drivers/kvm/kvm.h

   1 #ifndef __KVM_H
   2 #define __KVM_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/mm.h>
  17 #include <linux/preempt.h>
  18 #include <asm/signal.h>
  19
  20 #include <linux/kvm.h>
  21 #include <linux/kvm_para.h>
  22
  23 #include "types.h"
  24
  25 #define KVM_MAX_VCPUS 4
  26 #define KVM_ALIAS_SLOTS 4
  27 #define KVM_MEMORY_SLOTS 8
  28 /* memory slots that does not exposed to userspace */
  29 #define KVM_PRIVATE_MEM_SLOTS 4
  30 #define KVM_PERMILLE_MMU_PAGES 20
  31 #define KVM_MIN_ALLOC_MMU_PAGES 64
  32 #define KVM_NUM_MMU_PAGES 1024
  33 #define KVM_MIN_FREE_MMU_PAGES 5
  34 #define KVM_REFILL_PAGES 25
  35 #define KVM_MAX_CPUID_ENTRIES 40
  36
  37 #define KVM_PIO_PAGE_OFFSET 1
  38
  39 /*
  40  * vcpu->requests bit members
  41  */
  42 #define KVM_REQ_TLB_FLUSH          0
  43
  44 #define NR_PTE_CHAIN_ENTRIES 5
  45
  46 struct kvm_pte_chain {
  47         u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
  48         struct hlist_node link;
  49 };
  50
  51 /*
  52  * kvm_mmu_page_role, below, is defined as:
  53  *
  54  *   bits 0:3 - total guest paging levels (2-4, or zero for real mode)
  55  *   bits 4:7 - page table level for this shadow (1-4)
  56  *   bits 8:9 - page table quadrant for 2-level guests
  57  *   bit   16 - "metaphysical" - gfn is not a real page (huge page/real mode)
  58  *   bits 17:19 - common access permissions for all ptes in this shadow page
  59  */
  60 union kvm_mmu_page_role {
  61         unsigned word;
  62         struct {
  63                 unsigned glevels : 4;
  64                 unsigned level : 4;
  65                 unsigned quadrant : 2;
  66                 unsigned pad_for_nice_hex_output : 6;
  67                 unsigned metaphysical : 1;
  68                 unsigned access : 3;
  69         };
  70 };
  71
  72 struct kvm_mmu_page {
  73         struct list_head link;
  74         struct hlist_node hash_link;
  75
  76         /*
  77          * The following two entries are used to key the shadow page in the
  78          * hash table.
  79          */
  80         gfn_t gfn;
  81         union kvm_mmu_page_role role;
  82
  83         u64 *spt;
  84         /* hold the gfn of each spte inside spt */
  85         gfn_t *gfns;
  86         unsigned long slot_bitmap; /* One bit set per slot which has memory
  87                                     * in this shadow page.
  88                                     */
  89         int multimapped;         /* More than one parent_pte? */
  90         int root_count;          /* Currently serving as active root */
  91         union {
  92                 u64 *parent_pte;               /* !multimapped */
  93                 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
  94         };
  95 };
  96
  97 struct kvm_vcpu;
  98 extern struct kmem_cache *kvm_vcpu_cache;
  99
 100 /*
 101  * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
 102  * 32-bit).  The kvm_mmu structure abstracts the details of the current mmu
 103  * mode.
 104  */
 105 struct kvm_mmu {
 106         void (*new_cr3)(struct kvm_vcpu *vcpu);
 107         int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
 108         void (*free)(struct kvm_vcpu *vcpu);
 109         gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
 110         void (*prefetch_page)(struct kvm_vcpu *vcpu,
 111                               struct kvm_mmu_page *page);
 112         hpa_t root_hpa;
 113         int root_level;
 114         int shadow_root_level;
 115
 116         u64 *pae_root;
 117 };
 118
 119 #define KVM_NR_MEM_OBJS 40
 120
 121 /*
 122  * We don't want allocation failures within the mmu code, so we preallocate
 123  * enough memory for a single page fault in a cache.
 124  */
 125 struct kvm_mmu_memory_cache {
 126         int nobjs;
 127         void *objects[KVM_NR_MEM_OBJS];
 128 };
 129
 130 struct kvm_guest_debug {
 131         int enabled;
 132         unsigned long bp[4];
 133         int singlestep;
 134 };
 135
 136 struct kvm_pio_request {
 137         unsigned long count;
 138         int cur_count;
 139         struct page *guest_pages[2];
 140         unsigned guest_page_offset;
 141         int in;
 142         int port;
 143         int size;
 144         int string;
 145         int down;
 146         int rep;
 147 };
 148
 149 struct kvm_vcpu_stat {
 150         u32 pf_fixed;
 151         u32 pf_guest;
 152         u32 tlb_flush;
 153         u32 invlpg;
 154
 155         u32 exits;
 156         u32 io_exits;
 157         u32 mmio_exits;
 158         u32 signal_exits;
 159         u32 irq_window_exits;
 160         u32 halt_exits;
 161         u32 halt_wakeup;
 162         u32 request_irq_exits;
 163         u32 irq_exits;
 164         u32 host_state_reload;
 165         u32 efer_reload;
 166         u32 fpu_reload;
 167         u32 insn_emulation;
 168         u32 insn_emulation_fail;
 169 };
 170
 171 /*
 172  * It would be nice to use something smarter than a linear search, TBD...
