1 /* 2 * QEMU KVM support 3 * 4 * Copyright IBM, Corp. 2008 5 * 6 * Authors: 7 * Anthony Liguori <aliguori@us.ibm.com> 8 * 9 * This work is licensed under the terms of the GNU GPL, version 2 or later. 10 * See the COPYING file in the top-level directory. 11 * 12 */ 13 14 #ifndef QEMU_KVM_H 15 #define QEMU_KVM_H 16 17 #include <errno.h> 18 #include "config-host.h" 19 #include "qemu/queue.h" 20 #include "qom/cpu.h" 21 #include "exec/memattrs.h" 22 #include "hw/irq.h" 23 24 #ifdef CONFIG_KVM 25 #include <linux/kvm.h> 26 #include <linux/kvm_para.h> 27 #else 28 /* These constants must never be used at runtime if kvm_enabled() is false. 29 * They exist so we don't need #ifdefs around KVM-specific code that already 30 * checks kvm_enabled() properly. 31 */ 32 #define KVM_CPUID_SIGNATURE 0 33 #define KVM_CPUID_FEATURES 0 34 #define KVM_FEATURE_CLOCKSOURCE 0 35 #define KVM_FEATURE_NOP_IO_DELAY 0 36 #define KVM_FEATURE_MMU_OP 0 37 #define KVM_FEATURE_CLOCKSOURCE2 0 38 #define KVM_FEATURE_ASYNC_PF 0 39 #define KVM_FEATURE_STEAL_TIME 0 40 #define KVM_FEATURE_PV_EOI 0 41 #define KVM_FEATURE_CLOCKSOURCE_STABLE_BIT 0 42 #endif 43 44 extern bool kvm_allowed; 45 extern bool kvm_kernel_irqchip; 46 extern bool kvm_split_irqchip; 47 extern bool kvm_async_interrupts_allowed; 48 extern bool kvm_halt_in_kernel_allowed; 49 extern bool kvm_eventfds_allowed; 50 extern bool kvm_irqfds_allowed; 51 extern bool kvm_resamplefds_allowed; 52 extern bool kvm_msi_via_irqfd_allowed; 53 extern bool kvm_gsi_routing_allowed; 54 extern bool kvm_gsi_direct_mapping; 55 extern bool kvm_readonly_mem_allowed; 56 extern bool kvm_direct_msi_allowed; 57 extern bool kvm_ioeventfd_any_length_allowed; 58 59 #if defined CONFIG_KVM || !defined NEED_CPU_H 60 #define kvm_enabled() (kvm_allowed) 61 /** 62 * kvm_irqchip_in_kernel: 63 * 64 * Returns: true if the user asked us to create an in-kernel 65 * irqchip via the "kernel_irqchip=on" machine option. 66 * What this actually means is architecture and machine model 67 * specific: on PC, for instance, it means that the LAPIC, 68 * IOAPIC and PIT are all in kernel. This function should never 69 * be used from generic target-independent code: use one of the 70 * following functions or some other specific check instead. 71 */ 72 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip) 73 74 /** 75 * kvm_irqchip_is_split: 76 * 77 * Returns: true if the user asked us to split the irqchip 78 * implementation between user and kernel space. The details are 79 * architecture and machine specific. On PC, it means that the PIC, 80 * IOAPIC, and PIT are in user space while the LAPIC is in the kernel. 81 */ 82 #define kvm_irqchip_is_split() (kvm_split_irqchip) 83 84 /** 85 * kvm_async_interrupts_enabled: 86 * 87 * Returns: true if we can deliver interrupts to KVM 88 * asynchronously (ie by ioctl from any thread at any time) 89 * rather than having to do interrupt delivery synchronously 90 * (where the vcpu must be stopped at a suitable point first). 91 */ 92 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed) 93 94 /** 95 * kvm_halt_in_kernel 96 * 97 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run 98 * inside of kernel space. This only works if MP state is implemented. 99 */ 100 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed) 101 102 /** 103 * kvm_eventfds_enabled: 104 * 105 * Returns: true if we can use eventfds to receive notifications 106 * from a KVM CPU (ie the kernel supports eventds and we are running 107 * with a configuration where it is meaningful to use them). 108 */ 109 #define kvm_eventfds_enabled() (kvm_eventfds_allowed) 110 111 /** 112 * kvm_irqfds_enabled: 113 * 114 * Returns: true if we can use irqfds to inject interrupts into 115 * a KVM CPU (ie the kernel supports irqfds and we are running 116 * with a configuration where it is meaningful to use them). 117 */ 118 #define kvm_irqfds_enabled() (kvm_irqfds_allowed) 119 120 /** 121 * kvm_resamplefds_enabled: 122 * 123 * Returns: true if we can use resamplefds to inject interrupts into 124 * a KVM CPU (ie the kernel supports resamplefds and we are running 125 * with a configuration where it is meaningful to use them). 