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 "qemu/queue.h" 18 #include "qom/cpu.h" 19 #include "exec/memattrs.h" 20 #include "hw/irq.h" 21 22 #ifdef NEED_CPU_H 23 # ifdef CONFIG_KVM 24 # include <linux/kvm.h> 25 # define CONFIG_KVM_IS_POSSIBLE 26 # endif 27 #else 28 # define CONFIG_KVM_IS_POSSIBLE 29 #endif 30 31 #ifdef CONFIG_KVM_IS_POSSIBLE 32 33 extern bool kvm_allowed; 34 extern bool kvm_kernel_irqchip; 35 extern bool kvm_split_irqchip; 36 extern bool kvm_async_interrupts_allowed; 37 extern bool kvm_halt_in_kernel_allowed; 38 extern bool kvm_eventfds_allowed; 39 extern bool kvm_irqfds_allowed; 40 extern bool kvm_resamplefds_allowed; 41 extern bool kvm_msi_via_irqfd_allowed; 42 extern bool kvm_gsi_routing_allowed; 43 extern bool kvm_gsi_direct_mapping; 44 extern bool kvm_readonly_mem_allowed; 45 extern bool kvm_direct_msi_allowed; 46 extern bool kvm_ioeventfd_any_length_allowed; 47 extern bool kvm_msi_use_devid; 48 49 #define kvm_enabled() (kvm_allowed) 50 /** 51 * kvm_irqchip_in_kernel: 52 * 53 * Returns: true if the user asked us to create an in-kernel 54 * irqchip via the "kernel_irqchip=on" machine option. 55 * What this actually means is architecture and machine model 56 * specific: on PC, for instance, it means that the LAPIC, 57 * IOAPIC and PIT are all in kernel. This function should never 58 * be used from generic target-independent code: use one of the 59 * following functions or some other specific check instead. 60 */ 61 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip) 62 63 /** 64 * kvm_irqchip_is_split: 65 * 66 * Returns: true if the user asked us to split the irqchip 67 * implementation between user and kernel space. The details are 68 * architecture and machine specific. On PC, it means that the PIC, 69 * IOAPIC, and PIT are in user space while the LAPIC is in the kernel. 70 */ 71 #define kvm_irqchip_is_split() (kvm_split_irqchip) 72 73 /** 74 * kvm_async_interrupts_enabled: 75 * 76 * Returns: true if we can deliver interrupts to KVM 77 * asynchronously (ie by ioctl from any thread at any time) 78 * rather than having to do interrupt delivery synchronously 79 * (where the vcpu must be stopped at a suitable point first). 80 */ 81 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed) 82 83 /** 84 * kvm_halt_in_kernel 85 * 86 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run 87 * inside of kernel space. This only works if MP state is implemented. 88 */ 89 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed) 90 91 /** 92 * kvm_eventfds_enabled: 93 * 94 * Returns: true if we can use eventfds to receive notifications 95 * from a KVM CPU (ie the kernel supports eventds and we are running 96 * with a configuration where it is meaningful to use them). 97 */ 98 #define kvm_eventfds_enabled() (kvm_eventfds_allowed) 99 100 /** 101 * kvm_irqfds_enabled: 102 * 103 * Returns: true if we can use irqfds to inject interrupts into 104 * a KVM CPU (ie the kernel supports irqfds and we are running 105 * with a configuration where it is meaningful to use them). 106 */ 107 #define kvm_irqfds_enabled() (kvm_irqfds_allowed) 108 109 /** 110 * kvm_resamplefds_enabled: 111 * 112 * Returns: true if we can use resamplefds to inject interrupts into 113 * a KVM CPU (ie the kernel supports resamplefds and we are running 114 * with a configuration where it is meaningful to use them). 115 */ 116 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed) 117 118 /** 119 * kvm_msi_via_irqfd_enabled: 120 * 121 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt) 122 * to a KVM CPU via an irqfd. This requires that the kernel supports 123 * this and that we're running in a configuration that permits it. 124 */ 125 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed) 126 127 /** 128 * kvm_gsi_routing_enabled: 129 * 130 * Returns: true if GSI routing is enabled (ie the kernel supports 131 * it and we're running in a configuration that permits it). 132 */ 133 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed) 134 135 /** 136 * kvm_gsi_direct_mapping: 137 * 138 * Returns: true if GSI direct mapping is enabled. 139 */ 140 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping) 141 142 /** 143 * kvm_readonly_mem_enabled: 144 * 145 * Returns: true if KVM readonly memory is enabled (ie the kernel 146 * supports it and we're running in a configuration that permits it). 147 */ 148 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed) 149 150 /** 151 * kvm_direct_msi_enabled: 152 * 153 * Returns: true if KVM allows direct MSI injection. 