 173  * Thankfully we dont expect many devices to register (famous last words :),
 174  * so until then it will suffice.  At least its abstracted so we can change
 175  * in one place.
 176  */
 177 struct kvm_io_bus {
 178         int                   dev_count;
 179 #define NR_IOBUS_DEVS 6
 180         struct kvm_io_device *devs[NR_IOBUS_DEVS];
 181 };
 182
 183 void kvm_io_bus_init(struct kvm_io_bus *bus);
 184 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
 185 struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr);
 186 void kvm_io_bus_register_dev(struct kvm_io_bus *bus,
 187                              struct kvm_io_device *dev);
 188
 189 #ifdef CONFIG_HAS_IOMEM
 190 #define KVM_VCPU_MMIO                   \
 191         int mmio_needed;                \
 192         int mmio_read_completed;        \
 193         int mmio_is_write;              \
 194         int mmio_size;                  \
 195         unsigned char mmio_data[8];     \
 196         gpa_t mmio_phys_addr;
 197
 198 #else
 199 #define KVM_VCPU_MMIO
 200
 201 #endif
 202
 203 #define KVM_VCPU_COMM                                   \
 204         struct kvm *kvm;                                \
 205         struct preempt_notifier preempt_notifier;       \
 206         int vcpu_id;                                    \
 207         struct mutex mutex;                             \
 208         int   cpu;                                      \
 209         struct kvm_run *run;                            \
 210         int guest_mode;                                 \
 211         unsigned long requests;                         \
 212         struct kvm_guest_debug guest_debug;             \
 213         int fpu_active;                                 \
 214         int guest_fpu_loaded;                           \
 215         wait_queue_head_t wq;                           \
 216         int sigset_active;                              \
 217         sigset_t sigset;                                \
 218         struct kvm_vcpu_stat stat;                      \
 219         KVM_VCPU_MMIO
 220
 221 struct kvm_mem_alias {
 222         gfn_t base_gfn;
 223         unsigned long npages;
 224         gfn_t target_gfn;
 225 };
 226
 227 struct kvm_memory_slot {
 228         gfn_t base_gfn;
 229         unsigned long npages;
 230         unsigned long flags;
 231         unsigned long *rmap;
 232         unsigned long *dirty_bitmap;
 233         unsigned long userspace_addr;
 234         int user_alloc;
 235 };
 236
 237 struct kvm_vm_stat {
 238         u32 mmu_shadow_zapped;
 239         u32 mmu_pte_write;
 240         u32 mmu_pte_updated;
 241         u32 mmu_pde_zapped;
 242         u32 mmu_flooded;
 243         u32 mmu_recycled;
 244         u32 remote_tlb_flush;
 245 };
 246
 247 struct kvm {
 248         struct mutex lock; /* protects everything except vcpus */
 249         struct mm_struct *mm; /* userspace tied to this vm */
 250         int naliases;
 251         struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
 252         int nmemslots;
 253         struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
 254                                         KVM_PRIVATE_MEM_SLOTS];
 255         /*
 256          * Hash table of struct kvm_mmu_page.
