG_KVM_ASYNC_PF
 213struct kvm_async_pf {
 214        struct work_struct work;
 215        struct list_head link;
 216        struct list_head queue;
 217        struct kvm_vcpu *vcpu;
 218        struct mm_struct *mm;
 219        gpa_t cr2_or_gpa;
 220        unsigned long addr;
 221        struct kvm_arch_async_pf arch;
 222        bool   wakeup_all;
 223        bool notpresent_injected;
 224};
 225
 226void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
 227void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
 228bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
 229                        unsigned long hva, struct kvm_arch_async_pf *arch);
 230int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
 231#endif
 232
 233#ifdef KVM_ARCH_WANT_MMU_NOTIFIER
 234struct kvm_gfn_range {
 235        struct kvm_memory_slot *slot;
 236        gfn_t start;
 237        gfn_t end;
 238        pte_t pte;
 239        bool may_block;
 240};
 241bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
 242bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
 243bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
 244bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
 245#endif
 246
 247enum {
 248        OUTSIDE_GUEST_MODE,
 249        IN_GUEST_MODE,
 250        EXITING_GUEST_MODE,
 251        READING_SHADOW_PAGE_TABLES,
 252};
 253
 254#define KVM_UNMAPPED_PAGE       ((void *) 0x500 + POISON_POINTER_DELTA)
 255
 256struct kvm_host_map {
 257        /*
 258         * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
 259         * a 'struct page' for it. When using mem= kernel parameter some memory
 260         * can be used as guest memory but they are not managed by host
 261         * kernel).
 262         * If 'pfn' is not managed by the host kernel, this field is
 263         * initialized to KVM_UNMAPPED_PAGE.
 264         */
 265        struct page *page;
 266        void *hva;
 267        kvm_pfn_t pfn;
 268        kvm_pfn_t gfn;
 269};
 270
 271/*
 272 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
 273 * directly to check for that.
 274 */
 275static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
 276{
 277        return !!map->hva;
 278}
 279
 280static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
 281{
 282        return single_task_running() && !need_resched() && ktime_before(cur, stop);
 283}
 284
 285/*
 286 * Sometimes a large or cross-page mmio needs to be broken up into separate
 287 * exits for userspace servicing.
 288 */
 289struct kvm_mmio_fragment {
 290        gpa_t gpa;
 291        void *data;
 292        unsigned len;
 293};
 294
 295struct kvm_vcpu {
 296        struct kvm *kvm;
 297#ifdef CONFIG_PREEMPT_NOTIFIERS
 298        struct preempt_notifier preempt_notifier;
 299#endif
 300        int cpu;
 301        int vcpu_id; /* id given by userspace at creation */
 302        int vcpu_idx; /* index in kvm->vcpus array */
 303        int srcu_idx;
 304        int mode;
 305        u64 requests;
 306        unsigned long guest_debug;
 307
 308        int pre_pcpu;
 309        struct list_head blocked_vcpu_list;
 310
 311        struct mutex mutex;
 312        struct kvm_run *run;
 313
 314        struct rcuwait wait;
 315        struct pid __rcu *pid;
 316        int sigset_active;
 317        sigset_t sigset;
 318        unsigned int halt_poll_ns;
 319        bool valid_wakeup;
 320
 321#ifdef CONFIG_HAS_IOMEM
 322        int mmio_needed;
 323        int mmio_read_completed;
 324        int mmio_is_write;
 325        int mmio_cur_fragment;
 326        int mmio_nr_fragments;
 327        struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
 328#endif
 329
 330#ifdef CONFIG_KVM_ASYNC_PF
 331        struct {
 332                u32 queued;
 333                struct list_head queue;
 334                struct list_head done;
 335                spinlock_t lock;
 336        } async_pf;
 337#endif
 338
 339#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 340        /*
 341         * Cpu relax intercept or pause loop exit optimization
 342         * in_spin_loop: set when a vcpu does a pause loop exit
 343         *  or cpu relax intercepted.
 344         * dy_eligible: indicates whether vcpu is eligible for directed yield.
 345         */
 346        struct {
 347                bool in_spin_loop;
 348                bool dy_eligible;
 349        } spin_loop;
 350#endif
 351        bool preempted;
 352        bool ready;
 353        struct kvm_vcpu_arch arch;
 354        struct kvm_vcpu_stat stat;
 355        char stats_id[KVM_STATS_NAME_SIZE];
 356        struct kvm_dirty_ring dirty_ring;
 357
 358        /*
 359         * The index of the most recently used memslot by this vCPU. It's ok
 360         * if this becomes stale due to memslot changes since we always check
 361         * it is a valid slot.
 362         */
 363        int last_used_slot;
 364};
 365
 366/* must be called with irqs disabled */
 367static __always_inline void guest_enter_irqoff(void)
 368{
 369        /*
 370         * This is running in ioctl context so its safe to assume that it's the
 371         * stime pending cputime to flush.
 372         */
 373        instrumentation_begin();
 374        vtime_account_guest_enter();
 375        instrumentation_end();
 376
 377        /*
 378         * KVM does not hold any references to rcu protected data when it
 379         * switches CPU into a guest mode. In fact switching to a guest mode
 380         * is very similar to exiting to userspace from rcu point of view. In
 381         * addition CPU may stay in a guest mode for quite a long time (up to
 382         * one time slice). Lets treat guest mode as quiescent state, just like
 383         * we do with user-mode execution.
 384         */
 385        if (!context_tracking_guest_enter()) {
 386                instrumentation_begin();
 387                rcu_virt_note_context_switch(smp_processor_id());
 388                instrumentation_end();
 389        }
 390}
 391
 392static __always_inline void guest_exit_irqoff(void)
 393{
 394        context_tracking_guest_exit();
 395
 396        instrumentation_begin();
 397        /* Flush the guest cputime we spent on the guest */
 398        vtime_account_guest_exit();
 399        instrumentation_end();
 400}
 401
 402static inline void guest_exit(void)
 403{
 404        unsigned long flags;
 405
 406        local_irq_save(flags);
 407        guest_exit_irqoff();
 408        local_irq_restore(flags);
 409}
 410
 411static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
 412{
 413        /*
 414         * The memory barrier ensures a previous write to vcpu->requests cannot
 415         * be reordered with the read of vcpu->mode.  It pairs with the general
 416         * memory barrier following the write of vcpu->mode in VCPU RUN.
