qemu-kvm.h

   1 /*
   2  * qemu/kvm integration
   3  *
   4  * Copyright (C) 2006-2008 Qumranet Technologies
   5  *
   6  * Licensed under the terms of the GNU GPL version 2 or higher.
   7  */
   8 #ifndef THE_ORIGINAL_AND_TRUE_QEMU_KVM_H
   9 #define THE_ORIGINAL_AND_TRUE_QEMU_KVM_H
  10
  11 #include "cpu.h"
  12
  13 #include <signal.h>
  14
  15 #ifdef CONFIG_KVM
  16
  17 #if defined(__s390__)
  18 #include <asm/ptrace.h>
  19 #endif
  20
  21 #include <stdint.h>
  22
  23 #ifndef __user
  24 #define __user /* temporary, until installed via make headers_install */
  25 #endif
  26
  27 #include <linux/kvm.h>
  28
  29 #include <signal.h>
  30
  31 /* FIXME: share this number with kvm */
  32 /* FIXME: or dynamically alloc/realloc regions */
  33 #ifdef __s390__
  34 #define KVM_MAX_NUM_MEM_REGIONS 1u
  35 #define MAX_VCPUS 64
  36 #define LIBKVM_S390_ORIGIN (0UL)
  37 #elif defined(__ia64__)
  38 #define KVM_MAX_NUM_MEM_REGIONS 32u
  39 #define MAX_VCPUS 256
  40 #else
  41 #define KVM_MAX_NUM_MEM_REGIONS 32u
  42 #define MAX_VCPUS 16
  43 #endif
  44
  45 /* kvm abi verison variable */
  46 extern int kvm_abi;
  47
  48 /**
  49  * \brief The KVM context
  50  *
  51  * The verbose KVM context
  52  */
  53
  54 struct kvm_context {
  55         /// Filedescriptor to /dev/kvm
  56         int fd;
  57         int vm_fd;
  58         /// Callbacks that KVM uses to emulate various unvirtualizable functionality
  59         struct kvm_callbacks *callbacks;
  60         void *opaque;
  61         /// is dirty pages logging enabled for all regions or not
  62         int dirty_pages_log_all;
  63         /// do not create in-kernel irqchip if set
  64         int no_irqchip_creation;
  65         /// in-kernel irqchip status
  66         int irqchip_in_kernel;
  67         /// ioctl to use to inject interrupts
  68         int irqchip_inject_ioctl;
  69         /// do not create in-kernel pit if set
  70         int no_pit_creation;
  71         /// in-kernel pit status
  72         int pit_in_kernel;
  73         /// in-kernel coalesced mmio
  74         int coalesced_mmio;
  75 #ifdef KVM_CAP_IRQ_ROUTING
  76         struct kvm_irq_routing *irq_routes;
  77         int nr_allocated_irq_routes;
  78 #endif
  79         void *used_gsi_bitmap;
  80         int max_gsi;
  81 };
  82
  83 struct kvm_vcpu_context
  84 {
  85         int fd;
  86         struct kvm_run *run;
  87         struct kvm_context *kvm;
  88         uint32_t id;
  89 };
  90
  91 typedef struct kvm_context *kvm_context_t;
  92 typedef struct kvm_vcpu_context *kvm_vcpu_context_t;
  93
  94 #include "kvm.h"
  95 int kvm_alloc_kernel_memory(kvm_context_t kvm, unsigned long memory,
  96                                                                 void **vm_mem);
  97 int kvm_alloc_userspace_memory(kvm_context_t kvm, unsigned long memory,
  98                                                                 void **vm_mem);
  99
 100 int kvm_arch_create(kvm_context_t kvm, unsigned long phys_mem_bytes,
 101                         void **vm_mem);
 102 int kvm_arch_run(kvm_vcpu_context_t vcpu);
 103
 104
 105 void kvm_show_code(kvm_vcpu_context_t vcpu);
 106
 107 int handle_halt(kvm_vcpu_context_t vcpu);
 108 int handle_shutdown(kvm_context_t kvm, CPUState *env);
 109 void post_kvm_run(kvm_context_t kvm, CPUState *env);
 110 int pre_kvm_run(kvm_context_t kvm, CPUState *env);
 111 int handle_io_window(kvm_context_t kvm);
 112 int handle_debug(kvm_vcpu_context_t vcpu, void *env);
 113 int try_push_interrupts(kvm_context_t kvm);
 114
 115 #if defined(__x86_64__) || defined(__i386__)
 116 struct kvm_msr_list *kvm_get_msr_list(kvm_context_t);
 117 int kvm_get_msrs(kvm_vcpu_context_t, struct kvm_msr_entry *msrs, int n);
 118 int kvm_set_msrs(kvm_vcpu_context_t, struct kvm_msr_entry *msrs, int n);
 119 int kvm_get_mce_cap_supported(kvm_context_t, uint64_t *mce_cap, int *max_banks);
 120 int kvm_setup_mce(kvm_vcpu_context_t vcpu, uint64_t *mcg_cap);
 121 struct kvm_x86_mce;
 122 int kvm_set_mce(kvm_vcpu_context_t vcpu, struct kvm_x86_mce *mce);
 123 #endif
 124
 125 /*!
 126  * \brief Create new KVM context
 127  *
 128  * This creates a new kvm_context. A KVM context is a small area of data that
 129  * holds information about the KVM instance that gets created by this call.\n
 130  * This should always be your first call to KVM.
 131  *
 132  * \param opaque Not used
 133  * \return NULL on failure
 134  */
 135 int kvm_init(int smp_cpus);
 136
 137 /*!
 138  * \brief Disable the in-kernel IRQCHIP creation
 139  *
 140  * In-kernel irqchip is enabled by default. If userspace irqchip is to be used,
 141  * this should be called prior to kvm_create().
 142  *
 143  * \param kvm Pointer to the kvm_context
 144  */
 145 void kvm_disable_irqchip_creation(kvm_context_t kvm);
 146
 147 /*!
 148  * \brief Disable the in-kernel PIT creation
 149  *
 150  * In-kernel pit is enabled by default. If userspace pit is to be used,
 151  * this should be called prior to kvm_create().
 152  *
 153  *  \param kvm Pointer to the kvm_context
 154  */
 155 void kvm_disable_pit_creation(kvm_context_t kvm);
 156
 157 /*!
