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 57 #if defined CONFIG_KVM || !defined NEED_CPU_H 58 #define kvm_enabled() (kvm_allowed) 59 /** 60 * kvm_irqchip_in_kernel: 61 * 62 * Returns: true if the user asked us to create an in-kernel 63 * irqchip via the "kernel_irqchip=on" machine option. 64 * What this actually means is architecture and machine model 65 * specific: on PC, for instance, it means that the LAPIC, 66 * IOAPIC and PIT are all in kernel. This function should never 67 * be used from generic target-independent code: use one of the 68 * following functions or some other specific check instead. 69 */ 70 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip) 71 72 /** 73 * kvm_async_interrupts_enabled: 74 * 75 * Returns: true if we can deliver interrupts to KVM 76 * asynchronously (ie by ioctl from any thread at any time) 77 * rather than having to do interrupt delivery synchronously 78 * (where the vcpu must be stopped at a suitable point first). 79 */ 80 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed) 81 82 /** 83 * kvm_halt_in_kernel 84 * 85 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run 86 * inside of kernel space. This only works if MP state is implemented. 87 */ 88 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed) 89 90 /** 91 * kvm_eventfds_enabled: 92 * 93 * Returns: true if we can use eventfds to receive notifications 94 * from a KVM CPU (ie the kernel supports eventds and we are running 95 * with a configuration where it is meaningful to use them). 96 */ 97 #define kvm_eventfds_enabled() (kvm_eventfds_allowed) 98 99 /** 100 * kvm_irqfds_enabled: 101 * 102 * Returns: true if we can use irqfds to inject interrupts into 103 * a KVM CPU (ie the kernel supports irqfds and we are running 104 * with a configuration where it is meaningful to use them). 105 */ 106 #define kvm_irqfds_enabled() (kvm_irqfds_allowed) 107 108 /** 109 * kvm_resamplefds_enabled: 110 * 111 * Returns: true if we can use resamplefds to inject interrupts into 112 * a KVM CPU (ie the kernel supports resamplefds and we are running 113 * with a configuration where it is meaningful to use them). 114 */ 115 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed) 116 117 /** 118 * kvm_msi_via_irqfd_enabled: 119 * 120 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt) 121 * to a KVM CPU via an irqfd. This requires that the kernel supports 122 * this and that we're running in a configuration that permits it. 123 */ 124 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed) 125 126 /** 127 * kvm_gsi_routing_enabled: 128 * 129 * Returns: true if GSI routing is enabled (ie the kernel supports 130 * it and we're running in a configuration that permits it). 131 */ 132 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed) 133 134 /** 135 * kvm_gsi_direct_mapping: 136 * 137 * Returns: true if GSI direct mapping is enabled. 138 */ 139 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping) 140 141 /** 142 * kvm_readonly_mem_enabled: 143 * 144 * Returns: true if KVM readonly memory is enabled (ie the kernel 145 * supports it and we're running in a configuration that permits it). 146 */ 147 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed) 148 149 /** 150 * kvm_direct_msi_enabled: 151 * 152 * Returns: true if KVM allows direct MSI injection. 153 */ 154 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed) 155 156 #else 157 #define kvm_enabled() (0) 158 #define kvm_irqchip_in_kernel() (false) 159 #define kvm_async_interrupts_enabled() (false) 160 #define kvm_halt_in_kernel() (false) 161 #define kvm_eventfds_enabled() (false) 162 #define kvm_irqfds_enabled() (false) 163 #define kvm_resamplefds_enabled() (false) 164 #define kvm_msi_via_irqfd_enabled() (false) 165 #define kvm_gsi_routing_allowed() (false) 166 #define kvm_gsi_direct_mapping() (false) 167 #define kvm_readonly_mem_enabled() (false) 168 #define kvm_direct_msi_enabled() (false) 169 #endif 170 171 struct kvm_run; 172 struct kvm_lapic_state; 173 struct kvm_irq_routing_entry; 174 175 typedef struct KVMCapabilityInfo { 176 const char *name; 177 int value; 178 } KVMCapabilityInfo; 179 180 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP } 181 #define KVM_CAP_LAST_INFO { NULL, 0 } 182 183 