 417         */
 418        smp_mb__before_atomic();
 419        return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
 420}
 421
 422/*
 423 * Some of the bitops functions do not support too long bitmaps.
 424 * This number must be determined not to exceed such limits.
 425 */
 426#define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
 427
 428struct kvm_memory_slot {
 429        gfn_t base_gfn;
 430        unsigned long npages;
 431        unsigned long *dirty_bitmap;
 432        struct kvm_arch_memory_slot arch;
 433        unsigned long userspace_addr;
 434        u32 flags;
 435        short id;
 436        u16 as_id;
 437};
 438
 439static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
 440{
 441        return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
 442}
 443
 444static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
 445{
 446        return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 447}
 448
 449static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
 450{
 451        unsigned long len = kvm_dirty_bitmap_bytes(memslot);
 452
 453        return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
 454}
 455
 456#ifndef KVM_DIRTY_LOG_MANUAL_CAPS
 457#define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
 458#endif
 459
 460struct kvm_s390_adapter_int {
 461        u64 ind_addr;
 462        u64 summary_addr;
 463        u64 ind_offset;
 464        u32 summary_offset;
 465        u32 adapter_id;
 466};
 467
 468struct kvm_hv_sint {
 469        u32 vcpu;
 470        u32 sint;
 471};
 472
 473struct kvm_kernel_irq_routing_entry {
 474        u32 gsi;
 475        u32 type;
 476        int (*set)(struct kvm_kernel_irq_routing_entry *e,
 477                   struct kvm *kvm, int irq_source_id, int level,
 478                   bool line_status);
 479        union {
 480                struct {
 481                        unsigned irqchip;
 482                        unsigned pin;
 483                } irqchip;
 484                struct {
 485                        u32 address_lo;
 486                        u32 address_hi;
 487                        u32 data;
 488                        u32 flags;
 489                        u32 devid;
 490                } msi;
 491                struct kvm_s390_adapter_int adapter;
 492                struct kvm_hv_sint hv_sint;
 493        };
 494        struct hlist_node link;
 495};
 496
 497#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
 498struct kvm_irq_routing_table {
 499        int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
 500        u32 nr_rt_entries;
 501        /*
 502         * Array indexed by gsi. Each entry contains list of irq chips
 503         * the gsi is connected to.
 504         */
 505        struct hlist_head map[];
 506};
 507#endif
 508
 509#ifndef KVM_PRIVATE_MEM_SLOTS
 510#define KVM_PRIVATE_MEM_SLOTS 0
 511#endif
 512
 513#define KVM_MEM_SLOTS_NUM SHRT_MAX
 514#define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
 515
 516#ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
 517static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
 518{
 519        return 0;
 520}
 521#endif
 522
 523/*
 524 * Note:
 525 * memslots are not sorted by id anymore, please use id_to_memslot()
 526 * to get the memslot by its id.
 527 */
 528struct kvm_memslots {
 529        u64 generation;
 530        /* The mapping table from slot id to the index in memslots[]. */
 531        short id_to_index[KVM_MEM_SLOTS_NUM];
 532        atomic_t last_used_slot;
 533        int used_slots;
 534        struct kvm_memory_slot memslots[];
 535};
 536
 537struct kvm {
 538#ifdef KVM_HAVE_MMU_RWLOCK
 539        rwlock_t mmu_lock;
 540#else
 541        spinlock_t mmu_lock;
 542#endif /* KVM_HAVE_MMU_RWLOCK */
 543
 544        struct mutex slots_lock;
 545
 546        /*
 547         * Protects the arch-specific fields of struct kvm_memory_slots in
 548         * use by the VM. To be used under the slots_lock (above) or in a
 549         * kvm->srcu critical section where acquiring the slots_lock would
 550         * lead to deadlock with the synchronize_srcu in
 551         * install_new_memslots.
 552         */
 553        struct mutex slots_arch_lock;
 554        struct mm_struct *mm; /* userspace tied to this vm */
 555        struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
 556        struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 557
 558        /* Used to wait for completion of MMU notifiers.  */
 559        spinlock_t mn_invalidate_lock;
 560        unsigned long mn_active_invalidate_count;
 561        struct rcuwait mn_memslots_update_rcuwait;
 562
 563        /*
 564         * created_vcpus is protected by kvm->lock, and is incremented
 565         * at the beginning of KVM_CREATE_VCPU.  online_vcpus is only
 566         * incremented after storing the kvm_vcpu pointer in vcpus,
 567         * and is accessed atomically.
 568         */
 569        atomic_t online_vcpus;
 570        int created_vcpus;
 571        int last_boosted_vcpu;
 572        struct list_head vm_list;
 573        struct mutex lock;
 574        struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
 575#ifdef CONFIG_HAVE_KVM_EVENTFD
 576        struct {
 577                spinlock_t        lock;
 578                struct list_head  items;
 579                struct list_head  resampler_list;
 580                struct mutex      resampler_lock;
 581        } irqfds;
 582        struct list_head ioeventfds;
 583#endif
 584        struct kvm_vm_stat stat;
 585        struct kvm_arch arch;
 586        refcount_t users_count;
 587#ifdef CONFIG_KVM_MMIO
 588        struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
 589        spinlock_t ring_lock;
 590        struct list_head coalesced_zones;
 591#endif
 592
 593        struct mutex irq_lock;
 594#ifdef CONFIG_HAVE_KVM_IRQCHIP
 595        /*
 596         * Update side is protected by irq_lock.