154 */ 155 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed) 156 157 /** 158 * kvm_ioeventfd_any_length_enabled: 159 * Returns: true if KVM allows any length io eventfd. 160 */ 161 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed) 162 163 /** 164 * kvm_msi_devid_required: 165 * Returns: true if KVM requires a device id to be provided while 166 * defining an MSI routing entry. 167 */ 168 #define kvm_msi_devid_required() (kvm_msi_use_devid) 169 170 #else 171 172 #define kvm_enabled() (0) 173 #define kvm_irqchip_in_kernel() (false) 174 #define kvm_irqchip_is_split() (false) 175 #define kvm_async_interrupts_enabled() (false) 176 #define kvm_halt_in_kernel() (false) 177 #define kvm_eventfds_enabled() (false) 178 #define kvm_irqfds_enabled() (false) 179 #define kvm_resamplefds_enabled() (false) 180 #define kvm_msi_via_irqfd_enabled() (false) 181 #define kvm_gsi_routing_allowed() (false) 182 #define kvm_gsi_direct_mapping() (false) 183 #define kvm_readonly_mem_enabled() (false) 184 #define kvm_direct_msi_enabled() (false) 185 #define kvm_ioeventfd_any_length_enabled() (false) 186 #define kvm_msi_devid_required() (false) 187 188 #endif /* CONFIG_KVM_IS_POSSIBLE */ 189 190 struct kvm_run; 191 struct kvm_lapic_state; 192 struct kvm_irq_routing_entry; 193 194 typedef struct KVMCapabilityInfo { 195 const char *name; 196 int value; 197 } KVMCapabilityInfo; 198 199 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP } 200 #define KVM_CAP_LAST_INFO { NULL, 0 } 201 202 struct KVMState; 203 typedef struct KVMState KVMState; 204 extern KVMState *kvm_state; 205 206 /* external API */ 207 208 bool kvm_has_free_slot(MachineState *ms); 209 bool kvm_has_sync_mmu(void); 210 int kvm_has_vcpu_events(void); 211 int kvm_has_robust_singlestep(void); 212 int kvm_has_debugregs(void); 213 int kvm_has_pit_state2(void); 214 int kvm_has_many_ioeventfds(void); 215 int kvm_has_gsi_routing(void); 216 int kvm_has_intx_set_mask(void); 217 218 int kvm_init_vcpu(CPUState *cpu); 219 int kvm_cpu_exec(CPUState *cpu); 220 int kvm_destroy_vcpu(CPUState *cpu); 221 222 /** 223 * kvm_arm_supports_user_irq 224 * 225 * Not all KVM implementations support notifications for kernel generated 226 * interrupt events to user space. This function indicates whether the current 227 * KVM implementation does support them. 228 * 229 * Returns: true if KVM supports using kernel generated IRQs from user space 230 */ 231 bool kvm_arm_supports_user_irq(void); 232 233 /** 234 * kvm_memcrypt_enabled - return boolean indicating whether memory encryption 235 * is enabled 236 * Returns: 1 memory encryption is enabled 237 * 0 memory encryption is disabled 238 */ 239 bool kvm_memcrypt_enabled(void); 240 241 /** 242 * kvm_memcrypt_encrypt_data: encrypt the memory range 243 * 244 * Return: 1 failed to encrypt the range 245 * 0 succesfully encrypted memory region 246 */ 247 int kvm_memcrypt_encrypt_data(uint8_t *ptr, uint64_t len); 248 249 250 #ifdef NEED_CPU_H 251 #include "cpu.h" 252 253 void kvm_flush_coalesced_mmio_buffer(void); 254 255 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr, 256 target_ulong len, int type); 257 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr, 258 target_ulong len, int type); 259 void kvm_remove_all_breakpoints(CPUState *cpu); 260 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap); 261 262 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 263 int kvm_on_sigbus(int code, void *addr); 264 265 /* interface with exec.c */ 266 267 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align, bool shared)); 268 269 /* internal API */ 270 271 int kvm_ioctl(KVMState *s, int type, ...); 272 273 int kvm_vm_ioctl(KVMState *s, int type, ...); 274 275 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...); 276 277 /** 278 * kvm_device_ioctl - call an ioctl on a kvm device 279 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE 280 * @type: The device-ctrl ioctl number 281 * 282 * Returns: -errno on error, nonnegative on success 283 */ 284 int kvm_device_ioctl(int fd, int type, ...); 285 286 /** 287 * kvm_vm_check_attr - check for existence of a specific vm attribute 288 * @s: The KVMState pointer 289 * @group: the group 290 * @attr: the attribute of that group to query for 291 * 292 * Returns: 1 if the attribute exists 293 * 0 if the attribute either does not exist or if the vm device 294 * interface is unavailable 295 */ 296 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr); 