struct KVMState; 184 typedef struct KVMState KVMState; 185 extern KVMState *kvm_state; 186 187 /* external API */ 188 189 bool kvm_has_free_slot(MachineState *ms); 190 int kvm_has_sync_mmu(void); 191 int kvm_has_vcpu_events(void); 192 int kvm_has_robust_singlestep(void); 193 int kvm_has_debugregs(void); 194 int kvm_has_pit_state2(void); 195 int kvm_has_many_ioeventfds(void); 196 int kvm_has_gsi_routing(void); 197 int kvm_has_intx_set_mask(void); 198 199 int kvm_init_vcpu(CPUState *cpu); 200 int kvm_cpu_exec(CPUState *cpu); 201 202 #ifdef NEED_CPU_H 203 204 void kvm_setup_guest_memory(void *start, size_t size); 205 void kvm_flush_coalesced_mmio_buffer(void); 206 207 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr, 208 target_ulong len, int type); 209 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr, 210 target_ulong len, int type); 211 void kvm_remove_all_breakpoints(CPUState *cpu); 212 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap); 213 #ifndef _WIN32 214 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset); 215 #endif 216 217 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 218 int kvm_on_sigbus(int code, void *addr); 219 220 /* interface with exec.c */ 221 222 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align)); 223 224 /* internal API */ 225 226 int kvm_ioctl(KVMState *s, int type, ...); 227 228 int kvm_vm_ioctl(KVMState *s, int type, ...); 229 230 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...); 231 232 /** 233 * kvm_device_ioctl - call an ioctl on a kvm device 234 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE 235 * @type: The device-ctrl ioctl number 236 * 237 * Returns: -errno on error, nonnegative on success 238 */ 239 int kvm_device_ioctl(int fd, int type, ...); 240 241 /** 242 * kvm_vm_check_attr - check for existence of a specific vm attribute 243 * @s: The KVMState pointer 244 * @group: the group 245 * @attr: the attribute of that group to query for 246 * 247 * Returns: 1 if the attribute exists 248 * 0 if the attribute either does not exist or if the vm device 249 * interface is unavailable 250 */ 251 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr); 252 253 /** 254 * kvm_device_check_attr - check for existence of a specific device attribute 255 * @fd: The device file descriptor 256 * @group: the group 257 * @attr: the attribute of that group to query for 258 * 259 * Returns: 1 if the attribute exists 260 * 0 if the attribute either does not exist or if the vm device 261 * interface is unavailable 262 */ 263 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr); 264 265 /** 266 * kvm_device_access - set or get value of a specific vm attribute 267 * @fd: The device file descriptor 268 * @group: the group 269 * @attr: the attribute of that group to set or get 270 * @val: pointer to a storage area for the value 271 * @write: true for set and false for get operation 272 * 273 * This function is not allowed to fail. Use kvm_device_check_attr() 274 * in order to check for the availability of optional attributes. 275 */ 276 void kvm_device_access(int fd, int group, uint64_t attr, 277 void *val, bool write); 278 279 /** 280 * kvm_create_device - create a KVM device for the device control API 281 * @KVMState: The KVMState pointer 282 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the 283 * kernel source) 284 * @test: If true, only test if device can be created, but don't actually 285 * create the device. 286 * 287 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd; 288 */ 289 int kvm_create_device(KVMState *s, uint64_t type, bool test); 290 291 292 /* Arch specific hooks */ 293 294 extern const KVMCapabilityInfo kvm_arch_required_capabilities[]; 295 296 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run); 297 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run); 298 299 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run); 300 301 int kvm_arch_process_async_events(CPUState *cpu); 302 303 int kvm_arch_get_registers(CPUState *cpu); 304 305 /* state subset only touched by the VCPU itself during runtime */ 306 #define KVM_PUT_RUNTIME_STATE 1 307 /* state subset modified during VCPU reset */ 308 #define KVM_PUT_RESET_STATE 2 309 /* full state set, modified during initialization or