 158  * \brief Create new virtual machine
 159  *
 160  * This creates a new virtual machine, maps physical RAM to it, and creates a
 161  * virtual CPU for it.\n
 162  * \n
 163  * Memory gets mapped for addresses 0->0xA0000, 0xC0000->phys_mem_bytes
 164  *
 165  * \param kvm Pointer to the current kvm_context
 166  * \param phys_mem_bytes The amount of physical ram you want the VM to have
 167  * \param phys_mem This pointer will be set to point to the memory that
 168  * kvm_create allocates for physical RAM
 169  * \return 0 on success
 170  */
 171 int kvm_create(kvm_context_t kvm,
 172                unsigned long phys_mem_bytes,
 173                void **phys_mem);
 174 int kvm_create_vm(kvm_context_t kvm);
 175 int kvm_check_extension(kvm_context_t kvm, int ext);
 176 void kvm_create_irqchip(kvm_context_t kvm);
 177
 178 /*!
 179  * \brief Create a new virtual cpu
 180  *
 181  * This creates a new virtual cpu (the first vcpu is created by kvm_create()).
 182  * Should be called from a thread dedicated to the vcpu.
 183  *
 184  * \param kvm kvm context
 185  * \param slot vcpu number (> 0)
 186  * \return 0 on success, -errno on failure
 187  */
 188 kvm_vcpu_context_t kvm_create_vcpu(CPUState *env, int id);
 189
 190 /*!
 191  * \brief Start the VCPU
 192  *
 193  * This starts the VCPU and virtualization is started.\n
 194  * \n
 195  * This function will not return until any of these conditions are met:
 196  * - An IO/MMIO handler does not return "0"
 197  * - An exception that neither the guest OS, nor KVM can handle occurs
 198  *
 199  * \note This function will call the callbacks registered in kvm_init()
 200  * to emulate those functions
 201  * \note If you at any point want to interrupt the VCPU, kvm_run() will
 202  * listen to the EINTR signal. This allows you to simulate external interrupts
 203  * and asyncronous IO.
 204  *
 205  * \param kvm Pointer to the current kvm_context
 206  * \param vcpu Which virtual CPU should be started
 207  * \return 0 on success, but you really shouldn't expect this function to
 208  * return except for when an error has occured, or when you have sent it
 209  * an EINTR signal.
 210  */
 211 int kvm_run(kvm_vcpu_context_t vcpu, void *env);
 212
 213 /*!
 214  * \brief Get interrupt flag from on last exit to userspace
 215  *
 216  * This gets the CPU interrupt flag as it was on the last exit to userspace.
 217  *
 218  * \param kvm Pointer to the current kvm_context
 219  * \param vcpu Which virtual CPU should get dumped
 220  * \return interrupt flag value (0 or 1)
 221  */
 222 int kvm_get_interrupt_flag(kvm_vcpu_context_t vcpu);
 223
 224 /*!
 225  * \brief Get the value of the APIC_BASE msr as of last exit to userspace
 226  *
 227  * This gets the APIC_BASE msr as it was on the last exit to userspace.
 228  *
 229  * \param kvm Pointer to the current kvm_context
 230  * \param vcpu Which virtual CPU should get dumped
 231  * \return APIC_BASE msr contents
 232  */
 233 uint64_t kvm_get_apic_base(kvm_vcpu_context_t vcpu);
 234
 235 /*!
 236  * \brief Check if a vcpu is ready for interrupt injection
 237  *
 238  * This checks if vcpu interrupts are not masked by mov ss or sti.
 239  *
 240  * \param kvm Pointer to the current kvm_context
 241  * \param vcpu Which virtual CPU should get dumped
 242  * \return boolean indicating interrupt injection readiness
 243  */
 244 int kvm_is_ready_for_interrupt_injection(kvm_vcpu_context_t vcpu);
 245
 246 /*!
 247  * \brief Read VCPU registers
 248  *
 249  * This gets the GP registers from the VCPU and outputs them
 250  * into a kvm_regs structure
 251  *
 252  * \note This function returns a \b copy of the VCPUs registers.\n
 253  * If you wish to modify the VCPUs GP registers, you should call kvm_set_regs()
 254  *
 255  * \param kvm Pointer to the current kvm_context
 256  * \param vcpu Which virtual CPU should get dumped
 257  * \param regs Pointer to a kvm_regs which will be populated with the VCPUs
 258  * registers values
 259  * \return 0 on success
 260  */
 261 int kvm_get_regs(kvm_vcpu_context_t vcpu, struct kvm_regs *regs);
 262
 263 /*!
 264  * \brief Write VCPU registers
 265  *
 266  * This sets the GP registers on the VCPU from a kvm_regs structure
 267  *
 268  * \note When this function returns, the regs pointer and the data it points to
 269  * can be discarded
 270  * \param kvm Pointer to the current kvm_context
 271  * \param vcpu Which virtual CPU should get dumped
 272  * \param regs Pointer to a kvm_regs which will be populated with the VCPUs
 273  * registers values
 274  * \return 0 on success
 275  */
 276 int kvm_set_regs(kvm_vcpu_context_t vcpu, struct kvm_regs *regs);
 277 /*!
 278  * \brief Read VCPU fpu registers
 279  *
 280  * This gets the FPU registers from the VCPU and outputs them
 281  * into a kvm_fpu structure
 282  *
 283  * \note This function returns a \b copy of the VCPUs registers.\n
 284  * If you wish to modify the VCPU FPU registers, you should call kvm_set_fpu()
 285  *
 286  * \param kvm Pointer to the current kvm_context
 287  * \param vcpu Which virtual CPU should get dumped
 288  * \param fpu Pointer to a kvm_fpu which will be populated with the VCPUs
 289  * fpu registers values
 290  * \return 0 on success
 291  */
 292 int kvm_get_fpu(kvm_vcpu_context_t vcpu, struct kvm_fpu *fpu);
 293
 294 /*!
 295  * \brief Write VCPU fpu registers
 296  *
 297  * This sets the FPU registers on the VCPU from a kvm_fpu structure
 298  *
 299  * \note When this function returns, the fpu pointer and the data it points to
 300  * can be discarded
 301  * \param kvm Pointer to the current kvm_context
 302  * \param vcpu Which virtual CPU should get dumped
 303  * \param fpu Pointer to a kvm_fpu which holds the new vcpu fpu state
 304  * \return 0 on success
 305  */
 306 int kvm_set_fpu(kvm_vcpu_context_t vcpu, struct kvm_fpu *fpu);
 307
 308 /*!
