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_async_interrupts_allowed; 47 extern bool kvm_halt_in_kernel_allowed; 48 extern bool kvm_eventfds_allowed; 49 extern bool kvm_irqfds_allowed; 50 extern bool kvm_resamplefds_allowed; 51 extern bool kvm_msi_via_irqfd_allowed; 52 extern bool kvm_gsi_routing_allowed; 53 extern bool kvm_gsi_direct_mapping; 54 extern bool kvm_readonly_mem_allowed; 55 extern bool kvm_direct_msi_allowed; 56 extern bool kvm_ioeventfd_any_length_allowed; 57 58 #if defined CONFIG_KVM || !defined NEED_CPU_H 59 #define kvm_enabled() (kvm_allowed) 60 /** 61 * kvm_irqchip_in_kernel: 62 * 63 * Returns: true if the user asked us to create an in-kernel 64 * irqchip via the "kernel_irqchip=on" machine option. 65 * What this actually means is architecture and machine model 66 * specific: on PC, for instance, it means that the LAPIC, 67 * IOAPIC and PIT are all in kernel. This function should never 68 * be used from generic target-independent code: use one of the 69 * following functions or some other specific check instead. 70 */ 71 #define kvm_irqchip_in_kernel() (kvm_kernel_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 #else 164 #define kvm_enabled() (0) 165 #define kvm_irqchip_in_kernel() (false) 166 #define kvm_async_interrupts_enabled() (false) 167 #define kvm_halt_in_kernel() (false) 168 #define kvm_eventfds_enabled() (false) 169 #define kvm_irqfds_enabled() (false) 170 #define kvm_resamplefds_enabled() (false) 171 #define kvm_msi_via_irqfd_enabled() (false) 172 #define kvm_gsi_routing_allowed() (false) 173 #define kvm_gsi_direct_mapping() (false) 174 #define kvm_readonly_mem_enabled() (false) 175 #define kvm_direct_msi_enabled() (false) 176 #define kvm_ioeventfd_any_length_enabled() (false) 177 #endif 178 179 struct kvm_run; 180 struct kvm_lapic_state; 181 struct kvm_irq_routing_entry; 182 183 typedef struct KVMCapabilityInfo { 184 const char *name; 185 int value; 186 } KVMCapabilityInfo; 187 188 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP } 189 #define KVM_CAP_LAST_INFO { NULL, 0 } 190 191 struct KVMState; 192 typedef struct KVMState KVMState; 193 extern KVMState *kvm_state; 194 195 /* external API */ 196 197 bool kvm_has_free_slot(MachineState *ms); 198 int kvm_has_sync_mmu(void); 199 int kvm_has_vcpu_events(void); 200 int kvm_has_robust_singlestep(void); 201 int kvm_has_debugregs(void); 202 int kvm_has_pit_state2(void); 203 int kvm_has_many_ioeventfds(void); 204 int kvm_has_gsi_routing(void); 205 int kvm_has_intx_set_mask(void); 206 207 int kvm_init_vcpu(CPUState *cpu); 208 int kvm_cpu_exec(CPUState *cpu); 209 210 #ifdef NEED_CPU_H 211 212 void kvm_setup_guest_memory(void *start, size_t size); 213 void kvm_flush_coalesced_mmio_buffer(void); 214 215 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr, 216 target_ulong len, int type); 217 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr, 218 target_ulong len, int type); 219 void kvm_remove_all_breakpoints(CPUState *cpu); 220 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap); 221 #ifndef _WIN32 222 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset); 223 #endif 224 225 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 226 int kvm_on_sigbus(int code, void *addr); 227 228 /* interface with exec.c */ 229 230 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align)); 231 232 /* internal API */ 233 234 int kvm_ioctl(KVMState *s, int type, ...); 235 236 int kvm_vm_ioctl(KVMState *s, int type, ...); 237 238 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...); 239 240 /** 241 * kvm_device_ioctl - call an ioctl on a kvm device 242 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE 243 * @type: The device-ctrl ioctl number 244 * 245 * Returns: -errno on error, nonnegative on success 246 */ 247 int kvm_device_ioctl(int fd, int type, ...); 248 249 /** 250 * kvm_vm_check_attr - check for existence of a specific vm attribute 251 * @s: The KVMState pointer 252 * @group: the group 253 * @attr: the attribute of that group to query for 254 * 255 * Returns: 