297 298 /** 299 * kvm_device_check_attr - check for existence of a specific device attribute 300 * @fd: The device file descriptor 301 * @group: the group 302 * @attr: the attribute of that group to query for 303 * 304 * Returns: 1 if the attribute exists 305 * 0 if the attribute either does not exist or if the vm device 306 * interface is unavailable 307 */ 308 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr); 309 310 /** 311 * kvm_device_access - set or get value of a specific vm attribute 312 * @fd: The device file descriptor 313 * @group: the group 314 * @attr: the attribute of that group to set or get 315 * @val: pointer to a storage area for the value 316 * @write: true for set and false for get operation 317 * @errp: error object handle 318 * 319 * Returns: 0 on success 320 * < 0 on error 321 * Use kvm_device_check_attr() in order to check for the availability 322 * of optional attributes. 323 */ 324 int kvm_device_access(int fd, int group, uint64_t attr, 325 void *val, bool write, Error **errp); 326 327 /** 328 * kvm_create_device - create a KVM device for the device control API 329 * @KVMState: The KVMState pointer 330 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the 331 * kernel source) 332 * @test: If true, only test if device can be created, but don't actually 333 * create the device. 334 * 335 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd; 336 */ 337 int kvm_create_device(KVMState *s, uint64_t type, bool test); 338 339 /** 340 * kvm_device_supported - probe whether KVM supports specific device 341 * 342 * @vmfd: The fd handler for VM 343 * @type: type of device 344 * 345 * @return: true if supported, otherwise false. 346 */ 347 bool kvm_device_supported(int vmfd, uint64_t type); 348 349 /* Arch specific hooks */ 350 351 extern const KVMCapabilityInfo kvm_arch_required_capabilities[]; 352 353 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run); 354 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run); 355 356 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run); 357 358 int kvm_arch_process_async_events(CPUState *cpu); 359 360 int kvm_arch_get_registers(CPUState *cpu); 361 362 /* state subset only touched by the VCPU itself during runtime */ 363 #define KVM_PUT_RUNTIME_STATE 1 364 /* state subset modified during VCPU reset */ 365 #define KVM_PUT_RESET_STATE 2 366 /* full state set, modified during initialization or on vmload */ 367 #define KVM_PUT_FULL_STATE 3 368 369 int kvm_arch_put_registers(CPUState *cpu, int level); 370 371 int kvm_arch_init(MachineState *ms, KVMState *s); 372 373 int kvm_arch_init_vcpu(CPUState *cpu); 374 375 bool kvm_vcpu_id_is_valid(int vcpu_id); 376 377 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */ 378 unsigned long kvm_arch_vcpu_id(CPUState *cpu); 379 380 #ifdef TARGET_I386 381 #define KVM_HAVE_MCE_INJECTION 1 382 void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 383 #endif 384 385 void kvm_arch_init_irq_routing(KVMState *s); 386 387 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route, 388 uint64_t address, uint32_t data, PCIDevice *dev); 389 390 /* Notify arch about newly added MSI routes */ 391 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route, 392 int vector, PCIDevice *dev); 393 /* Notify arch about released MSI routes */ 394 int kvm_arch_release_virq_post(int virq); 395 396 int kvm_arch_msi_data_to_gsi(uint32_t data); 397 398 int kvm_set_irq(KVMState *s, int irq, int level); 399 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg); 400 401 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin); 402 403 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 404 405 struct kvm_guest_debug; 406 struct kvm_debug_exit_arch; 407 408 struct kvm_sw_breakpoint { 409 target_ulong pc; 410 target_ulong saved_insn; 411 int use_count; 412 QTAILQ_ENTRY(kvm_sw_breakpoint) entry; 413 }; 414 415 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint); 416 417 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu, 418 target_ulong pc); 419 420 int kvm_sw_breakpoints_active(CPUState *cpu); 421 422 int kvm_arch_insert_sw_breakpoint(CPUState *cpu, 423 struct kvm_sw_breakpoint *bp); 424 int kvm_arch_remove_sw_breakpoint(CPUState *cpu, 425 struct kvm_sw_breakpoint *bp); 426 int kvm_arch_insert_hw_breakpoint(target_ulong addr, 427 target_ulong len, int type); 428 int kvm_arch_remove_hw_breakpoint(target_ulong addr, 429 target_ulong len, int type); 430 void