 309  * \brief Read VCPU system registers
 310  *
 311  * This gets the non-GP registers from the VCPU and outputs them
 312  * into a kvm_sregs structure
 313  *
 314  * \note This function returns a \b copy of the VCPUs registers.\n
 315  * If you wish to modify the VCPUs non-GP registers, you should call
 316  * kvm_set_sregs()
 317  *
 318  * \param kvm Pointer to the current kvm_context
 319  * \param vcpu Which virtual CPU should get dumped
 320  * \param regs Pointer to a kvm_sregs which will be populated with the VCPUs
 321  * registers values
 322  * \return 0 on success
 323  */
 324 int kvm_get_sregs(kvm_vcpu_context_t vcpu, struct kvm_sregs *regs);
 325
 326 /*!
 327  * \brief Write VCPU system registers
 328  *
 329  * This sets the non-GP registers on the VCPU from a kvm_sregs structure
 330  *
 331  * \note When this function returns, the regs pointer and the data it points to
 332  * can be discarded
 333  * \param kvm Pointer to the current kvm_context
 334  * \param vcpu Which virtual CPU should get dumped
 335  * \param regs Pointer to a kvm_sregs which will be populated with the VCPUs
 336  * registers values
 337  * \return 0 on success
 338  */
 339 int kvm_set_sregs(kvm_vcpu_context_t vcpu, struct kvm_sregs *regs);
 340
 341 #ifdef KVM_CAP_MP_STATE
 342 /*!
 343  *  * \brief Read VCPU MP state
 344  *
 345  */
 346 int kvm_get_mpstate(kvm_vcpu_context_t vcpu, struct kvm_mp_state *mp_state);
 347
 348 /*!
 349  *  * \brief Write VCPU MP state
 350  *
 351  */
 352 int kvm_set_mpstate(kvm_vcpu_context_t vcpu, struct kvm_mp_state *mp_state);
 353 /*!
 354  *  * \brief Reset VCPU MP state
 355  *
 356  */
 357 static inline int kvm_reset_mpstate(kvm_vcpu_context_t vcpu)
 358 {
 359     struct kvm_mp_state mp_state = {.mp_state = KVM_MP_STATE_UNINITIALIZED};
 360     return kvm_set_mpstate(vcpu, &mp_state);
 361 }
 362 #endif
 363
 364 /*!
 365  * \brief Simulate an external vectored interrupt
 366  *
 367  * This allows you to simulate an external vectored interrupt.
 368  *
 369  * \param kvm Pointer to the current kvm_context
 370  * \param vcpu Which virtual CPU should get dumped
 371  * \param irq Vector number
 372  * \return 0 on success
 373  */
 374 int kvm_inject_irq(kvm_vcpu_context_t vcpu, unsigned irq);
 375
 376 #ifdef KVM_CAP_SET_GUEST_DEBUG
 377 int kvm_set_guest_debug(kvm_vcpu_context_t, struct kvm_guest_debug *dbg);
 378 #endif
 379
 380 #if defined(__i386__) || defined(__x86_64__)
 381 /*!
 382  * \brief Setup a vcpu's cpuid instruction emulation
 383  *
 384  * Set up a table of cpuid function to cpuid outputs.\n
 385  *
 386  * \param kvm Pointer to the current kvm_context
 387  * \param vcpu Which virtual CPU should be initialized
 388  * \param nent number of entries to be installed
 389  * \param entries cpuid function entries table
 390  * \return 0 on success, or -errno on error
 391  */
 392 int kvm_setup_cpuid(kvm_vcpu_context_t vcpu, int nent,
 393                     struct kvm_cpuid_entry *entries);
 394
 395 /*!
 396  * \brief Setup a vcpu's cpuid instruction emulation
 397  *
 398  * Set up a table of cpuid function to cpuid outputs.
 399  * This call replaces the older kvm_setup_cpuid interface by adding a few
 400  * parameters to support cpuid functions that have sub-leaf values.
 401  *
 402  * \param kvm Pointer to the current kvm_context
 403  * \param vcpu Which virtual CPU should be initialized
 404  * \param nent number of entries to be installed
 405  * \param entries cpuid function entries table
 406  * \return 0 on success, or -errno on error
 407  */
 408 int kvm_setup_cpuid2(kvm_vcpu_context_t vcpu, int nent,
 409                      struct kvm_cpuid_entry2 *entries);
 410
 411 /*!
 412  * \brief Setting the number of shadow pages to be allocated to the vm
 413  *
 414  * \param kvm pointer to kvm_context
 415  * \param nrshadow_pages number of pages to be allocated
 416  */
 417 int kvm_set_shadow_pages(kvm_context_t kvm, unsigned int nrshadow_pages);
 418
 419 /*!
 420  * \brief Getting the number of shadow pages that are allocated to the vm
 421  *
 422  * \param kvm pointer to kvm_context
 423  * \param nrshadow_pages number of pages to be allocated
 424  */
 425 int kvm_get_shadow_pages(kvm_context_t kvm , unsigned int *nrshadow_pages);
 426
 427 /*!
 428  * \brief Set up cr8 for next time the vcpu is executed
 429  *
 430  * This is a fast setter for cr8, which will be applied when the
 431  * vcpu next enters guest mode.
 432  *
 433  * \param kvm Pointer to the current kvm_context
 434  * \param vcpu Which virtual CPU should get dumped
 435  * \param cr8 next cr8 value
 436  */
 437 void kvm_set_cr8(kvm_vcpu_context_t vcpu, uint64_t cr8);
 438
 439 /*!
 440  * \brief Get cr8 for sync tpr in qemu apic emulation
 441  *
 442  * This is a getter for cr8, which used to sync with the tpr in qemu
 443  * apic emualtion.
 444  *
 445  * \param kvm Pointer to the current kvm_context
 446  * \param vcpu Which virtual CPU should get dumped
 447  */
 448 __u64 kvm_get_cr8(kvm_vcpu_context_t vcpu);
 449 #endif
 450
 451 /*!
 452  * \brief Set a vcpu's signal mask for guest mode
 453  *
 454  * A vcpu can have different signals blocked in guest mode and user mode.