126 */ 127 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed) 128 129 /** 130 * kvm_msi_via_irqfd_enabled: 131 * 132 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt) 133 * to a KVM CPU via an irqfd. This requires that the kernel supports 134 * this and that we're running in a configuration that permits it. 135 */ 136 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed) 137 138 /** 139 * kvm_gsi_routing_enabled: 140 * 141 * Returns: true if GSI routing is enabled (ie the kernel supports 142 * it and we're running in a configuration that permits it). 143 */ 144 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed) 145 146 /** 147 * kvm_gsi_direct_mapping: 148 * 149 * Returns: true if GSI direct mapping is enabled. 150 */ 151 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping) 152 153 /** 154 * kvm_readonly_mem_enabled: 155 * 156 * Returns: true if KVM readonly memory is enabled (ie the kernel 157 * supports it and we're running in a configuration that permits it). 158 */ 159 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed) 160 161 /** 162 * kvm_direct_msi_enabled: 163 * 164 * Returns: true if KVM allows direct MSI injection. 165 */ 166 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed) 167 168 /** 169 * kvm_ioeventfd_any_length_enabled: 170 * Returns: true if KVM allows any length io eventfd. 171 */ 172 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed) 173 174 #else 175 #define kvm_enabled() (0) 176 #define kvm_irqchip_in_kernel() (false) 177 #define kvm_irqchip_is_split() (false) 178 #define kvm_async_interrupts_enabled() (false) 179 #define kvm_halt_in_kernel() (false) 180 #define kvm_eventfds_enabled() (false) 181 #define kvm_irqfds_enabled() (false) 182 #define kvm_resamplefds_enabled() (false) 183 #define kvm_msi_via_irqfd_enabled() (false) 184 #define kvm_gsi_routing_allowed() (false) 185 #define kvm_gsi_direct_mapping() (false) 186 #define kvm_readonly_mem_enabled() (false) 187 #define kvm_direct_msi_enabled() (false) 188 #define kvm_ioeventfd_any_length_enabled() (false) 189 #endif 190 191 struct kvm_run; 192 struct kvm_lapic_state; 193 struct kvm_irq_routing_entry; 194 195 typedef struct KVMCapabilityInfo { 196 const char *name; 197 int value; 198 } KVMCapabilityInfo; 199 200 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP } 201 #define KVM_CAP_LAST_INFO { NULL, 0 } 202 203 struct KVMState; 204 typedef struct KVMState KVMState; 205 extern KVMState *kvm_state; 206 207 /* external API */ 208 209 bool kvm_has_free_slot(MachineState *ms); 210 int kvm_has_sync_mmu(void); 211 int kvm_has_vcpu_events(void); 212 int kvm_has_robust_singlestep(void); 213 int kvm_has_debugregs(void); 214 int kvm_has_pit_state2(void); 215 int kvm_has_many_ioeventfds(void); 216 int kvm_has_gsi_routing(void); 217 int kvm_has_intx_set_mask(void); 218 219 int kvm_init_vcpu(CPUState *cpu); 220 int kvm_cpu_exec(CPUState *cpu); 221 222 #ifdef NEED_CPU_H 223 224 void kvm_setup_guest_memory(void *start, size_t size); 225 void kvm_flush_coalesced_mmio_buffer(void); 226 227 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr, 228 target_ulong len, int type); 229 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr, 230 target_ulong len, int type); 231 void kvm_remove_all_breakpoints(CPUState *cpu); 232 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap); 233 #ifndef _WIN32 234 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset); 235 #endif 236 237 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 238 int kvm_on_sigbus(int code, void *addr); 239 240 /* interface with exec.c */ 241 242 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align)); 243 244 /* internal API */ 245 246 int kvm_ioctl(KVMState *s, int type, ...); 247 248 int kvm_vm_ioctl(KVMState *s, int type, ...); 249 250 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...); 251 252 /** 253 * kvm_device_ioctl - call an ioctl on a kvm device 254 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE 255 * @type: The device-ctrl ioctl number 256 * 257 * Returns: -errno on error, nonnegative on success 258 */ 259 int kvm_device_ioctl(int fd, int type, ...); 260 261 /** 262 * kvm_vm_check_attr - check for existence of a specific vm attribute 263 * @s: The KVMState pointer 264 * @group: the group 265 * @attr: the attribute of that group to query