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