 455  * This allows guest execution to be interrupted on a signal, without requiring
 456  * that the signal be delivered to a signal handler (the signal can be
 457  * dequeued using sigwait(2).
 458  *
 459  * \param kvm Pointer to the current kvm_context
 460  * \param vcpu Which virtual CPU should be initialized
 461  * \param sigset signal mask for guest mode
 462  * \return 0 on success, or -errno on error
 463  */
 464 int kvm_set_signal_mask(kvm_vcpu_context_t vcpu, const sigset_t *sigset);
 465
 466 /*!
 467  * \brief Dump VCPU registers
 468  *
 469  * This dumps some of the information that KVM has about a virtual CPU, namely:
 470  * - GP Registers
 471  *
 472  * A much more verbose version of this is available as kvm_dump_vcpu()
 473  *
 474  * \param kvm Pointer to the current kvm_context
 475  * \param vcpu Which virtual CPU should get dumped
 476  * \return 0 on success
 477  */
 478 void kvm_show_regs(kvm_vcpu_context_t vcpu);
 479
 480
 481 void *kvm_create_phys_mem(kvm_context_t, unsigned long phys_start,
 482                           unsigned long len, int log, int writable);
 483 void kvm_destroy_phys_mem(kvm_context_t, unsigned long phys_start,
 484                           unsigned long len);
 485 void kvm_unregister_memory_area(kvm_context_t, uint64_t phys_start,
 486                                 unsigned long len);
 487
 488 int kvm_is_containing_region(kvm_context_t kvm, unsigned long phys_start, unsigned long size);
 489 int kvm_register_phys_mem(kvm_context_t kvm,
 490                         unsigned long phys_start, void *userspace_addr,
 491                         unsigned long len, int log);
 492 int kvm_get_dirty_pages(kvm_context_t, unsigned long phys_addr, void *buf);
 493 int kvm_get_dirty_pages_range(kvm_context_t kvm, unsigned long phys_addr,
 494                               unsigned long end_addr, void*opaque,
 495                               int (*cb)(unsigned long start, unsigned long len,
 496                                         void*bitmap, void *opaque));
 497 int kvm_register_coalesced_mmio(kvm_context_t kvm,
 498                                 uint64_t addr, uint32_t size);
 499 int kvm_unregister_coalesced_mmio(kvm_context_t kvm,
 500                                   uint64_t addr, uint32_t size);
 501
 502 /*!
 503  * \brief Create a memory alias
 504  *
 505  * Aliases a portion of physical memory to another portion.  If the guest
 506  * accesses the alias region, it will behave exactly as if it accessed
 507  * the target memory.
 508  */
 509 int kvm_create_memory_alias(kvm_context_t,
 510                             uint64_t phys_start, uint64_t len,
 511                             uint64_t target_phys);
 512
 513 /*!
 514  * \brief Destroy a memory alias
 515  *
 516  * Removes an alias created with kvm_create_memory_alias().
 517  */
 518 int kvm_destroy_memory_alias(kvm_context_t, uint64_t phys_start);
 519
 520 /*!
 521  * \brief Get a bitmap of guest ram pages which are allocated to the guest.
 522  *
 523  * \param kvm Pointer to the current kvm_context
 524  * \param phys_addr Memory slot phys addr
 525  * \param bitmap Long aligned address of a big enough bitmap (one bit per page)
 526  */
 527 int kvm_get_mem_map(kvm_context_t kvm, unsigned long phys_addr, void *bitmap);
 528 int kvm_get_mem_map_range(kvm_context_t kvm, unsigned long phys_addr,
 529                            unsigned long len, void *buf, void *opaque,
 530                            int (*cb)(unsigned long start,unsigned long len,
 531                                      void* bitmap, void* opaque));
 532 int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status);
 533
 534 int kvm_dirty_pages_log_enable_slot(kvm_context_t kvm,
 535                                     uint64_t phys_start,
 536                                     uint64_t len);
 537 int kvm_dirty_pages_log_disable_slot(kvm_context_t kvm,
 538                                      uint64_t phys_start,
 539                                      uint64_t len);
 540 /*!
 541  * \brief Enable dirty-pages-logging for all memory regions
 542  *
 543  * \param kvm Pointer to the current kvm_context
 544  */
 545 int kvm_dirty_pages_log_enable_all(kvm_context_t kvm);
 546
 547 /*!
 548  * \brief Disable dirty-page-logging for some memory regions
 549  *
 550  * Disable dirty-pages-logging for those memory regions that were
 551  * created with dirty-page-logging disabled.
 552  *
 553  * \param kvm Pointer to the current kvm_context
 554  */
 555 int kvm_dirty_pages_log_reset(kvm_context_t kvm);
 556
 557 /*!
 558  * \brief Query whether in kernel irqchip is used
 559  *
 560  * \param kvm Pointer to the current kvm_context
 561  */
 562 int kvm_irqchip_in_kernel(kvm_context_t kvm);
 563
 564 #ifdef KVM_CAP_IRQCHIP
 565 /*!
 566  * \brief Dump in kernel IRQCHIP contents
 567  *
 568  * Dump one of the in kernel irq chip devices, including PIC (master/slave)
 569  * and IOAPIC into a kvm_irqchip structure
 570  *
 571  * \param kvm Pointer to the current kvm_context
 572  * \param chip The irq chip device to be dumped
 573  */
 574 int kvm_get_irqchip(kvm_context_t kvm, struct kvm_irqchip *chip);
 575
 576 /*!
 577  * \brief Set in kernel IRQCHIP contents
 578  *
 579  * Write one of the in kernel irq chip devices, including PIC (master/slave)
 580  * and IOAPIC
 581  *
 582  *
 583  * \param kvm Pointer to the current kvm_context
 584  * \param chip THe irq chip device to be written
 585  */
 586 int kvm_set_irqchip(kvm_context_t kvm, struct kvm_irqchip *chip);
 587
 588 #if defined(__i386__) || defined(__x86_64__)
 589 /*!
 590  * \brief Get in kernel local APIC for vcpu
 591  *
 592  * Save the local apic state including the timer of a virtual CPU
 593  *
 594  * \param kvm Pointer to the current kvm_context
 595  * \param vcpu Which virtual CPU should be accessed
 596  * \param s Local apic state of the specific virtual CPU
 597  */
 598 int kvm_get_lapic(kvm_vcpu_context_t vcpu, struct kvm_lapic_state *s);
 599
 600 /*!
