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