kvm_arch_remove_all_hw_breakpoints(void); 431 432 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg); 433 434 bool kvm_arch_stop_on_emulation_error(CPUState *cpu); 435 436 int kvm_check_extension(KVMState *s, unsigned int extension); 437 438 int kvm_vm_check_extension(KVMState *s, unsigned int extension); 439 440 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \ 441 ({ \ 442 struct kvm_enable_cap cap = { \ 443 .cap = capability, \ 444 .flags = cap_flags, \ 445 }; \ 446 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 447 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \ 448 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \ 449 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ 450 }) 451 452 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \ 453 ({ \ 454 struct kvm_enable_cap cap = { \ 455 .cap = capability, \ 456 .flags = cap_flags, \ 457 }; \ 458 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 459 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \ 460 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \ 461 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ 462 }) 463 464 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function, 465 uint32_t index, int reg); 466 uint32_t kvm_arch_get_supported_msr_feature(KVMState *s, uint32_t index); 467 468 469 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len); 470 471 #if !defined(CONFIG_USER_ONLY) 472 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr, 473 hwaddr *phys_addr); 474 #endif 475 476 #endif /* NEED_CPU_H */ 477 478 void kvm_cpu_synchronize_state(CPUState *cpu); 479 void kvm_cpu_synchronize_post_reset(CPUState *cpu); 480 void kvm_cpu_synchronize_post_init(CPUState *cpu); 481 void kvm_cpu_synchronize_pre_loadvm(CPUState *cpu); 482 483 void kvm_init_cpu_signals(CPUState *cpu); 484 485 /** 486 * kvm_irqchip_add_msi_route - Add MSI route for specific vector 487 * @s: KVM state 488 * @vector: which vector to add. This can be either MSI/MSIX 489 * vector. The function will automatically detect whether 490 * MSI/MSIX is enabled, and fetch corresponding MSI 491 * message. 492 * @dev: Owner PCI device to add the route. If @dev is specified 493 * as @NULL, an empty MSI message will be inited. 494 * @return: virq (>=0) when success, errno (<0) when failed. 495 */ 496 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev); 497 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg, 498 PCIDevice *dev); 499 void kvm_irqchip_commit_routes(KVMState *s); 500 void kvm_irqchip_release_virq(KVMState *s, int virq); 501 502 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter); 503 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint); 504 505 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 506 EventNotifier *rn, int virq); 507 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 508 int virq); 509 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n, 510 EventNotifier *rn, qemu_irq irq); 511 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, 512 qemu_irq irq); 513 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi); 514 void kvm_pc_gsi_handler(void *opaque, int n, int level); 515 void kvm_pc_setup_irq_routing(bool pci_enabled); 516 void kvm_init_irq_routing(KVMState *s); 517 518 /** 519 * kvm_arch_irqchip_create: 520 * @KVMState: The KVMState pointer 521 * @MachineState: The MachineState pointer 522 * 523 * Allow architectures to create an in-kernel irq chip themselves. 524 * 525 * Returns: < 0: error 526 * 0: irq chip was not created 527 * > 0: irq chip was created 528 */ 529 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s); 530 531 /** 532 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl 533 * @id: The register ID 534 * @source: The pointer to the value to be set. It must point to a variable 535 * of the correct type/size for the register being accessed. 536 * 537 * Returns: 0 on success, or a negative errno on failure. 538 */ 539 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source); 540 541 /** 542 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl 543 * @id: The register ID 544 * @target: The pointer where the value is to be stored. It must point to a 545 * variable of the correct type/size for the register being accessed. 546 * 547 * Returns: 0 on success, or a negative errno on failure. 548 */ 549 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target); 550 struct ppc_radix_page_info *kvm_get_radix_page_info(void); 551 int kvm_get_max_memslots(void); 552 #endif 553