 601  * \brief Set in kernel local APIC for vcpu
 602  *
 603  * Restore the local apic state including the timer of a virtual CPU
 604  *
 605  * \param kvm Pointer to the current kvm_context
 606  * \param vcpu Which virtual CPU should be accessed
 607  * \param s Local apic state of the specific virtual CPU
 608  */
 609 int kvm_set_lapic(kvm_vcpu_context_t vcpu, struct kvm_lapic_state *s);
 610
 611 #endif
 612
 613 /*!
 614  * \brief Simulate an NMI
 615  *
 616  * This allows you to simulate a non-maskable interrupt.
 617  *
 618  * \param kvm Pointer to the current kvm_context
 619  * \param vcpu Which virtual CPU should get dumped
 620  * \return 0 on success
 621  */
 622 int kvm_inject_nmi(kvm_vcpu_context_t vcpu);
 623
 624 #endif
 625
 626 /*!
 627  * \brief Simulate an x86 MCE
 628  *
 629  * This allows you to simulate a x86 MCE.
 630  *
 631  * \param cenv Which virtual CPU should get MCE injected
 632  * \param bank Bank number
 633  * \param status MSR_MCI_STATUS
 634  * \param mcg_status MSR_MCG_STATUS
 635  * \param addr MSR_MCI_ADDR
 636  * \param misc MSR_MCI_MISC
 637  */
 638 void kvm_inject_x86_mce(CPUState *cenv, int bank, uint64_t status,
 639                         uint64_t mcg_status, uint64_t addr, uint64_t misc);
 640
 641 /*!
 642  * \brief Query wheather in kernel pit is used
 643  *
 644  *  \param kvm Pointer to the current kvm_context
 645  */
 646 int kvm_pit_in_kernel(kvm_context_t kvm);
 647
 648 /*!
 649  * \brief Initialize coalesced MMIO
 650  *
 651  * Check for coalesced MMIO capability and store in context
 652  *
 653  * \param kvm Pointer to the current kvm_context
 654  */
 655 int kvm_init_coalesced_mmio(kvm_context_t kvm);
 656
 657 #ifdef KVM_CAP_PIT
 658
 659 #if defined(__i386__) || defined(__x86_64__)
 660 /*!
 661  * \brief Get in kernel PIT of the virtual domain
 662  *
 663  * Save the PIT state.
 664  *
 665  * \param kvm Pointer to the current kvm_context
 666  * \param s PIT state of the virtual domain
 667  */
 668 int kvm_get_pit(kvm_context_t kvm, struct kvm_pit_state *s);
 669
 670 /*!
 671  * \brief Set in kernel PIT of the virtual domain
 672  *
 673  * Restore the PIT state.
 674  * Timer would be retriggerred after restored.
 675  *
 676  * \param kvm Pointer to the current kvm_context
 677  * \param s PIT state of the virtual domain
 678  */
 679 int kvm_set_pit(kvm_context_t kvm, struct kvm_pit_state *s);
 680
 681 int kvm_reinject_control(kvm_context_t kvm, int pit_reinject);
 682
 683 #ifdef KVM_CAP_PIT_STATE2
 684 /*!
 685  * \brief Check for kvm support of kvm_pit_state2
 686  *
 687  * \param kvm Pointer to the current kvm_context
 688  * \return 0 on success
 689  */
 690 int kvm_has_pit_state2(kvm_context_t kvm);
 691
 692 /*!
 693  * \brief Set in kernel PIT state2 of the virtual domain
 694  *
 695  *
 696  * \param kvm Pointer to the current kvm_context
 697  * \param ps2 PIT state2 of the virtual domain
 698  * \return 0 on success
 699  */
 700 int kvm_set_pit2(kvm_context_t kvm, struct kvm_pit_state2 *ps2);
 701
 702 /*!
 703  * \brief Get in kernel PIT state2 of the virtual domain
 704  *
 705  *
 706  * \param kvm Pointer to the current kvm_context
 707  * \param ps2 PIT state2 of the virtual domain
 708  * \return 0 on success
 709  */
 710 int kvm_get_pit2(kvm_context_t kvm, struct kvm_pit_state2 *ps2);
 711
 712 #endif
 713 #endif
 714 #endif
 715
 716 #ifdef KVM_CAP_VAPIC
 717
 718 /*!
 719  * \brief Enable kernel tpr access reporting
 720  *
 721  * When tpr access reporting is enabled, the kernel will call the
 722  * ->tpr_access() callback every time the guest vcpu accesses the tpr.
 723  *
 724  * \param kvm Pointer to the current kvm_context
 725  * \param vcpu vcpu to enable tpr access reporting on
 726  */
 727 int kvm_enable_tpr_access_reporting(kvm_vcpu_context_t vcpu);
 728
 729 /*!
 730  * \brief Disable kernel tpr access reporting
 731  *
 732  * Undoes the effect of kvm_enable_tpr_access_reporting().
 733  *
 734  * \param kvm Pointer to the current kvm_context
 735  * \param vcpu vcpu to disable tpr access reporting on
 736  */
 737 int kvm_disable_tpr_access_reporting(kvm_vcpu_context_t vcpu);
 738
 739 int kvm_enable_vapic(kvm_vcpu_context_t vcpu, uint64_t vapic);
 740
 741 #endif
 742
 743 #if defined(__s390__)
 744 int kvm_s390_initial_reset(kvm_context_t kvm, int slot);
 745 int kvm_s390_interrupt(kvm_context_t kvm, int slot,
 746         struct kvm_s390_interrupt *kvmint);
 747 int kvm_s390_set_initial_psw(kvm_context_t kvm, int slot, psw_t psw);
 748 int kvm_s390_store_status(kvm_context_t kvm, int slot, unsigned long addr);
 749 #endif
 750
 751 #ifdef KVM_CAP_DEVICE_ASSIGNMENT
 752 /*!
 753  * \brief Notifies host kernel about a PCI device to be assigned to a guest
 754  *
 755  * Used for PCI device assignment, this function notifies the host
 756  * kernel about the assigning of the physical PCI device to a guest.