 597         */
 598        struct kvm_irq_routing_table __rcu *irq_routing;
 599#endif
 600#ifdef CONFIG_HAVE_KVM_IRQFD
 601        struct hlist_head irq_ack_notifier_list;
 602#endif
 603
 604#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 605        struct mmu_notifier mmu_notifier;
 606        unsigned long mmu_notifier_seq;
 607        long mmu_notifier_count;
 608        unsigned long mmu_notifier_range_start;
 609        unsigned long mmu_notifier_range_end;
 610#endif
 611        struct list_head devices;
 612        u64 manual_dirty_log_protect;
 613        struct dentry *debugfs_dentry;
 614        struct kvm_stat_data **debugfs_stat_data;
 615        struct srcu_struct srcu;
 616        struct srcu_struct irq_srcu;
 617        pid_t userspace_pid;
 618        unsigned int max_halt_poll_ns;
 619        u32 dirty_ring_size;
 620        bool vm_bugged;
 621
 622#ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
 623        struct notifier_block pm_notifier;
 624#endif
 625        char stats_id[KVM_STATS_NAME_SIZE];
 626};
 627
 628#define kvm_err(fmt, ...) \
 629        pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 630#define kvm_info(fmt, ...) \
 631        pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 632#define kvm_debug(fmt, ...) \
 633        pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
 634#define kvm_debug_ratelimited(fmt, ...) \
 635        pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
 636                             ## __VA_ARGS__)
 637#define kvm_pr_unimpl(fmt, ...) \
 638        pr_err_ratelimited("kvm [%i]: " fmt, \
 639                           task_tgid_nr(current), ## __VA_ARGS__)
 640
 641/* The guest did something we don't support. */
 642#define vcpu_unimpl(vcpu, fmt, ...)                                     \
 643        kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt,                  \
 644                        (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
 645
 646#define vcpu_debug(vcpu, fmt, ...)                                      \
 647        kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 648#define vcpu_debug_ratelimited(vcpu, fmt, ...)                          \
 649        kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id,           \
 650                              ## __VA_ARGS__)
 651#define vcpu_err(vcpu, fmt, ...)                                        \
 652        kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
 653
 654static inline void kvm_vm_bugged(struct kvm *kvm)
 655{
 656        kvm->vm_bugged = true;
 657        kvm_make_all_cpus_request(kvm, KVM_REQ_VM_BUGGED);
 658}
 659
 660#define KVM_BUG(cond, kvm, fmt...)                              \
 661({                                                              \
 662        int __ret = (cond);                                     \
 663                                                                \
 664        if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt))         \
 665                kvm_vm_bugged(kvm);                             \
 666        unlikely(__ret);                                        \
 667})
 668
 669#define KVM_BUG_ON(cond, kvm)                                   \
 670({                                                              \
 671        int __ret = (cond);                                     \
 672                                                                \
 673        if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged))           \
 674                kvm_vm_bugged(kvm);                             \
 675        unlikely(__ret);                                        \
 676})
 677
 678static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
 679{
 680        return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
 681}
 682
 683static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
 684{
 685        return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
 686                                      lockdep_is_held(&kvm->slots_lock) ||
 687                                      !refcount_read(&kvm->users_count));
 688}
 689
 690static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 691{
 692        int num_vcpus = atomic_read(&kvm->online_vcpus);
 693        i = array_index_nospec(i, num_vcpus);
 694
 695        /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu.  */
 696        smp_rmb();
 697        return kvm->vcpus[i];
 698}
 699
 700#define kvm_for_each_vcpu(idx, vcpup, kvm) \
 701        for (idx = 0; \
 702             idx < atomic_read(&kvm->online_vcpus) && \
 703             (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 704             idx++)
 705
 706static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 707{
 708        struct kvm_vcpu *vcpu = NULL;
 709        int i;
 710
 711        if (id < 0)
 712                return NULL;
 713        if (id < KVM_MAX_VCPUS)
 714                vcpu = kvm_get_vcpu(kvm, id);
 715        if (vcpu && vcpu->vcpu_id == id)
 716                return vcpu;
 717        kvm_for_each_vcpu(i, vcpu, kvm)
 718                if (vcpu->vcpu_id == id)
 719                        return vcpu;
 720        return NULL;
 721}
 722
 723#define kvm_for_each_memslot(memslot, slots)                            \
 724        for (memslot = &slots->memslots[0];                             \
 725             memslot < slots->memslots + slots->used_slots; memslot++)  \
 726                if (WARN_ON_ONCE(!memslot->npages)) {                   \
 727                } else
 728
 729void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
 730
 731void vcpu_load(struct kvm_vcpu *vcpu);
 732void vcpu_put(struct kvm_vcpu *vcpu);
 733
 734#ifdef __KVM_HAVE_IOAPIC
 735void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
 736void kvm_arch_post_irq_routing_update(struct kvm *kvm);
 737#else
 738static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
 739{
 740}
 741static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
 742{
 743}
 744#endif
 745
 746#ifdef CONFIG_HAVE_KVM_IRQFD
 747int kvm_irqfd_init(void);
 748void kvm_irqfd_exit(void);
 749#else
 750static inline int kvm_irqfd_init(void)
 751{
 752        return 0;
 753}
 754
 755static inline void kvm_irqfd_exit(void)
 756{
 757}
 758#endif
 759int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 760                  struct module *module);
 761void kvm_exit(void);
 762
 763void kvm_get_kvm(struct kvm *kvm);
 764bool kvm_get_kvm_safe(struct kvm *kvm);
 765void kvm_put_kvm(struct kvm *kvm);
 766bool file_is_kvm(struct file *file);
 767void kvm_put_kvm_no_destroy(struct kvm *kvm);
 768
 769static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
 770{
 771        as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
 772        return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
 773                        lockdep_is_held(&kvm->slots_lock) ||
 774                        !refcount_read(&kvm->users_count));
 775}
 776
 777static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
 778{
 779        return __kvm_memslots(kvm, 0);
 780}
 781
 782static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
 783{
 784        int as_id = kvm_arch_vcpu_memslots_id(vcpu);
 785
 786        return __kvm_memslots(vcpu->kvm, as_id);
 787}
 788
 789static inline
 790struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
 791{
 792        int index = slots->id_to_index[id];
 793        struct kvm_memory_slot *slot;
 794
 795        if (index < 0)
 796                return NULL;
 797
 798        slot = &slots->memslots[index];
 799
 800        WARN_ON(slot->id != id);
 801        return slot;
 802}
 803
 804/*
 805 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
 806 * - create a new memory slot
 807 * - delete an existing memory slot
 808 * - modify an existing memory slot
 809 *   -- move it in the guest physical memory space
 810 *   -- just change its flags
 811 *
 812 * Since flags can be changed by some of these operations, the following
 813 * differentiation is the best we can do for __kvm_set_memory_region():
 814 */
 815enum kvm_mr_change {
 816        KVM_MR_CREATE,
 817        KVM_MR_DELETE,
 818        KVM_MR_MOVE,
 819        KVM_MR_FLAGS_ONLY,
 820};
 821
 822int kvm_set_memory_region(struct kvm *kvm,
 823                          const struct kvm_userspace_memory_region *mem);
 824int __kvm_set_memory_region(struct kvm *kvm,
 825                            const struct kvm_userspace_memory_region *mem);
 826void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
 827void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
 828int kvm_arch_prepare_memory_region(struct kvm *kvm,
 829                                struct kvm_memory_slot *memslot,
 830                                const struct kvm_userspace_memory_region *mem,
 831                                enum kvm_mr_change change);
 832void kvm_arch_commit_memory_region(struct kvm *kvm,
 833                                const struct kvm_userspace_memory_region *mem,
