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