 757  *
 758  * \param kvm Pointer to the current kvm_context
 759  * \param assigned_dev Parameters, like bus, devfn number, etc
 760  */
 761 int kvm_assign_pci_device(kvm_context_t kvm,
 762                           struct kvm_assigned_pci_dev *assigned_dev);
 763
 764 /*!
 765  * \brief Assign IRQ for an assigned device
 766  *
 767  * Used for PCI device assignment, this function assigns IRQ numbers for
 768  * an physical device and guest IRQ handling.
 769  *
 770  * \param kvm Pointer to the current kvm_context
 771  * \param assigned_irq Parameters, like dev id, host irq, guest irq, etc
 772  */
 773 int kvm_assign_irq(kvm_context_t kvm,
 774                    struct kvm_assigned_irq *assigned_irq);
 775
 776 #ifdef KVM_CAP_ASSIGN_DEV_IRQ
 777 /*!
 778  * \brief Deassign IRQ for an assigned device
 779  *
 780  * Used for PCI device assignment, this function deassigns IRQ numbers
 781  * for an assigned device.
 782  *
 783  * \param kvm Pointer to the current kvm_context
 784  * \param assigned_irq Parameters, like dev id, host irq, guest irq, etc
 785  */
 786 int kvm_deassign_irq(kvm_context_t kvm,
 787                    struct kvm_assigned_irq *assigned_irq);
 788 #endif
 789 #endif
 790
 791 /*!
 792  * \brief Determines whether destroying memory regions is allowed
 793  *
 794  * KVM before 2.6.29 had a bug when destroying memory regions.
 795  *
 796  * \param kvm Pointer to the current kvm_context
 797  */
 798 int kvm_destroy_memory_region_works(kvm_context_t kvm);
 799
 800 #ifdef KVM_CAP_DEVICE_DEASSIGNMENT
 801 /*!
 802  * \brief Notifies host kernel about a PCI device to be deassigned from a guest
 803  *
 804  * Used for hot remove PCI device, this function notifies the host
 805  * kernel about the deassigning of the physical PCI device from a guest.
 806  *
 807  * \param kvm Pointer to the current kvm_context
 808  * \param assigned_dev Parameters, like bus, devfn number, etc
 809  */
 810 int kvm_deassign_pci_device(kvm_context_t kvm,
 811                             struct kvm_assigned_pci_dev *assigned_dev);
 812 #endif
 813
 814 /*!
 815  * \brief Checks whether the generic irq routing capability is present
 816  *
 817  * Checks whether kvm can reroute interrupts among the various interrupt
 818  * controllers.
 819  *
 820  * \param kvm Pointer to the current kvm_context
 821  */
 822 int kvm_has_gsi_routing(kvm_context_t kvm);
 823
 824 /*!
 825  * \brief Determines the number of gsis that can be routed
 826  *
 827  * Returns the number of distinct gsis that can be routed by kvm.  This is
 828  * also the number of distinct routes (if a gsi has two routes, than another
 829  * gsi cannot be used...)
 830  *
 831  * \param kvm Pointer to the current kvm_context
 832  */
 833 int kvm_get_gsi_count(kvm_context_t kvm);
 834
 835 /*!
 836  * \brief Clears the temporary irq routing table
 837  *
 838  * Clears the temporary irq routing table.  Nothing is committed to the
 839  * running VM.
 840  *
 841  * \param kvm Pointer to the current kvm_context
 842  */
 843 int kvm_clear_gsi_routes(kvm_context_t kvm);
 844
 845 /*!
 846  * \brief Adds an irq route to the temporary irq routing table
 847  *
 848  * Adds an irq route to the temporary irq routing table.  Nothing is
 849  * committed to the running VM.
 850  *
 851  * \param kvm Pointer to the current kvm_context
 852  */
 853 int kvm_add_irq_route(kvm_context_t kvm, int gsi, int irqchip, int pin);
 854
 855 /*!
 856  * \brief Removes an irq route from the temporary irq routing table
 857  *
 858  * Adds an irq route to the temporary irq routing table.  Nothing is
 859  * committed to the running VM.
 860  *
 861  * \param kvm Pointer to the current kvm_context
 862  */
 863 int kvm_del_irq_route(kvm_context_t kvm, int gsi, int irqchip, int pin);
 864
 865 struct kvm_irq_routing_entry;
 866 /*!
 867  * \brief Adds a routing entry to the temporary irq routing table
 868  *
 869  * Adds a filled routing entry to the temporary irq routing table. Nothing is
 870  * committed to the running VM.
 871  *
 872  * \param kvm Pointer to the current kvm_context
 873  */
 874 int kvm_add_routing_entry(kvm_context_t kvm,
 875                           struct kvm_irq_routing_entry* entry);
 876
 877 /*!
 878  * \brief Removes a routing from the temporary irq routing table
 879  *
 880  * Remove a routing to the temporary irq routing table.  Nothing is
 881  * committed to the running VM.
 882  *
 883  * \param kvm Pointer to the current kvm_context
 884  */
 885 int kvm_del_routing_entry(kvm_context_t kvm,
 886                           struct kvm_irq_routing_entry* entry);
 887
 888 /*!
 889  * \brief Updates a routing in the temporary irq routing table
 890  *
 891  * Update a routing in the temporary irq routing table
 892  * with a new value. entry type and GSI can not be changed.
 893  * Nothing is committed to the running VM.
 894  *
 895  * \param kvm Pointer to the current kvm_context
 896  */
 897 int kvm_update_routing_entry(kvm_context_t kvm,
 898                              struct kvm_irq_routing_entry* entry,
 899                              struct kvm_irq_routing_entry* newentry
 900 );
 901
 902 /*!
 903  * \brief Commit the temporary irq routing table
 904  *
 905  * Commit the temporary irq routing table to the running VM.
 906  *
 907  * \param kvm Pointer to the current kvm_context
 908  */
 909 int kvm_commit_irq_routes(kvm_context_t kvm);
 910
 911 /*!
 912  * \brief Get unused GSI number for irq routing table
 913  *
 914  * Get unused GSI number for irq routing table
 915  *
 916  * \param kvm Pointer to the current kvm_context
 917  */
 918 int kvm_get_irq_route_gsi(kvm_context_t kvm);
 919
 920 /*!