 834                                struct kvm_memory_slot *old,
 835                                const struct kvm_memory_slot *new,
 836                                enum kvm_mr_change change);
 837/* flush all memory translations */
 838void kvm_arch_flush_shadow_all(struct kvm *kvm);
 839/* flush memory translations pointing to 'slot' */
 840void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 841                                   struct kvm_memory_slot *slot);
 842
 843int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
 844                            struct page **pages, int nr_pages);
 845
 846struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 847unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 848unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
 849unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 850unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
 851                                      bool *writable);
 852void kvm_release_page_clean(struct page *page);
 853void kvm_release_page_dirty(struct page *page);
 854void kvm_set_page_accessed(struct page *page);
 855
 856kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 857kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
 858                      bool *writable);
 859kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
 860kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
 861kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
 862                               bool atomic, bool *async, bool write_fault,
 863                               bool *writable, hva_t *hva);
 864
 865void kvm_release_pfn_clean(kvm_pfn_t pfn);
 866void kvm_release_pfn_dirty(kvm_pfn_t pfn);
 867void kvm_set_pfn_dirty(kvm_pfn_t pfn);
 868void kvm_set_pfn_accessed(kvm_pfn_t pfn);
 869
 870void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
 871int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 872                        int len);
 873int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 874int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 875                           void *data, unsigned long len);
 876int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 877                                 void *data, unsigned int offset,
 878                                 unsigned long len);
 879int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 880                         int offset, int len);
 881int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 882                    unsigned long len);
 883int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 884                           void *data, unsigned long len);
 885int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 886                                  void *data, unsigned int offset,
 887                                  unsigned long len);
 888int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 889                              gpa_t gpa, unsigned long len);
 890
 891#define __kvm_get_guest(kvm, gfn, offset, v)                            \
 892({                                                                      \
 893        unsigned long __addr = gfn_to_hva(kvm, gfn);                    \
 894        typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
 895        int __ret = -EFAULT;                                            \
 896                                                                        \
 897        if (!kvm_is_error_hva(__addr))                                  \
 898                __ret = get_user(v, __uaddr);                           \
 899        __ret;                                                          \
 900})
 901
 902#define kvm_get_guest(kvm, gpa, v)                                      \
 903({                                                                      \
 904        gpa_t __gpa = gpa;                                              \
 905        struct kvm *__kvm = kvm;                                        \
 906                                                                        \
 907        __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT,                     \
 908                        offset_in_page(__gpa), v);                      \
 909})
 910
 911#define __kvm_put_guest(kvm, gfn, offset, v)                            \
 912({                                                                      \
 913        unsigned long __addr = gfn_to_hva(kvm, gfn);                    \
 914        typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
 915        int __ret = -EFAULT;                                            \
 916                                                                        \
 917        if (!kvm_is_error_hva(__addr))                                  \
 918                __ret = put_user(v, __uaddr);                           \
 919        if (!__ret)                                                     \
 920                mark_page_dirty(kvm, gfn);                              \
 921        __ret;                                                          \
 922})
 923
 924#define kvm_put_guest(kvm, gpa, v)                                      \
 925({                                                                      \
 926        gpa_t __gpa = gpa;                                              \
 927        struct kvm *__kvm = kvm;                                        \
 928                                                                        \
 929        __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT,                     \
 930                        offset_in_page(__gpa), v);                      \
 931})
 932
 933int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 934struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 935bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 936bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
 937unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
 938void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
 939void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 940
 941struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
 942struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
 943kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
 944kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
 945int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
 946int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
 947                struct gfn_to_pfn_cache *cache, bool atomic);
 948struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
 949void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
 950int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
 951                  struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
 952unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
 953unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
 954int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
 955                             int len);
 956int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
 957                               unsigned long len);
 958int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
 959                        unsigned long len);
 960int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
 961                              int offset, int len);
 962int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
 963                         unsigned long len);
 964void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
 965
 966void kvm_sigset_activate(struct kvm_vcpu *vcpu);
 967void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
 968
 969void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 970void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
 971void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
 972bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
 973void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
 974int kvm_vcpu_yield_to(struct kvm_vcpu *target);
 975void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
 976
 977void kvm_flush_remote_tlbs(struct kvm *kvm);
 978void kvm_reload_remote_mmus(struct kvm *kvm);
 979
 980#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
 981int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
 982int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
 983void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
 984void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
 985#endif
 986
 987void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
 988                                   unsigned long end);
 989void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
 990                                   unsigned long end);
 991
 992long kvm_arch_dev_ioctl(struct file *filp,
 993                        unsigned int ioctl, unsigned long arg);
 994long kvm_arch_vcpu_ioctl(struct file *filp,
 995                         unsigned int ioctl, unsigned long arg);