 257          */
 258         struct list_head active_mmu_pages;
 259         unsigned int n_free_mmu_pages;
 260         unsigned int n_requested_mmu_pages;
 261         unsigned int n_alloc_mmu_pages;
 262         struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
 263         struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
 264         struct list_head vm_list;
 265         struct file *filp;
 266         struct kvm_io_bus mmio_bus;
 267         struct kvm_io_bus pio_bus;
 268         struct kvm_pic *vpic;
 269         struct kvm_ioapic *vioapic;
 270         int round_robin_prev_vcpu;
 271         unsigned int tss_addr;
 272         struct page *apic_access_page;
 273         struct kvm_vm_stat stat;
 274 };
 275
 276 /* The guest did something we don't support. */
 277 #define pr_unimpl(vcpu, fmt, ...)                                       \
 278  do {                                                                   \
 279         if (printk_ratelimit())                                         \
 280                 printk(KERN_ERR "kvm: %i: cpu%i " fmt,                  \
 281                        current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
 282  } while (0)
 283
 284 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
 285 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
 286
 287 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
 288 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
 289
 290 void vcpu_load(struct kvm_vcpu *vcpu);
 291 void vcpu_put(struct kvm_vcpu *vcpu);
 292
 293 void decache_vcpus_on_cpu(int cpu);
 294
 295
 296 int kvm_init(void *opaque, unsigned int vcpu_size,
 297                   struct module *module);
 298 void kvm_exit(void);
 299
 300 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
 301 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
 302 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
 303 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
 304
 305 extern struct page *bad_page;
 306
 307 int is_error_page(struct page *page);
 308 int kvm_is_error_hva(unsigned long addr);
 309 int kvm_set_memory_region(struct kvm *kvm,
 310                           struct kvm_userspace_memory_region *mem,
 311                           int user_alloc);
 312 int __kvm_set_memory_region(struct kvm *kvm,
 313                             struct kvm_userspace_memory_region *mem,
 314                             int user_alloc);
 315 int kvm_arch_set_memory_region(struct kvm *kvm,
 316                                 struct kvm_userspace_memory_region *mem,
 317                                 struct kvm_memory_slot old,
 318                                 int user_alloc);
 319 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
 320 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 321 void kvm_release_page_clean(struct page *page);
 322 void kvm_release_page_dirty(struct page *page);
 323 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
 324                         int len);
 325 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
 326 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
 327                          int offset, int len);
 328 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
 329                     unsigned long len);
 330 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
 331 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
 332 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 333 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
 334 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 335
 336 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
 337 void kvm_resched(struct kvm_vcpu *vcpu);
 338 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
 339 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
 340 void kvm_flush_remote_tlbs(struct kvm *kvm);
 341
 342 long kvm_arch_dev_ioctl(struct file *filp,
 343                         unsigned int ioctl, unsigned long arg);
 344 long kvm_arch_vcpu_ioctl(struct file *filp,
 345                          unsigned int ioctl, unsigned long arg);
 346 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 347 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 348
 349 int kvm_dev_ioctl_check_extension(long ext);
 350
 351 int kvm_get_dirty_log(struct kvm *kvm,
 352                         struct kvm_dirty_log *log, int *is_dirty);
 353 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 354                                 struct kvm_dirty_log *log);
 355
 356 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 357                                    struct
 358                                    kvm_userspace_memory_region *mem,
 359                                    int user_alloc);
 360 long kvm_arch_vm_ioctl(struct file *filp,
 361                        unsigned int ioctl, unsigned long arg);
 362 void kvm_arch_destroy_vm(struct kvm *kvm);
 363
 364 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 365 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
 366
 367 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
 368                                     struct kvm_translation *tr);
 369
 370 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 371 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
 372 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 373                                   struct kvm_sregs *sregs);
 374 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 375                                   struct kvm_sregs *sregs);
 376 int kvm_arch_vcpu_ioctl_debug_guest(struct kvm_vcpu *vcpu,
 377                                     struct kvm_debug_guest *dbg);
 378 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
 379
 380 int kvm_arch_init(void *opaque);
 381 void kvm_arch_exit(void);
 382
 383 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
 384 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
 385
 386 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
 387 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
 388 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
 389 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
 390 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
 391 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
 392
 393 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
 394 void kvm_arch_hardware_enable(void *garbage);
 395 void kvm_arch_hardware_disable(void *garbage);
 396 int kvm_arch_hardware_setup(void);
 397 void kvm_arch_hardware_unsetup(void);
 398 void kvm_arch_check_processor_compat(void *rtn);
 399 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 400
 401 void kvm_free_physmem(struct kvm *kvm);
 402
 403 struct  kvm *kvm_arch_create_vm(void);
 404 void kvm_arch_destroy_vm(struct kvm *kvm);
 405
 406 int kvm_cpu_get_interrupt(struct kvm_vcpu *v);
 407 int kvm_cpu_has_interrupt(struct kvm_vcpu *v);
 408
 409 static inline void kvm_guest_enter(void)
 410 {
 411         account_system_vtime(current);
 412         current->flags |= PF_VCPU;
 413 }
 414
 415 static inline void kvm_guest_exit(void)
 416 {
 417         account_system_vtime(current);
 418         current->flags &= ~PF_VCPU;
 419 }
 420
 421 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
 422 {
 423         return slot - kvm->memslots;
 424 }
 425
 426 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 427 {
 428         return (gpa_t)gfn << PAGE_SHIFT;
 429 }
 430
 431 enum kvm_stat_kind {
 432         KVM_STAT_VM,
 433         KVM_STAT_VCPU,
 434 };
 435
 436 struct kvm_stats_debugfs_item {
 437         const char *name;
 438         int offset;
 439         enum kvm_stat_kind kind;
 440         struct dentry *dentry;
 441 };
 442 extern struct kvm_stats_debugfs_item debugfs_entries[];
 443
 444 #if defined(CONFIG_X86)
 445 #include "x86.h"
 446 #endif
 447
 448 #endif