 921  * \brief Create a file descriptor for injecting interrupts
 922  *
 923  * Creates an eventfd based file-descriptor that maps to a specific GSI
 924  * in the guest.  eventfd compliant signaling (write() from userspace, or
 925  * eventfd_signal() from kernelspace) will cause the GSI to inject
 926  * itself into the guest at the next available window.
 927  *
 928  * \param kvm Pointer to the current kvm_context
 929  * \param gsi GSI to assign to this fd
 930  * \param flags reserved, must be zero
 931  */
 932 int kvm_irqfd(kvm_context_t kvm, int gsi, int flags);
 933
 934 #ifdef KVM_CAP_DEVICE_MSIX
 935 int kvm_assign_set_msix_nr(kvm_context_t kvm,
 936                            struct kvm_assigned_msix_nr *msix_nr);
 937 int kvm_assign_set_msix_entry(kvm_context_t kvm,
 938                               struct kvm_assigned_msix_entry *entry);
 939 #endif
 940
 941 uint32_t kvm_get_supported_cpuid(kvm_context_t kvm, uint32_t function, int reg);
 942
 943 #else /* !CONFIG_KVM */
 944
 945 struct kvm_pit_state { };
 946
 947 #endif /* !CONFIG_KVM */
 948
 949
 950 int kvm_main_loop(void);
 951 int kvm_qemu_init(void);
 952 int kvm_qemu_create_context(void);
 953 int kvm_init_ap(void);
 954 int kvm_vcpu_inited(CPUState *env);
 955 void kvm_load_registers(CPUState *env);
 956 void kvm_save_registers(CPUState *env);
 957 void kvm_load_mpstate(CPUState *env);
 958 void kvm_save_mpstate(CPUState *env);
 959 int kvm_cpu_exec(CPUState *env);
 960 int kvm_insert_breakpoint(CPUState *current_env, target_ulong addr,
 961                           target_ulong len, int type);
 962 int kvm_remove_breakpoint(CPUState *current_env, target_ulong addr,
 963                           target_ulong len, int type);
 964 void kvm_remove_all_breakpoints(CPUState *current_env);
 965 int kvm_update_guest_debug(CPUState *env, unsigned long reinject_trap);
 966 int kvm_qemu_init_env(CPUState *env);
 967 int kvm_qemu_check_extension(int ext);
 968 void kvm_apic_init(CPUState *env);
 969 /* called from vcpu initialization */
 970 void qemu_kvm_load_lapic(CPUState *env);
 971
 972 void kvm_hpet_enable_kpit(void);
 973 void kvm_hpet_disable_kpit(void);
 974 int kvm_set_irq(int irq, int level, int *status);
 975
 976 int kvm_physical_memory_set_dirty_tracking(int enable);
 977 int kvm_update_dirty_pages_log(void);
 978 int kvm_get_phys_ram_page_bitmap(unsigned char *bitmap);
 979
 980 void qemu_kvm_call_with_env(void (*func)(void *), void *data, CPUState *env);
 981 void qemu_kvm_cpuid_on_env(CPUState *env);
 982 void kvm_inject_interrupt(CPUState *env, int mask);
 983 void kvm_update_after_sipi(CPUState *env);
 984 void kvm_update_interrupt_request(CPUState *env);
 985 void kvm_set_phys_mem(target_phys_addr_t start_addr, ram_addr_t size,
 986                       ram_addr_t phys_offset);
 987 void *kvm_cpu_create_phys_mem(target_phys_addr_t start_addr,
 988                               unsigned long size, int log, int writable);
 989
 990 void kvm_cpu_destroy_phys_mem(target_phys_addr_t start_addr,
 991                               unsigned long size);
 992 void kvm_qemu_log_memory(target_phys_addr_t start, target_phys_addr_t size,
 993                          int log);
 994 int kvm_setup_guest_memory(void *area, unsigned long size);
 995 int kvm_qemu_create_memory_alias(uint64_t phys_start,
 996                                  uint64_t len,
 997                                  uint64_t target_phys);
 998 int kvm_qemu_destroy_memory_alias(uint64_t phys_start);
 999
1000 int kvm_arch_qemu_create_context(void);
1001
1002 void kvm_arch_save_regs(CPUState *env);
1003 void kvm_arch_load_regs(CPUState *env);
1004 void kvm_arch_load_mpstate(CPUState *env);
1005 void kvm_arch_save_mpstate(CPUState *env);
1006 int kvm_arch_qemu_init_env(CPUState *cenv);
1007 void kvm_arch_pre_kvm_run(void *opaque, CPUState *env);
1008 void kvm_arch_post_kvm_run(void *opaque, CPUState *env);
1009 int kvm_arch_has_work(CPUState *env);
1010 void kvm_arch_process_irqchip_events(CPUState *env);
1011 int kvm_arch_try_push_interrupts(void *opaque);
1012 void kvm_arch_push_nmi(void *opaque);
1013 void kvm_arch_update_regs_for_sipi(CPUState *env);
1014 void kvm_arch_cpu_reset(CPUState *env);
1015 int kvm_set_boot_cpu_id(uint32_t id);
1016
1017 struct kvm_guest_debug;
1018 struct kvm_debug_exit_arch;
1019
1020 struct kvm_sw_breakpoint {
1021     target_ulong pc;
1022     target_ulong saved_insn;
1023     int use_count;
1024     TAILQ_ENTRY(kvm_sw_breakpoint) entry;
1025 };
1026
1027 TAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
1028
1029 int kvm_arch_debug(struct kvm_debug_exit_arch *arch_info);
1030 int kvm_sw_breakpoints_active(CPUState *env);
1031 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *env, target_ulong pc);
1032 int kvm_arch_insert_sw_breakpoint(CPUState *current_env,
1033                                   struct kvm_sw_breakpoint *bp);
1034 int kvm_arch_remove_sw_breakpoint(CPUState *current_env,
1035                                   struct kvm_sw_breakpoint *bp);
1036 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
1037                                   target_ulong len, int type);
1038 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