 996vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
 997
 998int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
 999
1000void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1001                                        struct kvm_memory_slot *slot,
1002                                        gfn_t gfn_offset,
1003                                        unsigned long mask);
1004void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1005
1006#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1007void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1008                                        const struct kvm_memory_slot *memslot);
1009#else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1010int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1011int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1012                      int *is_dirty, struct kvm_memory_slot **memslot);
1013#endif
1014
1015int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1016                        bool line_status);
1017int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1018                            struct kvm_enable_cap *cap);
1019long kvm_arch_vm_ioctl(struct file *filp,
1020                       unsigned int ioctl, unsigned long arg);
1021
1022int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1023int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1024
1025int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1026                                    struct kvm_translation *tr);
1027
1028int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1029int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1030int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1031                                  struct kvm_sregs *sregs);
1032int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1033                                  struct kvm_sregs *sregs);
1034int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1035                                    struct kvm_mp_state *mp_state);
1036int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1037                                    struct kvm_mp_state *mp_state);
1038int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1039                                        struct kvm_guest_debug *dbg);
1040int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1041
1042int kvm_arch_init(void *opaque);
1043void kvm_arch_exit(void);
1044
1045void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1046
1047void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1048void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1049int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1050int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1051void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1052void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1053
1054#ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1055int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1056#endif
1057
1058#ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1059void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1060#endif
1061
1062int kvm_arch_hardware_enable(void);
1063void kvm_arch_hardware_disable(void);
1064int kvm_arch_hardware_setup(void *opaque);
1065void kvm_arch_hardware_unsetup(void);
1066int kvm_arch_check_processor_compat(void *opaque);
1067int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1068bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1069int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1070bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1071bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1072int kvm_arch_post_init_vm(struct kvm *kvm);
1073void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1074int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1075
1076#ifndef __KVM_HAVE_ARCH_VM_ALLOC
1077/*
1078 * All architectures that want to use vzalloc currently also
1079 * need their own kvm_arch_alloc_vm implementation.
1080 */
1081static inline struct kvm *kvm_arch_alloc_vm(void)
1082{
1083        return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1084}
1085
1086static inline void kvm_arch_free_vm(struct kvm *kvm)
1087{
1088        kfree(kvm);
1089}
1090#endif
1091
1092#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1093static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1094{
1095        return -ENOTSUPP;
1096}
1097#endif
1098
1099#ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1100void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1101void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1102bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1103#else
1104static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1105{
1106}
1107
1108static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1109{
1110}
1111
1112static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1113{
1114        return false;
1115}
1116#endif
1117#ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1118void kvm_arch_start_assignment(struct kvm *kvm);
1119void kvm_arch_end_assignment(struct kvm *kvm);
1120bool kvm_arch_has_assigned_device(struct kvm *kvm);
1121#else
1122static inline void kvm_arch_start_assignment(struct kvm *kvm)
1123{
1124}
1125
1126static inline void kvm_arch_end_assignment(struct kvm *kvm)
1127{
1128}
1129
1130static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1131{
1132        return false;
1133}
1134#endif
1135
1136static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1137{
1138#ifdef __KVM_HAVE_ARCH_WQP
1139        return vcpu->arch.waitp;
1140#else
1141        return &vcpu->wait;
1142#endif
1143}
1144
1145#ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1146/*
1147 * returns true if the virtual interrupt controller is initialized and
1148 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1149 * controller is dynamically instantiated and this is not always true.
1150 */
1151bool kvm_arch_intc_initialized(struct kvm *kvm);
1152#else
1153static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1154{
1155        return true;
1156}
1157#endif
1158
1159int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1160void kvm_arch_destroy_vm(struct kvm *kvm);
1161void kvm_arch_sync_events(struct kvm *kvm);
1162
1163int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1164
1165bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1166bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1167bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1168
1169struct kvm_irq_ack_notifier {
1170        struct hlist_node link;
1171        unsigned gsi;
1172        void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1173};
1174
1175int kvm_irq_map_gsi(struct kvm *kvm,
1176                    struct kvm_kernel_irq_routing_entry *entries, int gsi);
1177int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1178
1179int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1180                bool line_status);
1181int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1182                int irq_source_id, int level, bool line_status);
1183int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1184                               struct kvm *kvm, int irq_source_id,
1185                               int level, bool line_status);
1186bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1187void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1188void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1189void kvm_register_irq_ack_notifier(struct kvm *kvm,
1190                                   struct kvm_irq_ack_notifier *kian);
1191void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1192                                   struct kvm_irq_ack_notifier *kian);
1193int kvm_request_irq_source_id(struct kvm *kvm);
1194void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1195bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1196
1197/*
1198 * Returns a pointer to the memslot at slot_index if it contains gfn.
1199 * Otherwise returns NULL.
1200 */
1201static inline struct kvm_memory_slot *
1202try_get_memslot(struct kvm_memslots *slots, int slot_index, gfn_t gfn)
1203{
1204        struct kvm_memory_slot *slot;
1205
1206        if (slot_index < 0 || slot_index >= slots->used_slots)
1207                return NULL;
1208
1209        /*
1210         * slot_index can come from vcpu->last_used_slot which is not kept
1211         * in sync with userspace-controllable memslot deletion. So use nospec
1212         * to prevent the CPU from speculating past the end of memslots[].
1213         */
1214        slot_index = array_index_nospec(slot_index, slots->used_slots);
1215        slot = &slots->memslots[slot_index];
1216
1217        if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1218                return slot;
1219        else
1220                return NULL;
1221}
1222
1223/*
1224 * Returns a pointer to the memslot that contains gfn and records the index of
1225 * the slot in index. Otherwise returns NULL.