for 266 * 267 * Returns: 1 if the attribute exists 268 * 0 if the attribute either does not exist or if the vm device 269 * interface is unavailable 270 */ 271 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr); 272 273 /** 274 * kvm_device_check_attr - check for existence of a specific device attribute 275 * @fd: The device file descriptor 276 * @group: the group 277 * @attr: the attribute of that group to query for 278 * 279 * Returns: 1 if the attribute exists 280 * 0 if the attribute either does not exist or if the vm device 281 * interface is unavailable 282 */ 283 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr); 284 285 /** 286 * kvm_device_access - set or get value of a specific vm attribute 287 * @fd: The device file descriptor 288 * @group: the group 289 * @attr: the attribute of that group to set or get 290 * @val: pointer to a storage area for the value 291 * @write: true for set and false for get operation 292 * 293 * This function is not allowed to fail. Use kvm_device_check_attr() 294 * in order to check for the availability of optional attributes. 295 */ 296 void kvm_device_access(int fd, int group, uint64_t attr, 297 void *val, bool write); 298 299 /** 300 * kvm_create_device - create a KVM device for the device control API 301 * @KVMState: The KVMState pointer 302 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the 303 * kernel source) 304 * @test: If true, only test if device can be created, but don't actually 305 * create the device. 306 * 307 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd; 308 */ 309 int kvm_create_device(KVMState *s, uint64_t type, bool test); 310 311 312 /* Arch specific hooks */ 313 314 extern const KVMCapabilityInfo kvm_arch_required_capabilities[]; 315 316 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run); 317 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run); 318 319 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run); 320 321 int kvm_arch_handle_ioapic_eoi(CPUState *cpu, struct kvm_run *run); 322 323 int kvm_arch_process_async_events(CPUState *cpu); 324 325 int kvm_arch_get_registers(CPUState *cpu); 326 327 /* state subset only touched by the VCPU itself during runtime */ 328 #define KVM_PUT_RUNTIME_STATE 1 329 /* state subset modified during VCPU reset */ 330 #define KVM_PUT_RESET_STATE 2 331 /* full state set, modified during initialization or on vmload */ 332 #define KVM_PUT_FULL_STATE 3 333 334 int kvm_arch_put_registers(CPUState *cpu, int level); 335 336 int kvm_arch_init(MachineState *ms, KVMState *s); 337 338 int kvm_arch_init_vcpu(CPUState *cpu); 339 340 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */ 341 unsigned long kvm_arch_vcpu_id(CPUState *cpu); 342 343 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 344 int kvm_arch_on_sigbus(int code, void *addr); 345 346 void kvm_arch_init_irq_routing(KVMState *s); 347 348 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route, 349 uint64_t address, uint32_t data, PCIDevice *dev); 350 351 int kvm_arch_msi_data_to_gsi(uint32_t data); 352 353 int kvm_set_irq(KVMState *s, int irq, int level); 354 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg); 355 356 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin); 357 void kvm_irqchip_commit_routes(KVMState *s); 358 359 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 360 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 361 362 struct kvm_guest_debug; 363 struct kvm_debug_exit_arch; 364 365 struct kvm_sw_breakpoint { 366 target_ulong pc; 367 target_ulong saved_insn; 368 int use_count; 369 QTAILQ_ENTRY(kvm_sw_breakpoint) entry; 370 }; 371 372 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint); 373 374 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu, 375 target_ulong pc); 376 377 int kvm_sw_breakpoints_active(CPUState *cpu); 378 379 int kvm_arch_insert_sw_breakpoint(CPUState *cpu, 380 struct kvm_sw_breakpoint *bp); 381 int kvm_arch_remove_sw_breakpoint(CPUState *cpu, 382 struct kvm_sw_breakpoint *bp); 383 int kvm_arch_insert_hw_breakpoint(target_ulong addr, 384 target_ulong len, int type); 385 int kvm_arch_remove_hw_breakpoint(target_ulong addr, 386 target_ulong len, int type); 387 void kvm_arch_remove_all_hw_breakpoints(void); 388 389 