1 if the attribute exists 256 * 0 if the attribute either does not exist or if the vm device 257 * interface is unavailable 258 */ 259 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr); 260 261 /** 262 * kvm_device_check_attr - check for existence of a specific device attribute 263 * @fd: The device file descriptor 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_device_check_attr(int fd, uint32_t group, uint64_t attr); 272 273 /** 274 * kvm_device_access - set or get value of a specific vm attribute 275 * @fd: The device file descriptor 276 * @group: the group 277 * @attr: the attribute of that group to set or get 278 * @val: pointer to a storage area for the value 279 * @write: true for set and false for get operation 280 * 281 * This function is not allowed to fail. Use kvm_device_check_attr() 282 * in order to check for the availability of optional attributes. 283 */ 284 void kvm_device_access(int fd, int group, uint64_t attr, 285 void *val, bool write); 286 287 /** 288 * kvm_create_device - create a KVM device for the device control API 289 * @KVMState: The KVMState pointer 290 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the 291 * kernel source) 292 * @test: If true, only test if device can be created, but don't actually 293 * create the device. 294 * 295 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd; 296 */ 297 int kvm_create_device(KVMState *s, uint64_t type, bool test); 298 299 300 /* Arch specific hooks */ 301 302 extern const KVMCapabilityInfo kvm_arch_required_capabilities[]; 303 304 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run); 305 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run); 306 307 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run); 308 309 int kvm_arch_process_async_events(CPUState *cpu); 310 311 int kvm_arch_get_registers(CPUState *cpu); 312 313 /* state subset only touched by the VCPU itself during runtime */ 314 #define KVM_PUT_RUNTIME_STATE 1 315 /* state subset modified during VCPU reset */ 316 #define KVM_PUT_RESET_STATE 2 317 /* full state set, modified during initialization or on vmload */ 318 #define KVM_PUT_FULL_STATE 3 319 320 int kvm_arch_put_registers(CPUState *cpu, int level); 321 322 int kvm_arch_init(MachineState *ms, KVMState *s); 323 324 int kvm_arch_init_vcpu(CPUState *cpu); 325 326 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */ 327 unsigned long kvm_arch_vcpu_id(CPUState *cpu); 328 329 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 330 int kvm_arch_on_sigbus(int code, void *addr); 331 332 void kvm_arch_init_irq_routing(KVMState *s); 333 334 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route, 335 uint64_t address, uint32_t data, PCIDevice *dev); 336 337 int kvm_arch_msi_data_to_gsi(uint32_t data); 338 339 int kvm_set_irq(KVMState *s, int irq, int level); 340 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg); 341 342 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin); 343 void kvm_irqchip_commit_routes(KVMState *s); 344 345 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 346 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 347 348 struct kvm_guest_debug; 349 struct kvm_debug_exit_arch; 350 351 struct kvm_sw_breakpoint { 352 target_ulong pc; 353 target_ulong saved_insn; 354 int use_count; 355 QTAILQ_ENTRY(kvm_sw_breakpoint) entry; 356 }; 357 358 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint); 359 360 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu, 361 target_ulong pc); 362 363 int kvm_sw_breakpoints_active(CPUState *cpu); 364 365 int kvm_arch_insert_sw_breakpoint(CPUState *cpu, 366 struct kvm_sw_breakpoint *bp); 367 int kvm_arch_remove_sw_breakpoint(CPUState *cpu, 368 struct kvm_sw_breakpoint *bp); 369 int kvm_arch_insert_hw_breakpoint(target_ulong addr, 370 target_ulong len, int type); 371 int kvm_arch_remove_hw_breakpoint(target_ulong addr, 372 target_ulong len, int type); 373 void kvm_arch_remove_all_hw_breakpoints(void); 374 375 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg); 376 377 bool