1226 *
1227 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1228 */
1229static inline struct kvm_memory_slot *
1230search_memslots(struct kvm_memslots *slots, gfn_t gfn, int *index)
1231{
1232        int start = 0, end = slots->used_slots;
1233        struct kvm_memory_slot *memslots = slots->memslots;
1234        struct kvm_memory_slot *slot;
1235
1236        if (unlikely(!slots->used_slots))
1237                return NULL;
1238
1239        while (start < end) {
1240                int slot = start + (end - start) / 2;
1241
1242                if (gfn >= memslots[slot].base_gfn)
1243                        end = slot;
1244                else
1245                        start = slot + 1;
1246        }
1247
1248        slot = try_get_memslot(slots, start, gfn);
1249        if (slot) {
1250                *index = start;
1251                return slot;
1252        }
1253
1254        return NULL;
1255}
1256
1257/*
1258 * __gfn_to_memslot() and its descendants are here because it is called from
1259 * non-modular code in arch/powerpc/kvm/book3s_64_vio{,_hv}.c. gfn_to_memslot()
1260 * itself isn't here as an inline because that would bloat other code too much.
1261 */
1262static inline struct kvm_memory_slot *
1263__gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1264{
1265        struct kvm_memory_slot *slot;
1266        int slot_index = atomic_read(&slots->last_used_slot);
1267
1268        slot = try_get_memslot(slots, slot_index, gfn);
1269        if (slot)
1270                return slot;
1271
1272        slot = search_memslots(slots, gfn, &slot_index);
1273        if (slot) {
1274                atomic_set(&slots->last_used_slot, slot_index);
1275                return slot;
1276        }
1277
1278        return NULL;
1279}
1280
1281static inline unsigned long
1282__gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1283{
1284        /*
1285         * The index was checked originally in search_memslots.  To avoid
1286         * that a malicious guest builds a Spectre gadget out of e.g. page
1287         * table walks, do not let the processor speculate loads outside
1288         * the guest's registered memslots.
1289         */
1290        unsigned long offset = gfn - slot->base_gfn;
1291        offset = array_index_nospec(offset, slot->npages);
1292        return slot->userspace_addr + offset * PAGE_SIZE;
1293}
1294
1295static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1296{
1297        return gfn_to_memslot(kvm, gfn)->id;
1298}
1299
1300static inline gfn_t
1301hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1302{
1303        gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1304
1305        return slot->base_gfn + gfn_offset;
1306}
1307
1308static inline gpa_t gfn_to_gpa(gfn_t gfn)
1309{
1310        return (gpa_t)gfn << PAGE_SHIFT;
1311}
1312
1313static inline gfn_t gpa_to_gfn(gpa_t gpa)
1314{
1315        return (gfn_t)(gpa >> PAGE_SHIFT);
1316}
1317
1318static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1319{
1320        return (hpa_t)pfn << PAGE_SHIFT;
1321}
1322
1323static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1324                                                gpa_t gpa)
1325{
1326        return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1327}
1328
1329static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1330{
1331        unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1332
1333        return kvm_is_error_hva(hva);
1334}
1335
1336enum kvm_stat_kind {
1337        KVM_STAT_VM,
1338        KVM_STAT_VCPU,
1339};
1340
1341struct kvm_stat_data {
1342        struct kvm *kvm;
1343        const struct _kvm_stats_desc *desc;
1344        enum kvm_stat_kind kind;
1345};
1346
1347struct _kvm_stats_desc {
1348        struct kvm_stats_desc desc;
1349        char name[KVM_STATS_NAME_SIZE];
1350};
1351
1352#define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz)                      \
1353        .flags = type | unit | base |                                          \
1354                 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) |              \
1355                 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) |              \
1356                 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK),               \
1357        .exponent = exp,                                                       \
1358        .size = sz,                                                            \
1359        .bucket_size = bsz
1360
1361#define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz)            \
1362        {                                                                      \
1363                {                                                              \
1364                        STATS_DESC_COMMON(type, unit, base, exp, sz, bsz),     \
1365                        .offset = offsetof(struct kvm_vm_stat, generic.stat)   \
1366                },                                                             \
1367                .name = #stat,                                                 \
1368        }
1369#define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz)          \
1370        {                                                                      \
1371                {                                                              \
1372                        STATS_DESC_COMMON(type, unit, base, exp, sz, bsz),     \
1373                        .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1374                },                                                             \
1375                .name = #stat,                                                 \
1376        }
1377#define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz)                    \
1378        {                                                                      \
1379                {                                                              \
1380                        STATS_DESC_COMMON(type, unit, base, exp, sz, bsz),     \
1381                        .offset = offsetof(struct kvm_vm_stat, stat)           \
1382                },                                                             \
1383                .name = #stat,                                                 \
1384        }
1385#define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz)                  \
1386        {                                                                      \
1387                {                                                              \
1388                        STATS_DESC_COMMON(type, unit, base, exp, sz, bsz),     \
1389                        .offset = offsetof(struct kvm_vcpu_stat, stat)         \
1390                },                                                             \
1391                .name = #stat,                                                 \
1392        }
1393/* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1394#define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz)                \
1395        SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1396
1397#define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent)               \
1398        STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE,                     \
1399                unit, base, exponent, 1, 0)
1400#define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent)                  \
1401        STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT,                        \
1402                unit, base, exponent, 1, 0)
1403#define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent)                     \
1404        STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK,                           \
1405                unit, base, exponent, 1, 0)
1406#define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz)     \
1407        STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST,                    \
1408                unit, base, exponent, sz, bsz)
1409#define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz)             \
1410        STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST,                       \
1411                unit, base, exponent, sz, 0)
1412
1413/* Cumulative counter, read/write */
1414#define STATS_DESC_COUNTER(SCOPE, name)                                        \
1415        STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE,                \
1416                KVM_STATS_BASE_POW10, 0)
1417/* Instantaneous counter, read only */
1418#define STATS_DESC_ICOUNTER(SCOPE, name)                                       \