on vmload */ 310 #define KVM_PUT_FULL_STATE 3 311 312 int kvm_arch_put_registers(CPUState *cpu, int level); 313 314 int kvm_arch_init(MachineState *ms, KVMState *s); 315 316 int kvm_arch_init_vcpu(CPUState *cpu); 317 318 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */ 319 unsigned long kvm_arch_vcpu_id(CPUState *cpu); 320 321 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 322 int kvm_arch_on_sigbus(int code, void *addr); 323 324 void kvm_arch_init_irq_routing(KVMState *s); 325 326 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route, 327 uint64_t address, uint32_t data, PCIDevice *dev); 328 329 int kvm_arch_msi_data_to_gsi(uint32_t data); 330 331 int kvm_set_irq(KVMState *s, int irq, int level); 332 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg); 333 334 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin); 335 void kvm_irqchip_commit_routes(KVMState *s); 336 337 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 338 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 339 340 struct kvm_guest_debug; 341 struct kvm_debug_exit_arch; 342 343 struct kvm_sw_breakpoint { 344 target_ulong pc; 345 target_ulong saved_insn; 346 int use_count; 347 QTAILQ_ENTRY(kvm_sw_breakpoint) entry; 348 }; 349 350 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint); 351 352 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu, 353 target_ulong pc); 354 355 int kvm_sw_breakpoints_active(CPUState *cpu); 356 357 int kvm_arch_insert_sw_breakpoint(CPUState *cpu, 358 struct kvm_sw_breakpoint *bp); 359 int kvm_arch_remove_sw_breakpoint(CPUState *cpu, 360 struct kvm_sw_breakpoint *bp); 361 int kvm_arch_insert_hw_breakpoint(target_ulong addr, 362 target_ulong len, int type); 363 int kvm_arch_remove_hw_breakpoint(target_ulong addr, 364 target_ulong len, int type); 365 void kvm_arch_remove_all_hw_breakpoints(void); 366 367 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg); 368 369 bool kvm_arch_stop_on_emulation_error(CPUState *cpu); 370 371 int kvm_check_extension(KVMState *s, unsigned int extension); 372 373 int kvm_vm_check_extension(KVMState *s, unsigned int extension); 374 375 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \ 376 ({ \ 377 struct kvm_enable_cap cap = { \ 378 .cap = capability, \ 379 .flags = cap_flags, \ 380 }; \ 381 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 382 int i; \ 383 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 384 i < ARRAY_SIZE(cap.args); i++) { \ 385 cap.args[i] = args_tmp[i]; \ 386 } \ 387 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ 388 }) 389 390 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \ 391 ({ \ 392 struct kvm_enable_cap cap = { \ 393 .cap = capability, \ 394 .flags = cap_flags, \ 395 }; \ 396 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 397 int i; \ 398 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \ 399 i < ARRAY_SIZE(cap.args); i++) { \ 400 cap.args[i] = args_tmp[i]; \ 401 } \ 402 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ 403 }) 404 405 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function, 406 uint32_t index, int reg); 407 408 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len); 409 410 #if !defined(CONFIG_USER_ONLY) 411 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr, 412 hwaddr *phys_addr); 413 #endif 414 415 #endif /* NEED_CPU_H */ 416 417 void kvm_cpu_synchronize_state(CPUState *cpu); 418 void kvm_cpu_synchronize_post_reset(CPUState *cpu); 419 void kvm_cpu_synchronize_post_init(CPUState *cpu); 420 void kvm_cpu_clean_state(CPUState *cpu); 421 422 /* generic hooks - to be moved/refactored once there are more users */ 423 424 static inline void cpu_synchronize_state(CPUState *cpu) 425 { 426 if (kvm_enabled()) { 427 kvm_cpu_synchronize_state(cpu); 428 } 429 } 430 431 static inline void cpu_synchronize_post_reset(CPUState *cpu) 432 { 433 if (kvm_enabled()) { 434 kvm_cpu_synchronize_post_reset(cpu); 435 } 436 } 437 438 static inline void cpu_synchronize_post_init(CPUState *cpu) 439 { 440 if (kvm_enabled()) { 441 kvm_cpu_synchronize_post_init(cpu); 442 } 443 } 444 445 static inline void cpu_clean_state(CPUState *cpu) 446 { 447 if (kvm_enabled()) { 448 kvm_cpu_clean_state(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