1039                                   target_ulong len, int type);
1040 void kvm_arch_remove_all_hw_breakpoints(void);
1041 void kvm_arch_update_guest_debug(CPUState *env, struct kvm_guest_debug *dbg);
1042
1043 void qemu_kvm_aio_wait_start(void);
1044 void qemu_kvm_aio_wait(void);
1045 void qemu_kvm_aio_wait_end(void);
1046
1047 void qemu_kvm_notify_work(void);
1048
1049 void kvm_tpr_opt_setup(void);
1050 void kvm_tpr_access_report(CPUState *env, uint64_t rip, int is_write);
1051 void kvm_tpr_vcpu_start(CPUState *env);
1052
1053 int qemu_kvm_get_dirty_pages(unsigned long phys_addr, void *buf);
1054 int kvm_coalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);
1055 int kvm_uncoalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);
1056
1057 int kvm_arch_init_irq_routing(void);
1058
1059 int kvm_mmio_read(void *opaque, uint64_t addr, uint8_t *data, int len);
1060 int kvm_mmio_write(void *opaque, uint64_t addr, uint8_t *data, int len);
1061
1062 #ifdef USE_KVM_DEVICE_ASSIGNMENT
1063 struct ioperm_data;
1064
1065 void kvm_ioperm(CPUState *env, void *data);
1066 void kvm_add_ioperm_data(struct ioperm_data *data);
1067 void kvm_remove_ioperm_data(unsigned long start_port, unsigned long num);
1068 void kvm_arch_do_ioperm(void *_data);
1069 #endif
1070
1071 #define ALIGN(x, y)  (((x)+(y)-1) & ~((y)-1))
1072 #define BITMAP_SIZE(m) (ALIGN(((m)>>TARGET_PAGE_BITS), HOST_LONG_BITS) / 8)
1073
1074 #ifdef CONFIG_KVM
1075 #include "sys-queue.h"
1076
1077 extern int kvm_allowed;
1078 extern int kvm_irqchip;
1079 extern int kvm_pit;
1080 extern int kvm_pit_reinject;
1081 extern int kvm_nested;
1082 extern kvm_context_t kvm_context;
1083
1084 struct ioperm_data {
1085     unsigned long start_port;
1086     unsigned long num;
1087     int turn_on;
1088     LIST_ENTRY(ioperm_data) entries;
1089 };
1090
1091 void qemu_kvm_cpu_stop(CPUState *env);
1092 int kvm_arch_halt(void *opaque, kvm_vcpu_context_t vcpu);
1093 int handle_tpr_access(void *opaque, kvm_vcpu_context_t vcpu,
1094                              uint64_t rip, int is_write);
1095 int kvm_has_sync_mmu(void);
1096
1097 #define kvm_enabled() (kvm_allowed)
1098 #define qemu_kvm_irqchip_in_kernel() kvm_irqchip_in_kernel(kvm_context)
1099 #define qemu_kvm_pit_in_kernel() kvm_pit_in_kernel(kvm_context)
1100 #define qemu_kvm_has_gsi_routing() kvm_has_gsi_routing(kvm_context)
1101 #ifdef TARGET_I386
1102 #define qemu_kvm_has_pit_state2() kvm_has_pit_state2(kvm_context)
1103 #endif
1104 void kvm_init_vcpu(CPUState *env);
1105 void kvm_load_tsc(CPUState *env);
1106 #else
1107 #define kvm_has_sync_mmu() (0)
1108 #define kvm_enabled() (0)
1109 #define kvm_nested 0
1110 #define qemu_kvm_irqchip_in_kernel() (0)
1111 #define qemu_kvm_pit_in_kernel() (0)
1112 #define qemu_kvm_has_gsi_routing() (0)
1113 #ifdef TARGET_I386
1114 #define qemu_kvm_has_pit_state2() (0)
1115 #endif
1116 #define kvm_load_registers(env) do {} while(0)
1117 #define kvm_save_registers(env) do {} while(0)
1118 #define qemu_kvm_cpu_stop(env) do {} while(0)
1119 static inline void kvm_init_vcpu(CPUState *env) { }
1120 static inline void kvm_load_tsc(CPUState *env) {}
1121 #endif
1122
1123 void kvm_mutex_unlock(void);
1124 void kvm_mutex_lock(void);
1125
1126 static inline void kvm_sleep_begin(void)
1127 {
1128     if (kvm_enabled())
1129         kvm_mutex_unlock();
1130 }
1131
1132 static inline void kvm_sleep_end(void)
1133 {
1134     if (kvm_enabled())
1135         kvm_mutex_lock();
1136 }
1137
1138 int kvm_physical_sync_dirty_bitmap(target_phys_addr_t start_addr, target_phys_addr_t end_addr);
1139
1140 int kvm_log_start(target_phys_addr_t phys_addr, target_phys_addr_t len);
1141 int kvm_log_stop(target_phys_addr_t phys_addr, target_phys_addr_t len);
1142
1143
1144 static inline int kvm_sync_vcpus(void) { return 0; }
1145
1146 static inline void kvm_arch_get_registers(CPUState *env)
1147 {
1148     kvm_save_registers(env);
1149     kvm_save_mpstate(env);
1150 }
1151
1152 static inline void kvm_arch_put_registers(CPUState *env)
1153 {
1154     kvm_load_registers(env);
1155     kvm_load_mpstate(env);
1156 }
1157
1158 static inline void cpu_synchronize_state(CPUState *env, int modified)
1159 {
1160     if (kvm_enabled()) {
1161         if (modified)
1162             kvm_arch_put_registers(env);
1163         else
1164             kvm_arch_get_registers(env);
1165     }
1166 }
1167
1168 uint32_t kvm_arch_get_supported_cpuid(CPUState *env, uint32_t function,
1169                                       int reg);
1170
1171
1172 static inline int kvm_set_migration_log(int enable)
1173 {
1174     return kvm_physical_memory_set_dirty_tracking(enable);
1175 }
1176
1177 typedef struct KVMSlot
1178 {
1179     target_phys_addr_t start_addr;
1180     ram_addr_t memory_size;
1181     ram_addr_t phys_offset;
1182     int slot;
1183     int flags;
1184 } KVMSlot;
1185
1186 typedef struct kvm_dirty_log KVMDirtyLog;
1187
1188 typedef struct KVMState
1189 {
1190     KVMSlot slots[32];
1191     int fd;
1192     int vmfd;
1193     int coalesced_mmio;
1194     int broken_set_mem_region;
1195     int migration_log;
1196 #ifdef KVM_CAP_SET_GUEST_DEBUG
1197     struct kvm_sw_breakpoint_head kvm_sw_breakpoints;
1198 #endif
1199     struct kvm_context kvm_context;
1200 } KVMState;
1201
1202
1203 #endif