kvm_arch_stop_on_emulation_error(CPUState *cpu); 378 379 int kvm_check_extension(KVMState *s, unsigned int extension); 380 381 int kvm_vm_check_extension(KVMState *s, unsigned int extension); 382 383 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \ 384 ({ \ 385 struct kvm_enable_cap cap = { \ 386 .cap = capability, \ 387 .flags = cap_flags, \ 388 }; \ 389 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 390 int i; \ 391 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 392 i < ARRAY_SIZE(cap.args); i++) { \ 393 cap.args[i] = args_tmp[i]; \ 394 } \ 395 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ 396 }) 397 398 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \ 399 ({ \ 400 struct kvm_enable_cap cap = { \ 401 .cap = capability, \ 402 .flags = cap_flags, \ 403 }; \ 404 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 405 int i; \ 406 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 407 i < ARRAY_SIZE(cap.args); i++) { \ 408 cap.args[i] = args_tmp[i]; \ 409 } \ 410 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ 411 }) 412 413 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function, 414 uint32_t index, int reg); 415 416 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len); 417 418 #if !defined(CONFIG_USER_ONLY) 419 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr, 420 hwaddr *phys_addr); 421 #endif 422 423 #endif /* NEED_CPU_H */ 424 425 void kvm_cpu_synchronize_state(CPUState *cpu); 426 void kvm_cpu_synchronize_post_reset(CPUState *cpu); 427 void kvm_cpu_synchronize_post_init(CPUState *cpu); 428 429 /* generic hooks - to be moved/refactored once there are more users */ 430 431 static inline void cpu_synchronize_state(CPUState *cpu) 432 { 433 if (kvm_enabled()) { 434 kvm_cpu_synchronize_state(cpu); 435 } 436 } 437 438 static inline void cpu_synchronize_post_reset(CPUState *cpu) 439 { 440 if (kvm_enabled()) { 441 kvm_cpu_synchronize_post_reset(cpu); 442 } 443 } 444 445 static inline void cpu_synchronize_post_init(CPUState *cpu) 446 { 447 if (kvm_enabled()) { 448 kvm_cpu_synchronize_post_init(cpu); 449 } 450 } 451 452 int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg, PCIDevice *dev); 453 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg, 454 PCIDevice *dev); 455 void kvm_irqchip_release_virq(KVMState *s, int virq); 456 457 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter); 458 459 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 460 EventNotifier *rn, int virq); 461 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n, 462 int virq); 463 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n, 464 EventNotifier *rn, qemu_irq irq); 465 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, 466 qemu_irq irq); 467 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi); 468 void kvm_pc_gsi_handler(void *opaque, int n, int level); 469 void kvm_pc_setup_irq_routing(bool pci_enabled); 470 void kvm_init_irq_routing(KVMState *s); 471 472 /** 473 * kvm_arch_irqchip_create: 474 * @KVMState: The KVMState pointer 475 * 476 * Allow architectures to create an in-kernel irq chip themselves. 477 * 478 * Returns: < 0: error 479 * 0: irq chip was not created 480 * > 0: irq chip was created 481 */ 482 int kvm_arch_irqchip_create(KVMState *s); 483 484 /** 485 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl 486 * @id: The register ID 487 * @source: The pointer to the value to be set. It must point to a variable 488 * of the correct type/size for the register being accessed. 489 * 490 * Returns: 0 on success, or a negative errno on failure. 491 */ 492 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source); 493 494 /** 495 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl 496 * @id: The register ID 497 * @target: The pointer where the value is to be stored. It must point to a 498 * variable of the correct type/size for the register being accessed. 499 * 500 * Returns: 0 on success, or a negative errno on failure. 501 */ 502 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target); 503 #endif 504