1419        STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE,                   \
1420                KVM_STATS_BASE_POW10, 0)
1421/* Peak counter, read/write */
1422#define STATS_DESC_PCOUNTER(SCOPE, name)                                       \
1423        STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE,                      \
1424                KVM_STATS_BASE_POW10, 0)
1425
1426/* Cumulative time in nanosecond */
1427#define STATS_DESC_TIME_NSEC(SCOPE, name)                                      \
1428        STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS,             \
1429                KVM_STATS_BASE_POW10, -9)
1430/* Linear histogram for time in nanosecond */
1431#define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz)                     \
1432        STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS,            \
1433                KVM_STATS_BASE_POW10, -9, sz, bsz)
1434/* Logarithmic histogram for time in nanosecond */
1435#define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz)                          \
1436        STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS,               \
1437                KVM_STATS_BASE_POW10, -9, sz)
1438
1439#define KVM_GENERIC_VM_STATS()                                                 \
1440        STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush),                      \
1441        STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1442
1443#define KVM_GENERIC_VCPU_STATS()                                               \
1444        STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll),                \
1445        STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll),                 \
1446        STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid),                   \
1447        STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup),                         \
1448        STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns),              \
1449        STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns),                 \
1450        STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns),                      \
1451        STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist,     \
1452                        HALT_POLL_HIST_COUNT),                                 \
1453        STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist,        \
1454                        HALT_POLL_HIST_COUNT),                                 \
1455        STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist,             \
1456                        HALT_POLL_HIST_COUNT)
1457
1458extern struct dentry *kvm_debugfs_dir;
1459
1460ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1461                       const struct _kvm_stats_desc *desc,
1462                       void *stats, size_t size_stats,
1463                       char __user *user_buffer, size_t size, loff_t *offset);
1464
1465/**
1466 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1467 * statistics data.
1468 *
1469 * @data: start address of the stats data
1470 * @size: the number of bucket of the stats data
1471 * @value: the new value used to update the linear histogram's bucket
1472 * @bucket_size: the size (width) of a bucket
1473 */
1474static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1475                                                u64 value, size_t bucket_size)
1476{
1477        size_t index = div64_u64(value, bucket_size);
1478
1479        index = min(index, size - 1);
1480        ++data[index];
1481}
1482
1483/**
1484 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1485 * statistics data.
1486 *
1487 * @data: start address of the stats data
1488 * @size: the number of bucket of the stats data
1489 * @value: the new value used to update the logarithmic histogram's bucket
1490 */
1491static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1492{
1493        size_t index = fls64(value);
1494
1495        index = min(index, size - 1);
1496        ++data[index];
1497}
1498
1499#define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize)                      \
1500        kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1501#define KVM_STATS_LOG_HIST_UPDATE(array, value)                                \
1502        kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1503
1504
1505extern const struct kvm_stats_header kvm_vm_stats_header;
1506extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1507extern const struct kvm_stats_header kvm_vcpu_stats_header;
1508extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1509
1510#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1511static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1512{
1513        if (unlikely(kvm->mmu_notifier_count))
1514                return 1;
1515        /*
1516         * Ensure the read of mmu_notifier_count happens before the read
1517         * of mmu_notifier_seq.  This interacts with the smp_wmb() in
1518         * mmu_notifier_invalidate_range_end to make sure that the caller
1519         * either sees the old (non-zero) value of mmu_notifier_count or
1520         * the new (incremented) value of mmu_notifier_seq.
1521         * PowerPC Book3s HV KVM calls this under a per-page lock
1522         * rather than under kvm->mmu_lock, for scalability, so
1523         * can't rely on kvm->mmu_lock to keep things ordered.
1524         */
1525        smp_rmb();
1526        if (kvm->mmu_notifier_seq != mmu_seq)
1527                return 1;
1528        return 0;
1529}
1530
1531static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1532                                         unsigned long mmu_seq,
1533                                         unsigned long hva)
1534{
1535        lockdep_assert_held(&kvm->mmu_lock);
1536        /*
1537         * If mmu_notifier_count is non-zero, then the range maintained by
1538         * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1539         * might be being invalidated. Note that it may include some false
1540         * positives, due to shortcuts when handing concurrent invalidations.
1541         */
1542        if (unlikely(kvm->mmu_notifier_count) &&
1543            hva >= kvm->mmu_notifier_range_start &&
1544            hva < kvm->mmu_notifier_range_end)
1545                return 1;
1546        if (kvm->mmu_notifier_seq != mmu_seq)
1547                return 1;
1548        return 0;
1549}
1550#endif
1551
1552#ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1553
1554#define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1555
1556bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1557int kvm_set_irq_routing(struct kvm *kvm,
1558                        const struct kvm_irq_routing_entry *entries,
1559                        unsigned nr,
1560                        unsigned flags);
1561int kvm_set_routing_entry(struct kvm *kvm,
1562                          struct kvm_kernel_irq_routing_entry *e,
1563                          const struct kvm_irq_routing_entry *ue);
1564void kvm_free_irq_routing(struct kvm *kvm);
1565
1566#else
1567
1568static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1569
1570#endif
1571
1572int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1573
1574#ifdef CONFIG_HAVE_KVM_EVENTFD
1575
1576void kvm_eventfd_init(struct kvm *kvm);
1577int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1578
1579#ifdef CONFIG_HAVE_KVM_IRQFD
1580int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1581void kvm_irqfd_release(struct kvm *kvm);
1582void kvm_irq_routing_update(struct kvm *);
1583#else
1584static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1585{
1586        return -EINVAL;
1587}
1588
1589static inline void kvm_irqfd_release(struct kvm *kvm) {}
1590#endif
1591
1592#else
1593
1594static inline void kvm_eventfd_init(struct kvm *kvm) {}
1595
1596static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1597{
1598        return -EINVAL;
1599}
1600
1601static inline void kvm_irqfd_release(struct kvm *kvm) {}
1602
1603#ifdef CONFIG_HAVE_KVM_IRQCHIP
1604static inline void kvm_irq_routing_update(struct kvm *kvm)
1605{
1606}
1607#endif
1608
1609static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1610{
1611        return -ENOSYS;
1612}
1613
1614#endif /* CONFIG_HAVE_KVM_EVENTFD */
1615
1616void kvm_arch_irq_routing_update(struct kvm *kvm);
1617
1618static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1619{
1620        /*
1621         * Ensure the rest of the request is published to kvm_check_request's
1622         * caller.  Paired with the smp_mb__after_atomic in kvm_check_request.