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg); 390 391 bool kvm_arch_stop_on_emulation_error(CPUState *cpu); 392 393 int kvm_check_extension(KVMState *s, unsigned int extension); 394 395 int kvm_vm_check_extension(KVMState *s, unsigned int extension); 396 397 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \ 398 ({ \ 399 struct kvm_enable_cap cap = { \ 400 .cap = capability, \ 401 .flags = cap_flags, \ 402 }; \ 403 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 404 int i; \ 405 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 406 i < ARRAY_SIZE(cap.args); i++) { \ 407 cap.args[i] = args_tmp[i]; \ 408 } \ 409 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ 410 }) 411 412 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \ 413 ({ \ 414 struct kvm_enable_cap cap = { \ 415 .cap = capability, \ 416 .flags = cap_flags, \ 417 }; \ 418 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 419 int i; \ 420 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 421 i < ARRAY_SIZE(cap.args); i++) { \ 422 cap.args[i] = args_tmp[i]; \ 423 } \ 424 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ 425 }) 426 427 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function, 428 uint32_t index, int reg); 429 430 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len); 431 432 #if !defined(CONFIG_USER_ONLY) 433 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr, 434 hwaddr *phys_addr); 435 #endif 436 437 #endif /* NEED_CPU_H */ 438 439 void kvm_cpu_synchronize_state(CPUState *cpu); 440 void kvm_cpu_synchronize_post_reset(CPUState *cpu); 441 void kvm_cpu_synchronize_post_init(CPUState *cpu); 442 443 /* generic hooks - to be moved/refactored once there are more users */ 444 445 static inline void cpu_synchronize_state(CPUState *cpu) 446 { 447 if (kvm_enabled()) { 448 kvm_cpu_synchronize_state(cpu); 449 } 450 } 451 452 static inline void cpu_synchronize_post_reset(CPUState *cpu) 453 { 454 if (kvm_enabled()) { 455 kvm_cpu_synchronize_post_reset(cpu); 456 } 457 } 458 459 static inline void cpu_synchronize_post_init(CPUState *cpu) 460 { 461 if (kvm_enabled()) { 462 kvm_cpu_synchronize_post_init(cpu); 463 } 464 } 465 466 int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg, PCIDevice *dev); 467 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg, 468 PCIDevice *dev); 469 void kvm_irqchip_release_virq(KVMState *s, int virq); 470 471 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter); 472 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint); 473 474 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 475 EventNotifier *rn, int virq); 476 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 477 int virq); 478 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n, 479 EventNotifier *rn, qemu_irq irq); 480 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, 481 qemu_irq irq); 482 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi); 483 void kvm_pc_gsi_handler(void *opaque, int n, int level); 484 void kvm_pc_setup_irq_routing(bool pci_enabled); 485 void kvm_init_irq_routing(KVMState *s); 486 487 /** 488 * kvm_arch_irqchip_create: 489 * @KVMState: The KVMState pointer 490 * @MachineState: The MachineState pointer 491 * 492 * Allow architectures to create an in-kernel irq chip themselves. 493 * 494 * Returns: < 0: error 495 * 0: irq chip was not created 496 * > 0: irq chip was created 497 */ 498 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s); 499 500 /** 501 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl 502 * @id: The register ID 503 * @source: The pointer to the value to be set. It must point to a variable 504 * of the correct type/size for the register being accessed. 505 * 506 * Returns: 0 on success, or a negative errno on failure. 507 */ 508 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source); 509 510 /** 511 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl 512 * @id: The register ID 513 * @target: The pointer where the value is to be stored. It must point to a 514 * variable of the correct type/size for the register being accessed. 515 * 516 * Returns: 0 on success, or a negative errno on failure. 517 */ 518 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target); 519 #endif 520