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