1623         */
1624        smp_wmb();
1625        set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1626}
1627
1628static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1629{
1630        return READ_ONCE(vcpu->requests);
1631}
1632
1633static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1634{
1635        return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1636}
1637
1638static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1639{
1640        clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1641}
1642
1643static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1644{
1645        if (kvm_test_request(req, vcpu)) {
1646                kvm_clear_request(req, vcpu);
1647
1648                /*
1649                 * Ensure the rest of the request is visible to kvm_check_request's
1650                 * caller.  Paired with the smp_wmb in kvm_make_request.
1651                 */
1652                smp_mb__after_atomic();
1653                return true;
1654        } else {
1655                return false;
1656        }
1657}
1658
1659extern bool kvm_rebooting;
1660
1661extern unsigned int halt_poll_ns;
1662extern unsigned int halt_poll_ns_grow;
1663extern unsigned int halt_poll_ns_grow_start;
1664extern unsigned int halt_poll_ns_shrink;
1665
1666struct kvm_device {
1667        const struct kvm_device_ops *ops;
1668        struct kvm *kvm;
1669        void *private;
1670        struct list_head vm_node;
1671};
1672
1673/* create, destroy, and name are mandatory */
1674struct kvm_device_ops {
1675        const char *name;
1676
1677        /*
1678         * create is called holding kvm->lock and any operations not suitable
1679         * to do while holding the lock should be deferred to init (see
1680         * below).
1681         */
1682        int (*create)(struct kvm_device *dev, u32 type);
1683
1684        /*
1685         * init is called after create if create is successful and is called
1686         * outside of holding kvm->lock.
1687         */
1688        void (*init)(struct kvm_device *dev);
1689
1690        /*
1691         * Destroy is responsible for freeing dev.
1692         *
1693         * Destroy may be called before or after destructors are called
1694         * on emulated I/O regions, depending on whether a reference is
1695         * held by a vcpu or other kvm component that gets destroyed
1696         * after the emulated I/O.
1697         */
1698        void (*destroy)(struct kvm_device *dev);
1699
1700        /*
1701         * Release is an alternative method to free the device. It is
1702         * called when the device file descriptor is closed. Once
1703         * release is called, the destroy method will not be called
1704         * anymore as the device is removed from the device list of
1705         * the VM. kvm->lock is held.
1706         */
1707        void (*release)(struct kvm_device *dev);
1708
1709        int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1710        int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1711        int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1712        long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1713                      unsigned long arg);
1714        int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1715};
1716
1717void kvm_device_get(struct kvm_device *dev);
1718void kvm_device_put(struct kvm_device *dev);
1719struct kvm_device *kvm_device_from_filp(struct file *filp);
1720int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1721void kvm_unregister_device_ops(u32 type);
1722
1723extern struct kvm_device_ops kvm_mpic_ops;
1724extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1725extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1726
1727#ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1728
1729static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1730{
1731        vcpu->spin_loop.in_spin_loop = val;
1732}
1733static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1734{
1735        vcpu->spin_loop.dy_eligible = val;
1736}
1737
1738#else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1739
1740static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1741{
1742}
1743
1744static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1745{
1746}
1747#endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1748
1749static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1750{
1751        return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1752                !(memslot->flags & KVM_MEMSLOT_INVALID));
1753}
1754
1755struct kvm_vcpu *kvm_get_running_vcpu(void);
1756struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1757
1758#ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1759bool kvm_arch_has_irq_bypass(void);
1760int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1761                           struct irq_bypass_producer *);
1762void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1763                           struct irq_bypass_producer *);
1764void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1765void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1766int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1767                                  uint32_t guest_irq, bool set);
1768#endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1769
1770#ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1771/* If we wakeup during the poll time, was it a sucessful poll? */
1772static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1773{
1774        return vcpu->valid_wakeup;
1775}
1776
1777#else
1778static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1779{
1780        return true;
1781}
1782#endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1783
1784#ifdef CONFIG_HAVE_KVM_NO_POLL
1785/* Callback that tells if we must not poll */
1786bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1787#else
1788static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1789{
1790        return false;
1791}
1792#endif /* CONFIG_HAVE_KVM_NO_POLL */
1793
1794#ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1795long kvm_arch_vcpu_async_ioctl(struct file *filp,
1796                               unsigned int ioctl, unsigned long arg);
1797#else
1798static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1799                                             unsigned int ioctl,
1800                                             unsigned long arg)
1801{
1802        return -ENOIOCTLCMD;
1803}
1804#endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1805
1806void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1807                                            unsigned long start, unsigned long end);
1808
1809#ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1810int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1811#else
1812static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1813{
1814        return 0;
1815}
1816#endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1817
1818typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1819
1820int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1821                                uintptr_t data, const char *name,
1822                                struct task_struct **thread_ptr);
1823
1824#ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1825static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1826{
1827        vcpu->run->exit_reason = KVM_EXIT_INTR;
1828        vcpu->stat.signal_exits++;
1829}
1830#endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1831
1832/*
1833 * This defines how many reserved entries we want to keep before we
1834 * kick the vcpu to the userspace to avoid dirty ring full.  This
1835 * value can be tuned to higher if e.g. PML is enabled on the host.
1836 */
1837#define  KVM_DIRTY_RING_RSVD_ENTRIES  64
1838
1839/* Max number of entries allowed for each kvm dirty ring */
1840#define  KVM_DIRTY_RING_MAX_ENTRIES  65536
1841
1842#endif
1843